What is Microsoft Discovery and how does it differ from Copilot Studio?

Microsoft Discovery is a purpose-built agent platform designed specifically for research and development workflows, not a general-purpose copilot tool. Unlike Copilot Studio, it comes with domain-specific primitives like scientific data structures and simulation orchestration built in, so teams spend far less time engineering workarounds for R&D-specific tasks.

How does Microsoft Discovery agentic RD improve research and development workflows?

Because the platform already understands R&D concepts like hypothesis tracking, candidate scoring, and simulation runs, agents do not need lengthy prompts or custom-built logic to handle them. This reduces build time significantly compared to trying to model the same workflows on a generic agent platform.

When should I choose a domain-specific agent platform over a generic one like Copilot Studio?

A domain-specific platform makes sense when your workflows map closely to the vertical it was designed for, since the built-in primitives cut build time and reduce fragility. If your use case is too broad or does not fit the platform shape, a general-purpose tool with custom configuration will give you more flexibility.

Why do generic agentic workflows often fail for complex enterprise use cases?

General-purpose platforms require teams to manually encode domain logic through prompts, tool definitions, and automation flows, which produces brittle systems that are hard to maintain. When the platform has no native understanding of the domain, complexity accumulates quickly and the resulting agent is difficult to scale or debug.

Which is better, Claude vs ChatGPT for automation workflows?

Choosing between Claude and ChatGPT for automation is less about which model performs better in general benchmarks and more about how each behaves inside a pipeline. The criteria that matter are structured output reliability, tool-calling accuracy, and how well the model holds instructions as context grows. Testing both models against your specific workflow conditions will tell you far more than any consumer comparison.

Why does my AI agent start producing errors after working fine at first?

This often happens because the model experiences instruction drift as context accumulates over time. Long flows gather user inputs, tool results, and retrieved documents, and some models struggle to maintain consistent behaviour at high token counts. Testing your model under realistic context lengths before going to production can help catch this early.

How do I choose an AI model for a Power Automate or Copilot Studio flow?

Focus on how the model handles structured outputs, selects the right tools when multiple connectors are available, and behaves when context is long rather than short. Consumer benchmarks test models in isolation, but real automation pipelines require consistent, predictable behaviour across repeated calls with varying inputs. Running your own tests against your actual schema and tools will give you more reliable answers.

What causes silent failures in AI automation workflows?

One of the most common causes is inconsistent structured output, where a model occasionally adds unexpected formatting, drops required fields, or wraps a response in a code block instead of returning clean JSON. These errors can pass through without triggering obvious alerts while still breaking downstream actions. Testing output stability across many varied inputs is one of the most important steps before deploying a model-powered flow.

How do I set up a Power Platform agent GitHub integration in an enterprise environment?

Start by mapping which Power Platform environments need to connect to which GitHub organisations, and what DLP policies apply to each. You cannot assume a connection that works in Dev or UAT will behave the same way in Production, since connector classifications can differ across environments. Sorting out service account credentials and token rotation policies before you build will save significant rework later.

Why does my Power Platform agent work in UAT but fail in Production when connecting to GitHub?

The most common cause is a difference in DLP policies between environments. The GitHub connector may be classified as allowed in your UAT environment but blocked or restricted in Production, which stops the connection from loading at all. The agent will often still respond but return empty results, making the failure easy to miss until users report it.

What is environment sprawl and why does it matter for Power Platform agent deployments?

Environment sprawl refers to the accumulation of multiple Power Platform environments across an organisation, each with its own DLP rules, security posture, and connector permissions. It matters for agent deployments because a GitHub connection that is permitted in one environment may be completely blocked in another, and most teams do not map these differences before they start building.

When should I use a service account instead of personal authentication for a GitHub connector in Power Platform?

Any time the agent is intended for a team or production use case, a service account is the right choice over personal authentication. Personal credentials tie the connection to an individual user, which creates access and continuity risks when that person changes roles or leaves the organisation. A shared service account also makes token rotation and permission auditing much easier to manage.

Why does agentic workflow latency get so bad in multi-step agents?

Each individual call in an agent loop may seem fast, but the overhead adds up across 8 to 15 round trips per user request. Connection setup, authentication, and token streaming costs stack on every single step, turning individually acceptable delays into a frustrating overall wait time.

What is connection-scoped caching and how does it help agent performance?

Connection-scoped caching keeps expensive context like system prompts, tool definitions, and prior conversation state ready across multiple calls instead of reprocessing it each time. This avoids redundant work on every step of an agent loop and significantly reduces the overhead that accumulates across a multi-turn interaction.

How do I reduce latency in Copilot Studio and Power Automate agents?

Start by auditing how many round trips your agent actually makes for a single user request, since this is where most hidden latency lives. Look for opportunities to batch tool calls, reduce unnecessary steps in your loop, and watch for platform-level improvements like persistent connections that reduce per-call overhead.

Why do users abandon AI agents even when the agent gives correct answers?

If the response takes too long, users lose confidence and revert to familiar alternatives before the agent finishes. Technical correctness does not matter if the experience feels slow enough to suggest something has gone wrong.

How do I get started with Copilot Studio in 2026 without wasting time on outdated tutorials?

Skip the classic topic-and-trigger-phrase approach and go straight to building a generative orchestration agent. Focus on writing clear instructions, connecting knowledge sources, and designing reliable tools rather than learning the node-based dialog editor.

What is generative orchestration in Copilot Studio and why does it matter?

Generative orchestration is now the default approach for new agents in Copilot Studio, where the agent uses AI to handle conversation flow rather than following a rigid dialog tree. It reduces the need to manually write dozens of trigger phrases and scales far better than the older topic-based model.

Why does my Copilot Studio agent feel rigid and break when users ask unexpected questions?

This usually happens when the agent is built around a topic tree with trigger phrases, which only handles the exact variations you anticipated. Switching to a generative orchestration approach with well-written instructions and grounded knowledge sources makes the agent much more flexible.

When should I still use topics in Copilot Studio?

Topics are still useful when a specific flow must follow a strict, predictable script, such as a regulated process or a form-style interaction with no variation. They work best as guardrails for those edge cases rather than as the foundation of the entire agent.

When should I use Copilot Studio instead of another tool?

Copilot Studio works best when the interaction is genuinely conversational, scoped to a single domain, and involves a defined set of intents. If the task is transactional, point-in-time, or better served by a simple form or search interface, tools like Power Automate or Power Apps are likely a faster and more maintainable choice.

What is the difference between Copilot Studio and Power Automate?

Power Automate is built for workflow and process automation, such as form submissions, approvals, and scheduled triggers. Copilot Studio is designed for conversational agent experiences. Using Copilot Studio for single-action tasks adds unnecessary complexity without improving the user experience.

Why does my Copilot Studio agent keep routing users to the wrong topic?

Topic routing breaks down when an agent is built to handle too many domains or intents within a single session. When a user's phrasing falls between two topics, the agent will confidently pick one and get it wrong. Keeping each agent focused on a narrow scope reduces these edge cases and makes failures easier to diagnose.

How do I know if my use case actually needs a chatbot?

Start by defining what the interaction needs to do before choosing a tool. If users need a back-and-forth conversation to complete a task, a conversational agent may be appropriate. If they need a search result, a status update, or a simple action, a canvas app or improved search interface will often deliver a better outcome in less time.

Why does my Copilot Studio agent testing pass in demos but fail in production?

Most Copilot Studio agent testing is built around ideal user inputs and predictable conversation paths, which do not reflect how real users actually behave. In production, users ask incomplete questions, switch intent mid-conversation, and use the agent in unintended ways that no demo ever surfaces. Testing needs to go beyond the happy path to catch these edge cases before go-live.

What are the most common failure modes in Copilot Studio agents?

The three patterns that appear most often are topic misrouting when user phrasing falls between two intents, action failures that complete silently without triggering any error or fallback, and prompt instructions that break down when they encounter messy real-world data. Each of these can go undetected in testing because they only emerge under realistic conditions.

How do I know if a Power Automate action failed inside my Copilot Studio agent?

Silent action failures are a serious risk because the agent can continue the conversation and imply a task completed when it did not. You need explicit error handling and fallback logic in your flows so that failures surface to the user rather than being masked by a confident-sounding response.

When should I test my Copilot Studio agent against real production data?

You should test against realistic data before promotion to production, not after. System prompts written against clean test data can behave unpredictably when they encounter unexpected characters, mixed languages, or long strings that only appear in live environments. Incorporating a sample of real or representative data into your test suite is a necessary step before sign-off.

Tag: Copilot Studio

Microsoft Discovery Is the First Real Glimpse of Domain-Specific Agent Platforms
I came across the Azure Blog post about Microsoft Discovery expanding its preview, and it crystallised something I have been chewing on for months. Most enterprise AI conversation right now is stuck on horizontal agents. Generic copilots doing generic things across generic data. Microsoft Discovery agentic RD goes the other direction, and that direction is where the interesting architectural decisions are about to happen.

What Microsoft Discovery Actually Is If You Skip the Marketing

Strip the announcement language away and Discovery is a vertical agent platform shaped specifically for research and development workflows. It is not a chatbot. It is not Copilot Studio with a science skin. It is a purpose-built layer with domain primitives baked in: scientific data structures, simulation orchestration, multi-agent coordination tuned for R&D problems instead of generic enterprise tasks.

The important word is shape. A horizontal agent platform gives you a blank canvas and a set of generic tools. A domain-shaped platform gives you a canvas where the grid lines already match the work. You give up flexibility. You gain a tenth of the build time when the shape fits.

Why Domain-Shaped Agent Platforms Beat Generic Copilots for R&D Workflows

I have written before about how most agentic workflows are just fancy if/then logic in a trench coat. The reason is almost always the same. Teams use a general-purpose tool to model a domain it does not understand, then spend weeks bolting domain logic on top through prompts and tool definitions.

R&D is the perfect example. A real research workflow involves hypothesis tracking, simulation runs, candidate scoring, lineage of why a decision was made three steps ago. None of that is native to a generic Copilot Studio agent. You can build it. I have seen people try. It ends up as a fragile stack of topics, variables, and Power Automate flows pretending to be a state machine.

A domain-shaped platform encodes those primitives directly. The agent does not need a 4000-token system prompt explaining what a candidate molecule is, because the platform already knows. That is the productivity unlock, and it is also why I think we are about to see a lot more of these.

How This Changes the Build vs Buy Decision for Power Platform Teams

Here is the part Power Platform people should pay attention to. The skill that matters going forward is not how well you can build in Power Automate or Copilot Studio. It is picking the right altitude for the automation in front of you.

I keep seeing teams default to building everything in Copilot Studio because that is the tool they know. Someone wants a research assistant. Someone wants a contract review agent. Someone wants a finance close helper. All of it gets crammed into Copilot Studio topics and custom connectors, and six months later the build is brittle, slow, and three people deep in technical debt. If you are just getting started, getting started with Copilot Studio in 2026 means skipping the chatbot tutorials entirely and learning to think in terms of orchestration first.

The decision tree is going to look more like this:
- Is there a domain-shaped platform that already models this work? Use it. Customise on top.
- Is the workflow generic but cross-system? Copilot Studio agent with deterministic Power Automate flows underneath.
- Is the workflow narrow, predictable, high volume? Raw Power Automate. No agent. No reasoning layer. Just a flow.
- Is the workflow heavy on judgment with messy unstructured inputs? Reasoning model in the orchestration layer, not the response layer. I covered this in my post on Claude as orchestration brain.
Picking the wrong altitude is the most expensive mistake I see. Discovery is interesting precisely because it adds a new altitude that did not exist in the Microsoft stack before. R&D teams who would have been forced into Copilot Studio now have a layer that fits their work natively.

What I Would Watch For Next in the Microsoft Agent Stack

Discovery is the canary. R&D is just the first vertical because Microsoft has obvious customers there and the workflows are well-understood. The pattern will repeat. I would expect domain-shaped agent layers for clinical workflows, manufacturing operations, financial close, regulatory review. Each one will sit above the general-purpose Copilot stack and offer the same trade: less flexibility, much faster time to a working system.

The thing I am watching is interoperability. Can a domain platform like Discovery call out to a Copilot Studio agent for a side task? Can a Power Automate flow trigger a Discovery workflow? If yes, the stack becomes composable and the architectural decisions get genuinely interesting. If no, we end up with another round of silos with their own latency problems and integration debt.

For now, the practical move is to stop treating Copilot Studio as the universal hammer. In my experience, the teams who consistently ship working automations are the ones who match the tool to the shape of the work. Discovery just made that decision a little more interesting.

Frequently Asked Questions

What is Microsoft Discovery and how does it differ from Copilot Studio?

Microsoft Discovery is a purpose-built agent platform designed specifically for research and development workflows, not a general-purpose copilot tool. Unlike Copilot Studio, it comes with domain-specific primitives like scientific data structures and simulation orchestration built in, so teams spend far less time engineering workarounds for R&D-specific tasks.

How does Microsoft Discovery agentic RD improve research and development workflows?

Because the platform already understands R&D concepts like hypothesis tracking, candidate scoring, and simulation runs, agents do not need lengthy prompts or custom-built logic to handle them. This reduces build time significantly compared to trying to model the same workflows on a generic agent platform.

When should I choose a domain-specific agent platform over a generic one like Copilot Studio?

A domain-specific platform makes sense when your workflows map closely to the vertical it was designed for, since the built-in primitives cut build time and reduce fragility. If your use case is too broad or does not fit the platform shape, a general-purpose tool with custom configuration will give you more flexibility.

Why do generic agentic workflows often fail for complex enterprise use cases?

General-purpose platforms require teams to manually encode domain logic through prompts, tool definitions, and automation flows, which produces brittle systems that are hard to maintain. When the platform has no native understanding of the domain, complexity accumulates quickly and the resulting agent is difficult to scale or debug.

Source: Microsoft Discovery: Advancing agentic R&D at scale (Azure Blog).
April 25, 2026
Claude vs ChatGPT Is the Wrong Question When You Are Building Automations

Another Claude vs ChatGPT comparison landed in my feed this week. I came across a piece on the Zapier Blog running the usual head-to-head: reasoning, coding, writing, ethical dilemmas. Useful if you are picking a chat assistant for personal use. Almost useless if you are deciding claude vs chatgpt for automation inside a real enterprise flow.

I keep seeing people pick a model based on a consumer benchmark and then act confused when their Copilot Studio agent starts returning malformed JSON in week three. The criteria that matter when a model sits behind a connector are not the criteria that make for a good blog post.

Why Head to Head Model Comparisons Stop Being Useful the Moment You Add a Connector

Consumer comparisons test the model in isolation. One prompt in, one answer out, a human judges the output. That setup tells you nothing about what happens when the model has to call a tool, parse a response, call another tool, and feed a structured result into a downstream action.

Inside an automation, the model is not the product. The model is one component in a pipeline. The question is not which one writes better poetry. The question is which one fails in ways your orchestration layer can actually handle.

I wrote about this angle in a previous post on agentic workflows. The LLM is the reasoning layer, not the agent. Picking the reasoning layer on vibes from a consumer benchmark is how you end up with a beautifully worded confident response for a task that never completed.

The Four Things That Actually Matter When a Model Sits Inside an Automation

These are what I actually test for. None of them show up in head-to-head comparisons.

Structured output stability under load. Ask the same model for the same JSON schema a hundred times with slightly different inputs. Count how often it adds a trailing comma, drops a required field, wraps the JSON in a code fence, or decides today is the day to add a helpful explanation before the response. This is the single biggest source of silent failures I see in production.

Tool-calling predictability with multiple connectors. Give the model five tools. Watch how it picks. A model that is 95 percent accurate on tool selection with two tools can drop to 70 percent with five because the descriptions start competing. Consumer tests never measure this.

Behaviour when context gets long. Most real flows accumulate context: user input, previous tool results, system instructions, retrieved documents. I want to know how the model behaves at 40k tokens of accumulated state, not at 500. Instruction drift usually shows up here first.

Pricing behaviour under loops. An agent that retries three times on a failed tool call can quietly 10x your cost. The cheaper model on paper is not always the cheaper model in production once you account for retry patterns and token accumulation. Latency Is the Quiet Killer of Agentic Workflows covers how round-trip costs compound in ways most people never budget for until it is too late.

How I Pick Between Claude and GPT for a Specific Flow

I do not pick a model for the whole platform. I pick per use case.

For long-context reasoning where the model needs to hold a lot of state and follow detailed instructions without drifting, Claude has been the more predictable option in my testing. Fewer surprise deviations from the system prompt when the context gets messy. If you want to go deeper on why Claude works well as a reasoning layer inside enterprise pipelines, Claude as an Orchestration Brain Is the Most Interesting Thing Happening in Enterprise AI Right Now gets into the architecture side of that decision.

For fast, cheap, high-volume classification or extraction where the schema is simple and the input is short, GPT models tend to win on cost-per-call and latency. If the task is “read this email and return one of five categories,” I am not paying for a heavyweight reasoning model.

For tool-calling inside a Copilot Studio agent with multiple Power Automate actions, I test both. There is no universal winner. It depends on how the tool descriptions are written, how many there are, and how ambiguous the user input gets.

The honest answer most of the time is: it does not matter as much as the people arguing about it think it does. The bigger wins come from tool design, prompt structure, and failure handling. A well-designed flow with a mid-tier model beats a sloppy flow with the flagship every time.

What to Test Before You Commit a Model to Production

Before a model goes behind a production flow, I run four checks. Not benchmarks. Checks against the actual flow.

Run the real schema a hundred times with production-like inputs. Measure malformed output rate. Anything above one percent and you need a validation and retry layer, no matter which model you picked.

Run the tool-calling logic with the real connector set, not a simplified test set. Watch for the model picking the wrong tool when two descriptions overlap. This is where I lost the most time the hard way.

Simulate a long session. Feed it accumulated context that looks like a real user journey, not a single clean turn. Watch for instruction drift.

Load test with the pricing model in mind. Know what a retry storm costs you before it happens in production, not after finance asks questions. The Power Automate documentation covers retry policies, but most people never configure them until something breaks.

The Claude vs ChatGPT question is the wrong frame. The right question is: which model handles the specific shape of failure my flow is most exposed to. Answer that and the comparison stops mattering. That is the part I keep trying to explain when people ask me, and it still gets pushed aside for whichever model topped a benchmark last week.

Frequently Asked Questions

Which is better, Claude vs ChatGPT for automation workflows?

Choosing between Claude and ChatGPT for automation is less about which model performs better in general benchmarks and more about how each behaves inside a pipeline. The criteria that matter are structured output reliability, tool-calling accuracy, and how well the model holds instructions as context grows. Testing both models against your specific workflow conditions will tell you far more than any consumer comparison.

Why does my AI agent start producing errors after working fine at first?

This often happens because the model experiences instruction drift as context accumulates over time. Long flows gather user inputs, tool results, and retrieved documents, and some models struggle to maintain consistent behaviour at high token counts. Testing your model under realistic context lengths before going to production can help catch this early.

How do I choose an AI model for a Power Automate or Copilot Studio flow?

Focus on how the model handles structured outputs, selects the right tools when multiple connectors are available, and behaves when context is long rather than short. Consumer benchmarks test models in isolation, but real automation pipelines require consistent, predictable behaviour across repeated calls with varying inputs. Running your own tests against your actual schema and tools will give you more reliable answers.

What causes silent failures in AI automation workflows?

One of the most common causes is inconsistent structured output, where a model occasionally adds unexpected formatting, drops required fields, or wraps a response in a code block instead of returning clean JSON. These errors can pass through without triggering obvious alerts while still breaking downstream actions. Testing output stability across many varied inputs is one of the most important steps before deploying a model-powered flow.

Source: Claude vs. ChatGPT: Which is best? [2026] (Zapier Blog).

April 23, 2026
Power Platform Agents Talking to GitHub Sounds Simple Until You Hit Enterprise Environment Sprawl

I keep seeing the same demo on LinkedIn. Someone wires a Power Platform agent to GitHub in ten minutes, the agent answers questions about a repo, everyone claps. Then a team at a real enterprise tries to copy the pattern and stalls for three weeks. The problem is never the connector. The problem is power platform agent github integration enterprise reality, where you have twelve environments, three GitHub orgs, DLP policies that differ per environment, and a security team that wants tokens rotated quarterly.

The connector works. The environment strategy around it does not.

Why the GitHub Connector Demo Lies to You

In a demo, one person builds in their default environment, authenticates with their own GitHub account, and points the agent at a public or personal repo. Every layer of that setup is the easy path.

Personal auth hides the service account problem. Default environment hides the DLP problem. A single repo hides the org sprawl problem. Running it locally hides the fact that the Prod environment sits in a different security posture entirely.

I have watched teams ship an agent to UAT, watch it work, promote it to Prod, and then hit a wall where the connection simply cannot see the repo. No error message that is useful. Just empty knowledge and a confident agent saying it could not find the information. Same failure mode I wrote about in agent testing versus production behavior. The agent answers. The answer is empty. Nobody notices for a week.

The Environment Sprawl Problem Nobody Scopes For

Large enterprises do not have one Power Platform environment. They have Dev, multiple UAT environments per business unit, a shared services environment, regional Prod environments, a sandbox for citizen devs, and a couple of orphaned ones nobody wants to delete. Each one has its own DLP policy group at the tenant level.

GitHub is not one thing either. Most large orgs I hear about run at least two GitHub organisations. Sometimes an enterprise account with several orgs under it. One for platform code, one for product code, maybe one for security-sensitive work behind SSO with stricter SAML enforcement.

Now ask the real question. Which Power Platform environment is allowed to talk to which GitHub org, with which connector classification, using which identity? Most teams have never drawn this map. They build in Dev, it works, and they assume Prod behaves the same. It does not.

The GitHub connector can sit in the Business data group in one environment and the Blocked group in another. The agent that passed UAT will not even load the connection in Prod because the policy blocks it. You find out at go-live.

How I Would Map Power Platform Environments to GitHub Orgs and Repos

Start with the principle that environment design is the hard part, not the connector. Then draw the map before you build anything.

Dev Power Platform environment talks to a Dev GitHub org, or a dedicated sandbox org, with a scoped token and loose DLP. UAT talks to the same repos as Prod but through a read-only identity, so you are testing against real structure without write risk. Prod talks to Prod GitHub orgs through a managed service identity, with the connector in the Business data group and DLP exceptions documented.

The knowledge source in the agent has to point at a repo that the target environment’s connection can actually see. This is where most builds break. The agent was built in Dev pointing at a repo in the Dev org. In Prod, that repo does not exist, and the Prod connection has no permission on the Prod equivalent. The fix is not technical. It is an environment variable pattern where the repo reference is parameterised per environment, and solution deployment swaps the value.

The Microsoft Learn docs on environment strategy cover the platform side. They do not cover the mapping to external orgs. That part is on you.

Service Accounts Tokens and the Stuff That Actually Breaks in Prod

Personal Access Tokens are how every demo works and how no enterprise should run anything. The person who created the PAT leaves the company, the token is revoked, the agent goes dark. I have seen this happen. Twice.

GitHub Apps are the right answer for Prod. Fine-grained permissions, installable per org, rotate credentials without losing the identity. The connector supports GitHub App auth. Use it. The trade-off is setup time. You have to get the security team to approve the app installation on the target org, which takes weeks in a large enterprise. Plan for that before you commit to a go-live date.

Service account seats are the other thing that breaks quietly. The identity your Prod connection uses needs a seat on the target repo. In a GitHub Enterprise plan with seat limits, this is a budget conversation, not a technical one. I have seen agent deployments stall because nobody wanted to pay for an extra seat.

Token rotation policy is the last piece. If your security team rotates every ninety days, build the rotation into your deployment pipeline, not into a calendar reminder. Otherwise the agent fails silently on day ninety-one and the confident-but-empty response problem shows up again. And if those silent failures start compounding across chained steps, you are looking at the kind of agentic workflow latency problem that is easy to miss until it is already affecting users.

The connector is not the hard part. It never was. The teams that succeed stop treating integration as a connector problem and start treating it as an environment design problem. If you are still early in this process, getting started with Copilot Studio in 2026 means thinking through environment strategy from day one, not after your first failed Prod deployment. Draw the map first. Build second.

Frequently Asked Questions

How do I set up a Power Platform agent GitHub integration in an enterprise environment?

Start by mapping which Power Platform environments need to connect to which GitHub organisations, and what DLP policies apply to each. You cannot assume a connection that works in Dev or UAT will behave the same way in Production, since connector classifications can differ across environments. Sorting out service account credentials and token rotation policies before you build will save significant rework later.

Why does my Power Platform agent work in UAT but fail in Production when connecting to GitHub?

The most common cause is a difference in DLP policies between environments. The GitHub connector may be classified as allowed in your UAT environment but blocked or restricted in Production, which stops the connection from loading at all. The agent will often still respond but return empty results, making the failure easy to miss until users report it.

What is environment sprawl and why does it matter for Power Platform agent deployments?

Environment sprawl refers to the accumulation of multiple Power Platform environments across an organisation, each with its own DLP rules, security posture, and connector permissions. It matters for agent deployments because a GitHub connection that is permitted in one environment may be completely blocked in another, and most teams do not map these differences before they start building.

When should I use a service account instead of personal authentication for a GitHub connector in Power Platform?

Any time the agent is intended for a team or production use case, a service account is the right choice over personal authentication. Personal credentials tie the connection to an individual user, which creates access and continuity risks when that person changes roles or leaves the organisation. A shared service account also makes token rotation and permission auditing much easier to manage.

April 23, 2026
Latency Is the Quiet Killer of Agentic Workflows and Almost Nobody Talks About It
Everyone obsesses over model quality, tool design, and prompt structure when building agents. The thing that actually kills adoption in production is something else entirely. Agentic workflow latency is the quiet killer, and most Power Platform and Copilot Studio builders are not thinking about it until users start abandoning the tool.

I came across a post from OpenAI about using WebSockets and connection-scoped caching in the Responses API to speed up their Codex loop. It confirmed something I keep running into building multi-step agents internally. The math is brutal once you do it honestly.

Why Agent Loops Feel Slow Even When Each Call Is Fast

A single model call at 800ms feels fine. A tool call at 300ms feels fine. A Dataverse lookup at 500ms feels fine. Everyone looks at these numbers in isolation and says the platform is fast enough.

Then you build an actual agent. It reasons, calls a tool, reads the result, reasons again, calls another tool, checks a condition, calls a third tool, summarises, responds. That is 8 to 15 round trips for one user request. Each round trip carries connection setup, authentication overhead, token streaming setup, and the model’s own time to first token.

A 400ms overhead per call sounds small. Multiply by 12 calls. That is almost 5 seconds of pure overhead before any actual thinking or work happens. Users do not wait 15 seconds for a confident answer. They ask once, get nothing for a few seconds, and switch back to the old way of doing it.

I have watched this kill internal tools that were technically correct. The agent did the right thing. Nobody used it.

What OpenAI Just Shipped and Why It Matters Beyond Codex

The short version of what they did: move from repeated HTTP requests to a persistent WebSocket connection, and keep cache state scoped to that connection so repeat context does not need to be re-processed on every turn.

This is not a Codex-only trick. It is a general pattern. Connection-scoped caching means the expensive part of a call, the part that handles your system prompt and tool definitions and prior context, does not get redone from scratch every time your agent takes another step.

For anyone building agents that loop, this is the shape of the next year of infrastructure work. The platforms that expose this properly will feel instant. The ones that do not will feel like they are thinking through molasses.

What This Looks Like Inside Copilot Studio and Power Automate

Here is where it gets uncomfortable. In Copilot Studio, you do not see the round trips. You see a topic, a few actions, a generative answer node. The platform hides every call behind its own orchestration.

That hiding is the problem. A Copilot Studio agent doing generative orchestration with three tool calls backed by Power Automate flows is making far more round trips than most builders realise. Each tool call is a Power Automate HTTP trigger plus whatever that flow does internally, often including another connector call to SharePoint, Dataverse, or an external API. The agent then reads the response and decides what to do next, which is another model call. And if you are hitting Power Automate throttling limits under real load, every one of those round trips gets longer.

I built one recently that felt snappy in testing with one user. In production with ten concurrent sessions, response times doubled. Nothing in the flow was slow on its own. The sum was slow, and throttling on shared connectors made it worse. This is the same class of problem I wrote about in Most Agentic Workflows Are Just Fancy If/Then Logic in a Trench Coat. The difference between a real agent loop and a glorified flow shows up in latency first.

How I Would Budget Latency Before I Build the Agent

I treat latency as a first-class design constraint now, not something I measure after the fact. Before I build, I do this:
- Estimate the number of model calls per user request. Not best case. Typical case.
- Estimate the number of tool calls and what each one hits. A SharePoint list call in the same tenant is not a Graph API call with auth handshake.
- Set a budget. I aim for under 4 seconds total for anything conversational, under 10 seconds for anything that is clearly doing work.
- Cut calls aggressively. Can two tools be one? Can I pre-fetch context in a single call instead of three? Can the agent skip a reasoning step when the intent is obvious?
- Parallelise where I can. Power Automate lets you run actions in parallel branches. Most builders do not use them.
The other thing I stopped doing: chaining LLM calls for steps that do not need reasoning. If a step is deterministic, I call the tool directly, not through the model. Every model call I can remove from the loop gives me back 500 to 1500ms.

Latency is also where the question of who owns the decision in an agentic workflow becomes a performance problem, not just a governance one. Every checkpoint that routes back to a human approver adds another wait state to the loop. The more of those you have, the more your total response time is dominated by human latency, not model latency.

I have written more about my approach to this kind of trade-off on my LinkedIn, because I keep having the same conversation with people at other organisations who hit the wall when their demo hits real users.

The agents that win in production are not the smartest ones. They are the ones that answer before the user gives up.

Frequently Asked Questions

Why does agentic workflow latency get so bad in multi-step agents?

Each individual call in an agent loop may seem fast, but the overhead adds up across 8 to 15 round trips per user request. Connection setup, authentication, and token streaming costs stack on every single step, turning individually acceptable delays into a frustrating overall wait time.

What is connection-scoped caching and how does it help agent performance?

Connection-scoped caching keeps expensive context like system prompts, tool definitions, and prior conversation state ready across multiple calls instead of reprocessing it each time. This avoids redundant work on every step of an agent loop and significantly reduces the overhead that accumulates across a multi-turn interaction.

How do I reduce latency in Copilot Studio and Power Automate agents?

Start by auditing how many round trips your agent actually makes for a single user request, since this is where most hidden latency lives. Look for opportunities to batch tool calls, reduce unnecessary steps in your loop, and watch for platform-level improvements like persistent connections that reduce per-call overhead.

Why do users abandon AI agents even when the agent gives correct answers?

If the response takes too long, users lose confidence and revert to familiar alternatives before the agent finishes. Technical correctness does not matter if the experience feels slow enough to suggest something has gone wrong.

Source: Speeding up agentic workflows with WebSockets in the Responses API (OpenAI).
April 22, 2026
If You Are Starting Copilot Studio in 2026 Skip the Chatbot Tutorials

I keep running into people getting started with Copilot Studio in 2026 who spent their first two weeks building a topic tree with trigger phrases and a node-by-node dialog flow. Then they ask me why their agent feels rigid and why every new question breaks something. The answer is not that they built it badly. They learned the wrong paradigm from tutorials written three years ago.

If you are starting today, skip the chatbot tutorials. Go straight to a generative orchestration agent with a couple of tools and knowledge sources. That is how real agents are being built now.

Why the 2023 Copilot Studio learning path is actively misleading in 2026

The classic learning path walks you through creating a topic, adding trigger phrases, building a dialog tree with question nodes, conditions, and variables, then testing it with the exact phrases you wrote. It feels productive because you see progress quickly. It is also teaching you a way of thinking that does not scale past an FAQ bot.

Microsoft itself has de-emphasised this path. Generative orchestration is the default for new agents. Topics still exist and still have their place, but as guardrails for specific flows that must follow a strict script, not as the primary way to build conversation.

The problem with learning topics first is unlearning them later. People who start with trigger phrases end up writing dozens of them trying to catch every variation a user might type, then wonder why the agent still misroutes at the edges. Your Copilot Studio Agent Passed Every Test and Still Failed in Production explores exactly this kind of failure. Generative orchestration does not eliminate this, but it changes the shape of the problem entirely.

The four things that actually matter when you start today

When someone joins the team and asks where to begin with Copilot Studio, I tell them to focus on four things before anything else.

Instructions. The system prompt is where most of the agent’s behaviour comes from now. Learning to write clear, specific, scoped instructions is more valuable than learning the node editor. Bad instructions cause instruction drift the moment real data hits the agent.

Knowledge sources. Connecting a SharePoint site, a Dataverse table, or a public website and understanding how the agent grounds answers in them. This is where the actual information lives. Most agents do not need a dialog tree, they need good grounding.

Tools. A tool is a Power Automate flow, a connector action, an MCP server, or a prompt. Learning how to design tools so the agent can call them reliably is the real skill. Returning a status of done is not enough. The agent needs structured output it can reason about.

Orchestration behaviour. Understanding that the LLM picks which tool to call, in what order, and how to handle the result. You are not drawing the flow. You are writing instructions and designing tools so the LLM can draw the flow at runtime. If you are deciding whether Copilot Studio is even the right tool for what you are building, Copilot Studio Is Not Always the Answer is worth reading before you go further.

A first build that teaches you the right mental model

Pick a small real scenario. Something like: surface open tickets assigned to the current user and let them add a comment. Not a chatbot. An agent with a job.

Create a new agent in Copilot Studio with generative orchestration enabled. Skip topics entirely for this first build. Write clear instructions describing what the agent does, what it does not do, and how it should respond when it cannot help.

Add one knowledge source. A SharePoint site or a Dataverse table works well. Ask it questions against that knowledge and watch how it grounds.

Add two tools. One Power Automate flow that reads data, one that writes. Make sure both return structured output, not just a success flag. Test what happens when the write fails. If the agent does not know it failed, fix the tool, not the agent.

That is the entire first build. No trigger phrases. No dialog tree. No variables you manage manually. You will learn more in a week of this than a month of the traditional tutorial path.

What to learn next once your first agent works

Once that agent behaves reliably, then learn topics. Use them for the narrow cases where you need a deterministic script. Identity confirmation. Regulated disclosures. Multi-step forms with strict validation. Topics are good at this. They are bad as the primary way to build an agent.

After topics, learn evaluation. Not just output testing. Behavioural testing. Does the agent handle incomplete questions, mid-conversation intent switching, and edge case inputs the way you expect? Most agentic workflows that feel intelligent in a demo turn out to be just fancy if/then logic in a trench coat when real users hit them. This is the single skill that separates agents that pass a demo from agents that survive production.

Then learn MCP, custom connectors, and multi-agent orchestration. By that point you will have the mental model to evaluate whether you actually need them or whether a simpler design solves the problem.

The fastest way to be useful with Copilot Studio in 2026 is not to learn every feature. It is to learn the four things above, build one real agent, and resist the pull of outdated tutorials that still dominate the first page of search results.

Frequently Asked Questions

How do I get started with Copilot Studio in 2026 without wasting time on outdated tutorials?

Skip the classic topic-and-trigger-phrase approach and go straight to building a generative orchestration agent. Focus on writing clear instructions, connecting knowledge sources, and designing reliable tools rather than learning the node-based dialog editor.

What is generative orchestration in Copilot Studio and why does it matter?

Generative orchestration is now the default approach for new agents in Copilot Studio, where the agent uses AI to handle conversation flow rather than following a rigid dialog tree. It reduces the need to manually write dozens of trigger phrases and scales far better than the older topic-based model.

Why does my Copilot Studio agent feel rigid and break when users ask unexpected questions?

This usually happens when the agent is built around a topic tree with trigger phrases, which only handles the exact variations you anticipated. Switching to a generative orchestration approach with well-written instructions and grounded knowledge sources makes the agent much more flexible.

When should I still use topics in Copilot Studio?

Topics are still useful when a specific flow must follow a strict, predictable script, such as a regulated process or a form-style interaction with no variation. They work best as guardrails for those edge cases rather than as the foundation of the entire agent.

April 20, 2026
Most Agentic Workflows Are Just Fancy If/Then Logic in a Trench Coat

People keep asking in the community what makes an agentic workflow actually useful. The honest answer is that most things being called agentic workflows right now are not. They are linear automations with a language model bolted on for the response step. That distinction matters more than most teams realise when they start building.

What a Useful Agentic Workflow Actually Does

A useful agentic workflow does something a standard Power Automate flow cannot: it makes decisions mid-execution based on context it discovered during the run, not based on conditions you hard-coded before it started.

That sounds obvious. It is not, in practice.

A flow that checks a field value and routes left or right is not an agent. An agent is something that can retrieve information it did not start with, reason about what that information means for the current task, and take a different action than you would have anticipated when you designed it. The key word is discovered. The agent had to go and find out something, then act on it.

If you can fully diagram the execution path before the workflow runs, it is probably not agentic. It is a well-structured flow. There is nothing wrong with a well-structured flow. But you should not be paying the overhead of agent infrastructure to build one.

Where Teams Go Wrong Building Agentic Workflows

The most common mistake I see is treating the language model as the agent. The LLM is not the agent. The LLM is the reasoning layer. The agent is the system that decides when to call what tool, handles what comes back, and determines whether the result is good enough to proceed or whether it needs to try something else.

When that orchestration layer is weak or missing, you get a workflow that calls one tool, takes the output at face value, and moves on. That is not reasoning under uncertainty. That is a glorified lookup with a friendly response message.

I wrote about silent action failures in the context of Copilot Studio earlier (the production testing post covers this in detail). The same failure mode appears in agentic workflows, but it is worse because the agent has more steps where it can silently accept a bad result and keep going. A flow fails at a specific action. An agent can propagate a bad intermediate result through three more steps before anything looks wrong.

The Two Things That Make or Break an Agentic Workflow

Based on what I have built internally and what I hear from people at other organisations, it comes down to two things.

First: tool design. The actions available to your agent need to return enough context for the agent to evaluate them, not just a success or failure signal. If your Power Automate flow returns {"status": "done"}, the agent has no way to assess whether done means what the user needed. It will treat it as success. Your tools need to return structured, interpretable output. This is not a language model problem. It is an API design problem.

Second: failure handling that is explicit, not optimistic. A useful agent knows when it is stuck and does something about it. That might mean escalating to a human, asking the user for clarification, or stopping cleanly with an honest message. What it does not do is generate a confident-sounding response for a task that did not complete. That is the failure mode that destroys trust in agents faster than anything else, because the user finds out later, not immediately.

I covered how this plays out in Copilot Studio specifically in the post on when Copilot Studio is the wrong choice. But the principle applies regardless of the tooling. An agent that cannot fail gracefully is not useful in production. It is a liability.

What Agentic Workflows Are Actually Good For

The use cases where agentic workflows justify their complexity share a few characteristics. The task has multiple possible paths and you cannot enumerate them all upfront. The inputs are unstructured or variable enough that rule-based routing breaks down. The system needs to recover from partial failures without a human in the loop for every edge case.

Document processing that involves extracting, validating, cross-referencing, and then acting on extracted data is a reasonable fit. Multi-step research tasks where what you search for next depends on what you found are a reasonable fit. Anything where the decision logic changes frequently and hard-coding it into a flow becomes a maintenance problem is worth evaluating. Before committing to that architecture, though, it is worth asking whether the underlying process is actually sound — automating a bad process just makes it fail faster, and agentic workflows are no exception.

A status check is not a fit. A single-action task triggered by a button is not a fit. Anything you can build cleanly as a Power Automate flow with proper error handling is probably not worth the overhead of an agentic architecture. The orchestration cost is real and the debugging surface is larger.

The Test I Use

Before committing to an agentic workflow architecture, I ask one question: does this task require the system to discover something during execution that changes what it does next, and would that discovery be different for different runs? As Halilcan Soran on LinkedIn, I have found this single question filters out more false positives than any other test I use.

If yes, agents are worth the investment. If no, you are adding complexity to solve a problem that a well-built flow could handle, and you will spend more time debugging Claude or other AI agent behaviour than you saved on logic design.

The technology is not the constraint. Knowing what you are actually building is.

Frequently Asked Questions

What

April 17, 2026
Adding Copilot to Your Power App Is Not the Same as Making It Smarter
Microsoft published a post this week about making business apps smarter by embedding Copilot, app skills, and agents directly into Power Apps. The features are real and some of them are genuinely useful. But I keep seeing teams read announcements like that and immediately open their existing apps to start wiring things in. That is where it goes wrong. Adding Copilot to Power Apps does not make the app smarter. It makes the AI visible. Those are different things.

What App Skills and Agent Integration Actually Do Under the Hood

When you expose a Power App as an app skill or embed a Copilot Studio agent into a canvas app, you are giving the AI a surface to operate on. The agent can read context from the app, trigger actions, and return responses into the UI. In theory, the AI bridges what the user needs and what the app can do.

In practice, the agent is only as capable as what you hand it. It reads data from your app’s data sources. It calls the actions you have defined. It interprets user intent against the topics and instructions you have written. If your data model is inconsistent, your actions are incomplete, or your process logic has gaps, the agent does not compensate for any of that. It just operates on top of it and returns confident-sounding responses anyway.

I wrote about this problem in a different context when covering why Copilot Studio agents fail in production. Silent action failures are one of the nastiest issues: the agent completes its response, the user thinks something happened, nothing actually did. That risk does not disappear when you move the agent inside a Power App. If anything, it gets harder to spot because users expect the app to be reliable.

Why the Data Model and UX Structure Matter More Than the AI Feature

Most Power Apps I have seen built inside large organisations were designed around a specific, narrow workflow. The data model reflects decisions made at the time of build, often under time pressure, often by someone who is no longer on the team. Fields are repurposed. Status columns hold values that mean three different things depending on which team is using them. Lookup tables have orphaned records nobody cleaned up.

When you put an agent on top of that, the agent queries this data and tries to give useful answers. The answers will be coherent. They will not be correct. Not reliably.

The UX structure compounds this. Canvas apps built for point-and-click navigation do not automatically become good AI surfaces. If a user can ask the agent to update a record, but the app’s own form has fifteen required fields and three conditional rules that only run client-side, you now have a conflict between what the agent can do via a Power Automate action and what the app enforces through its UI. One of them will win. It will not always be the right one.

This is the same argument I made about automating a bad process. The automation does not fix the process, it executes it faster and more consistently, including the broken parts. Embedding AI into a poorly structured app works the same way.

What I Check Before Wiring Any Agent Into an Existing App

Before I connect anything to a Copilot Studio agent or enable app skills on an existing Power App, I go through a short audit. Not a formal document. Just four questions that save a lot of cleanup later.
- Is the data model clean enough to query? If the same concept is stored in three different columns across two tables with inconsistent naming, the agent will surface that inconsistency directly to the user. Fix the model first.
- Are the actions the agent can trigger complete and safe? Every Power Automate flow an agent can call needs proper error handling and a defined failure response. Silent failures inside agent topics are a known problem. If the flow does not return a clear success or failure, the agent cannot respond accurately.
- Does the app enforce rules that the agent needs to know about? If business logic lives only in Power Fx expressions inside the app’s forms, the agent does not see it. Validation that matters needs to exist at the data layer or inside the flows the agent calls.
- Is the process the app supports well-defined enough to describe to an AI? If I cannot write a clear system prompt describing what the agent should and should not do in this app, the process is not ready. Ambiguity in the process becomes ambiguity in agent behaviour.
When Embedding AI in a Power App Is Worth It and When It Is Not

There are genuinely good cases for this. An app where users regularly need to find records across complex filters is a reasonable candidate. Surfacing a conversational shortcut to navigate a large dataset, trigger a common action, or get a summary of a record without clicking through multiple screens can reduce real friction. I have seen it work well when the underlying data is clean and the scope of what the agent can do is narrow and explicit.

The cases where it is not worth it yet are more common. An app with inconsistent data. A process with unresolved exceptions. A UX that was never designed with AI interaction in mind. In those situations, embedding an agent creates a new layer of support burden without a proportional benefit.

I also want to be direct about something I mentioned in my post on when Copilot Studio is the wrong choice: not every interaction benefits from being conversational. Some things in a Power App are faster as a button. The AI control is not always an upgrade on a well-placed filter or a clear form layout.

Connect with Halilcan Soran on LinkedIn for more insights on Power Apps and AI integration.
April 16, 2026
Copilot Studio Is Not Always the Answer

I keep seeing this on LinkedIn and in community forums. Someone describes an internal use case, and the first five replies are all “have you tried Copilot Studio?” The tool has gotten good enough that it has become the reflexive answer to any question involving automation, conversation, or AI. That reflex is causing real problems. Knowing when Copilot Studio is the wrong tool is as important as knowing how to build with it well.

When Copilot Studio Is the Wrong Tool for the Job

Most misuse I see falls into one of three situations. The use case is purely transactional. The interaction model is not conversational. Or the team wants a workflow, not an agent.

If someone needs to submit a form, approve a request, or trigger a process on a schedule, that is Power Automate territory. Putting a conversational interface in front of a single-action task does not make it better. It makes it slower, harder to test, and harder to maintain. Users do not want to type a sentence to do something they could do in two clicks.

The second situation is harder to spot. Some interactions look conversational but are not. A knowledge base search, a document lookup, a status check. These are point-in-time queries with no real back-and-forth. You could build them in Copilot Studio. You could also build them as a Power Apps canvas app with a simple search interface and ship it in a day with less moving parts and a much more predictable failure surface.

The Agent Complexity Problem

There is also a complexity ceiling that teams hit faster than expected. Copilot Studio agents work well when the conversation scope is tight. One domain. A few topics. Defined intents. When someone tries to build a single agent that handles HR queries, IT requests, and finance approvals inside the same session, topic routing starts failing at the edges. I wrote about this in Your Copilot Studio Agent Passed Every Test and Still Failed in Production. When a user’s phrasing sits between two topics, the agent picks one confidently and gets it wrong. The more topics you add, the more edge cases you create, and the harder they are to test systematically.

The instinct to build one agent that does everything is understandable. It feels cleaner. In practice it produces an agent that does everything poorly and fails in ways that are genuinely difficult to diagnose.

Where the Wrong Choice Usually Starts

It usually starts with the framing of the requirement. Someone says “we want a chatbot” and that phrase triggers Copilot Studio before anyone has defined what the interaction actually needs to do. I have seen teams spend weeks building agent topics, writing generative AI prompts, and wiring up Power Automate actions, when what the users actually wanted was a better SharePoint search and a weekly digest email.

The honest question to ask before opening Copilot Studio is this: does this use case genuinely require back-and-forth conversation, or does it just need to surface information or move data? If the answer is the second one, there is almost always a simpler path.

This is not a knock on Copilot Studio. The tool is genuinely capable when it fits the problem. Handling multi-turn conversations, routing across complex intent patterns, integrating generative answers with structured actions, those are things it does well. But that capability comes with a real operational cost. There is a topic structure to maintain, system prompts that drift when production data introduces edge cases, Power Automate actions that can fail silently inside a topic and return a confident-sounding response for work that was never done.

What to Reach for Instead

Power Apps for anything with a fixed interaction model. Canvas apps are underrated for internal tooling. They give you a defined UI, predictable state, and a clear place to debug when something breaks.

Power Automate for anything triggered, scheduled, or event-driven. If there is no user in the loop having a conversation, there is no reason for Copilot Studio to be involved. Keep in mind that even straightforward flows can run into issues at scale, as Power Automate throttling limits will break your flow in production under real load if you have not accounted for them.

SharePoint or Dataverse with a search interface for knowledge retrieval. If users are looking something up, build a search experience, not a conversational one.

In enterprise environments, the governance overhead of Copilot Studio also matters. You are managing an agent that generates natural language responses. That response quality needs to be reviewed, monitored, and occasionally corrected. Most teams I talk to underestimate this cost until they are three months into production and someone in legal asks why the agent said something it should not have.

The Right Question Before You Build

Before any Copilot Studio project starts, the question worth asking is not “how do we build this agent” but “does this use case actually need an agent.” If the answer requires you to stretch the definition of conversation to make it fit, that is a sign to stop and pick the simpler tool.

Copilot Studio is a good tool. It is not a default. Using it where it fits produces something worth building. Using it where it does not produces something you will be maintaining and explaining for a long time.

Frequently Asked Questions

When should I use Copilot Studio instead of another tool?

Copilot Studio works best when the interaction is genuinely conversational, scoped to a single domain, and involves a defined set of intents. If the task is transactional, point-in-time, or better served by a simple form or search interface, tools like Power Automate or Power Apps are likely a faster and more maintainable choice.

What is the difference between Copilot Studio and Power Automate?

Power Automate is built for workflow and process automation, such as form submissions, approvals, and scheduled triggers. Copilot Studio is designed for conversational agent experiences. Using Copilot Studio for single-action tasks adds unnecessary complexity without improving the user experience.

Why does my Copilot Studio agent keep routing users to the wrong topic?

Topic routing breaks down when an agent is built to handle too many domains or intents within a single session. When a user’s phrasing falls between two topics, the agent will confidently pick one and get it wrong. Keeping each agent focused on a narrow scope reduces these edge cases and makes failures easier to diagnose.

How do I know if my use case actually needs a chatbot?

Start by defining what the interaction needs to do before choosing a tool. If users need a back-and-forth conversation to complete a task, a conversational agent may be appropriate. If they need a search result, a status update, or a simple action, a canvas app or improved search interface will often deliver a better outcome in less time.

April 15, 2026
Your Copilot Studio Agent Passed Every Test and Still Failed in Production

I came across a post from Zapier Blog about AI agent evaluation, and it described something I keep seeing inside large organisations: an agent that looks perfect in a demo, gets signed off, goes live, and then immediately starts doing things nobody expected. Wrong tool calls. Conversation loops that never resolve. Outputs that look confident and are completely wrong. The post frames this well as a sandbox problem. But the fix it describes, better test coverage and smarter metrics, only gets you partway there. The deeper issue with Copilot Studio agent testing is not the quantity of your tests. It is what you are actually testing for.

Why Demo-Passing Agents Break in Real Workflows

When a team builds an agent in Copilot Studio, they test it against the happy path. A user asks a clean question. The agent triggers the right topic or action. The response looks good. Someone in the review meeting says it works great. The agent gets promoted to production.

The problem is that real users do not ask clean questions. They ask incomplete ones. They switch intent halfway through a conversation. They paste in text that includes formatting your prompt never anticipated. They use your agent for things it was never designed to do, because nothing in the interface tells them not to.

None of that shows up in a demo. It shows up three days after go-live when someone forwards you a conversation log that reads like a stress test you forgot to run.

The Three Failure Modes I Keep Seeing in Copilot Studio Agents

After building and reviewing a number of agents internally, the failures cluster into three patterns.

Topic misrouting at the edges. Your agent routes correctly when the user says exactly what you expected. But natural language is messy. When a user’s phrasing sits between two topics, the agent picks one confidently and gets it wrong. You only discover this when someone captures a failed session and traces it back. By then, a dozen other users have hit the same wall and just stopped using the agent.

Action failures that degrade silently. A Power Automate flow or a connector action fails in the background and the agent carries on as if nothing happened. No error surfaced. No fallback triggered. The user gets a response that implies the task completed. It did not. This is the agent equivalent of a flow that retries quietly and masks the problem until the load goes up. I wrote about that pattern in the context of Power Automate throttling limits breaking flows under real load. The same logic applies here: silent success is not success.

Prompt instruction drift under real data. Your system prompt was written against clean test data. Production data is not clean. It has unexpected characters, long strings, mixed languages, or values that push the model toward an interpretation you did not intend. The agent’s behaviour drifts. Not catastrophically. Just enough to become unreliable in ways that are hard to reproduce and harder to explain to stakeholders.

How to Build a Behavioral Test Suite Instead of an Output Checklist

Most teams build an output checklist. Did the agent return the right answer for these ten questions? That tells you almost nothing about production behaviour.

What you actually need is a behavioral test suite. The difference is this: output testing checks what the agent said. Behavioral testing checks how the agent handled the situation.

Here is how I approach it inside Copilot Studio before promoting anything to production.

Build adversarial input sets, not just representative ones. For every topic your agent handles, write three versions of the trigger: the clean version, an ambiguous version that could belong to two topics, and a broken version with incomplete or oddly formatted input. If the agent routes all three correctly, you have something worth shipping. If it fails on the ambiguous case, you have a routing gap that will hit real users constantly.

Test conversation state, not just single turns. Copilot Studio agents hold context across a conversation. Test what happens when a user changes their mind on turn three. Test what happens when they ask a follow-up that assumes context the agent should have retained but might not. Single-turn testing misses an entire class of failure that only appears in multi-turn sessions. This is also why agentic workflows require a fundamentally different design approach, not just an AI layer placed on top of existing processes.

Inject real data samples into action inputs. Pull a sample of actual data from your environment and run it through the actions your agent calls. Do not use synthetic test data if you can avoid it. Real data has edge cases your synthetic data will never cover. If your agent calls a flow that queries a SharePoint list, run the query against the actual list with actual entries, including the ones with blank fields and formatting you did not anticipate.

Define explicit fallback behaviour and test it deliberately. Every agent should have a defined behaviour for when it cannot complete a task. Most teams add a fallback topic and assume it works. Test it by constructing inputs that should trigger it. If the fallback does not fire, or fires on the wrong inputs, fix it before go-live. A graceful failure is far better than a confident wrong answer.

What to Monitor After Go-Live and When to Pull an Agent Back

Testing before launch is necessary but not sufficient. Agent behaviour shifts as the inputs it receives in production diverge from what you tested against. You need monitoring in place from day one.

Track escalation rate and abandon rate per topic. If a topic is seeing significantly higher escalations than others, that is a signal of routing or response quality problems, not user error. Track action failure rates separately from conversation outcomes. An agent can complete a conversation and still have failed to do the thing the user needed.

Set a threshold before launch. If escalation rate exceeds a number you agree on in advance, or if a specific action is failing more than a defined percentage of the time, you pull the agent back or disable the affected topic. The threshold is arbitrary. Having no threshold at all is not.

The agents I have seen hold up in production are not the ones with the most sophisticated prompts. They are the ones where someone spent real time on the failure cases before launch and built actual monitoring into the plan from the start.

If you are still signing off agents based on demo performance, you are not testing. You are hoping.

Frequently Asked Questions

Why does my Copilot Studio agent testing pass in demos but fail in production?

Most Copilot Studio agent testing is built around ideal user inputs and predictable conversation paths, which do not reflect how real users actually behave. In production, users ask incomplete questions, switch intent mid-conversation, and use the agent in unintended ways that no demo ever surfaces. Testing needs to go beyond the happy path to catch these edge cases before go-live.

What are the most common failure modes in Copilot Studio agents?

The three patterns that appear most often are topic misrouting when user phrasing falls between two intents, action failures that complete silently without triggering any error or fallback, and prompt instructions that break down when they encounter messy real-world data. Each of these can go undetected in testing because they only emerge under realistic conditions.

How do I know if a Power Automate action failed inside my Copilot Studio agent?

Silent action failures are a serious risk because the agent can continue the conversation and imply a task completed when it did not. You need explicit error handling and fallback logic in your flows so that failures surface to the user rather than being masked by a confident-sounding response.

When should I test my Copilot Studio agent against real production data?

You should test against realistic data before promotion to production, not after. System prompts written against clean test data can behave unpredictably when they encounter unexpected characters, mixed languages, or long strings that only appear in live environments. Incorporating a sample of real or representative data into your test suite is a necessary step before sign-off.

Source: AI agent evaluation: How to test + improve AI agents (Zapier Blog).

April 14, 2026
Stop Bolting Agents Onto Power Automate Flows
Updated July 2026 rewrote the orchestration section now that multi-agent handoffs and MCP tool access are generally available, and added a section on the accountability problem nobody designed for.

The mistake I keep seeing

A team hears about AI agents. They want to add AI to their approval workflow. So they take the existing 10-step Power Automate flow, stick a Copilot Studio agent somewhere in the middle, and call it an agentic workflow.

It isn’t. It’s the old process with a chatbot attached.

I ran into this internally, and people I talk to at other orgs describe the exact same pattern. The agent becomes a fancy input form. The process stays broken. And when it fails, and it does, everyone blames the AI. Then the whole automation program takes a credibility hit for a design mistake that had nothing to do with the model.

What actually makes a workflow agentic

An agentic workflow is not about adding a language model to a flow. It’s about giving the system the ability to reason about what to do next, not just execute a predefined sequence.

The difference matters. In a traditional flow, you define every branch. Every condition. Every outcome. The machine follows instructions. In an agentic workflow, the agent interprets a goal, decides which tools or actions to use, and adjusts based on what it gets back.

Anthropic made this distinction cleanly in their Building Effective Agents piece: workflows are systems where you orchestrate the steps, agents are systems where the model directs its own process. Most of what I see labeled agentic in enterprise Power Platform work is still a workflow. That’s fine. Workflows are great. Just call them what they are and design them accordingly.

Agentic design requires a fundamentally different approach. You’re not mapping steps. You’re defining boundaries, tools, and acceptable outcomes.

Three things that have to change in your process design
- Stop thinking in sequences. Agentic workflows are goal-driven, not step-driven. Define what done looks like, not every micro-step to get there. If your flow diagram looks like a subway map, you’re still in traditional automation mode.
- Give the agent real tools, not just data. An agent that can only read a SharePoint list and send an email is not doing much reasoning. It needs to call APIs, query systems, write back to records, trigger sub-flows. This got significantly easier once MCP support in Copilot Studio matured. You can now expose internal systems as tool servers instead of building a custom connector for every single action. But the design question hasn’t changed: people still give agents access to everything or nothing. Neither works. Scope each tool to the narrowest useful action.
- Build in failure handling at the goal level. Traditional flows handle errors at the step level. If this action fails, go here. Agentic workflows need you to think about what happens when the agent reaches a dead end, produces a low-confidence result, or loops without resolution. I’ve seen agents spin for 40 iterations on a task that should have escalated to a human after three. Set iteration caps. Set confidence thresholds. Define the escalation path before you define anything else.
Where this actually works in business processes

Not everywhere. I want to be direct about that.

Agentic design makes sense when the process has variability you cannot fully predict upfront. Invoice exceptions. Complex customer complaints. Multi-system data reconciliation where the right answer depends on context you only know at runtime.

It does not make sense for processes that are well-defined and stable. If your purchase order approval follows the same 6 steps every time, a standard Power Automate flow is the right tool. It’s deterministic, cheap to run, and easy to audit. Don’t add an agent to it because a roadmap slide told you to.

The teams getting real results are the ones who identify a process where exceptions are eating their staff’s time, then let the agent handle the exceptions rather than replacing the whole flow. The happy path stays deterministic. The agent works the mess.

The orchestration layer, revisited

When I first wrote this post, coordinating multiple agents in the Power Platform meant building the handoff logic almost entirely yourself. That has changed. Multi-agent orchestration in Copilot Studio now handles agent-to-agent handoffs natively, and connected agents can share context across a task.

Here’s what hasn’t changed: the platform routes the handoff, but it doesn’t design your boundaries for you. Agent B still won’t know what Agent A already tried unless you decide what context gets passed and what gets dropped. I’ve watched a document processing agent and a system update agent duplicate work for weeks because nobody defined which one owned the retry logic.

Document your agent boundaries explicitly. What does each agent know? What can it do? What should it never do? Treat it like designing a team of junior staff who are fast and tireless but have no common sense unless you’ve given them the right context. The tooling got better. The thinking is still on you.

The accountability gap

One thing I underweighted the first time around: agents don’t take responsibility. A colleague put it to me bluntly. You can’t have the numbers sometimes right and sometimes not. Enterprise processes need someone accountable for the outcome, and an agent is not someone.

This is a design constraint, not a reason to avoid agents. It means every agentic workflow needs a named human owner, a decision log the owner can actually read, and a clear line for which decisions the agent makes autonomously versus which ones it recommends. If you can’t answer “who is responsible when this agent gets it wrong” in one sentence, the workflow isn’t ready for production. I’ve written about this pattern before and it comes up in nearly every conversation on my LinkedIn.

Start smaller than you think you should

Pick one process. One that has clear exceptions, high manual effort, and a measurable outcome. Build the agent, give it two or three tools, test it against real historical cases before you deploy it anywhere near live data.

The teams succeeding with agentic workflows in 2026 are not the ones with the biggest ambitions. They’re the ones who are rigorous about scope, honest about where the agent is making decisions versus guessing, and fast to pull the agent out of the loop when something looks wrong.

Agentic is a design philosophy. Apply it where it earns its complexity.
April 13, 2026

Tag: Copilot Studio

What Microsoft Discovery Actually Is If You Skip the Marketing

Why Domain-Shaped Agent Platforms Beat Generic Copilots for R&D Workflows

How This Changes the Build vs Buy Decision for Power Platform Teams

What I Would Watch For Next in the Microsoft Agent Stack

Frequently Asked Questions

What is Microsoft Discovery and how does it differ from Copilot Studio?

How does Microsoft Discovery agentic RD improve research and development workflows?

When should I choose a domain-specific agent platform over a generic one like Copilot Studio?

Why do generic agentic workflows often fail for complex enterprise use cases?

Why Head to Head Model Comparisons Stop Being Useful the Moment You Add a Connector

The Four Things That Actually Matter When a Model Sits Inside an Automation

How I Pick Between Claude and GPT for a Specific Flow

What to Test Before You Commit a Model to Production

Frequently Asked Questions

Which is better, Claude vs ChatGPT for automation workflows?

Why does my AI agent start producing errors after working fine at first?

How do I choose an AI model for a Power Automate or Copilot Studio flow?

What causes silent failures in AI automation workflows?

Why the GitHub Connector Demo Lies to You

The Environment Sprawl Problem Nobody Scopes For

How I Would Map Power Platform Environments to GitHub Orgs and Repos

Service Accounts Tokens and the Stuff That Actually Breaks in Prod

Frequently Asked Questions

How do I set up a Power Platform agent GitHub integration in an enterprise environment?

Why does my Power Platform agent work in UAT but fail in Production when connecting to GitHub?

What is environment sprawl and why does it matter for Power Platform agent deployments?

When should I use a service account instead of personal authentication for a GitHub connector in Power Platform?

Why Agent Loops Feel Slow Even When Each Call Is Fast

What OpenAI Just Shipped and Why It Matters Beyond Codex

What This Looks Like Inside Copilot Studio and Power Automate

How I Would Budget Latency Before I Build the Agent

Frequently Asked Questions

Why does agentic workflow latency get so bad in multi-step agents?

What is connection-scoped caching and how does it help agent performance?

How do I reduce latency in Copilot Studio and Power Automate agents?

Why do users abandon AI agents even when the agent gives correct answers?

Why the 2023 Copilot Studio learning path is actively misleading in 2026

The four things that actually matter when you start today

A first build that teaches you the right mental model

What to learn next once your first agent works

Frequently Asked Questions

How do I get started with Copilot Studio in 2026 without wasting time on outdated tutorials?

What is generative orchestration in Copilot Studio and why does it matter?

Why does my Copilot Studio agent feel rigid and break when users ask unexpected questions?

When should I still use topics in Copilot Studio?

What a Useful Agentic Workflow Actually Does

Where Teams Go Wrong Building Agentic Workflows

The Two Things That Make or Break an Agentic Workflow

What Agentic Workflows Are Actually Good For

The Test I Use

Frequently Asked Questions

What

What App Skills and Agent Integration Actually Do Under the Hood

Why the Data Model and UX Structure Matter More Than the AI Feature

What I Check Before Wiring Any Agent Into an Existing App

When Embedding AI in a Power App Is Worth It and When It Is Not

When Copilot Studio Is the Wrong Tool for the Job

The Agent Complexity Problem

Where the Wrong Choice Usually Starts

What to Reach for Instead

The Right Question Before You Build

Frequently Asked Questions

When should I use Copilot Studio instead of another tool?

What is the difference between Copilot Studio and Power Automate?

Why does my Copilot Studio agent keep routing users to the wrong topic?

How do I know if my use case actually needs a chatbot?

Why Demo-Passing Agents Break in Real Workflows

The Three Failure Modes I Keep Seeing in Copilot Studio Agents

How to Build a Behavioral Test Suite Instead of an Output Checklist

What to Monitor After Go-Live and When to Pull an Agent Back

Frequently Asked Questions

Why does my Copilot Studio agent testing pass in demos but fail in production?

What are the most common failure modes in Copilot Studio agents?

How do I know if a Power Automate action failed inside my Copilot Studio agent?

When should I test my Copilot Studio agent against real production data?

The mistake I keep seeing

What actually makes a workflow agentic

Three things that have to change in your process design

Where this actually works in business processes