Home / Copilot agents
Plain English guide
Plain-English guidance for any team deciding where agents fit, how to govern them and when Copilot Studio is the right build route.
Good first agent signals
There are two kinds of agent. Agents you talk to help people find and draft things, and Microsoft's own Copilot now does a lot of that out of the box. Agents that run start on their own when something happens: an email arrives, a meeting ends, a report is exported, a date comes round. They do the checking, drafting and filing, then stop for one of your team to approve. The second kind is where a firm gets hours back, and it's what we build.
Both kinds are built around a particular job, source of knowledge, audience or workflow, rather than asking general Copilot to help with whatever is in front of you. Microsoft keeps making Copilot better, and we'll tell you when it already does the job. What it doesn't know is your firm: your thresholds, your templates, your filing conventions and who signs off what. That's the part we build and look after.
In an accountancy or law firm, an agent you talk to might be a knowledge assistant or a client intake guide. An agent that runs might file the morning post, push WIP exceptions to managers or turn a client meeting into a filed note. The important detail is not that the agent is clever. It is that the scope is narrow enough to trust.
Agents work best when they sit between broad Copilot use and hard automation. They can guide a person through a process, answer from approved material and collect useful context. They should not remove professional judgement, client responsibility or the need to check the output. When the job needs a build, this connects to Copilot Studio consultancy; when it only needs rule-based steps, Power Automate consultancy is usually the better route.
For the wider picture of how agent capability grows, the Agent Journey maps five stages from using a first agent to running a team of agents, with a learn or build route at every stage.
If your team would rather learn the basics together, the Copilot agents course covers what an agent is, how to use one safely and how to build a simple one, in two or three hours.
Types of agents
The right route depends on the job, the source content, the audience and how much control the firm needs over testing, actions and handover.
In Microsoft 365
Agents that sit close to everyday Microsoft 365 work are usually best for helping people find knowledge, follow internal guidance or complete a repeated task without leaving the flow of work.
In Copilot Studio
Copilot Studio agents are better when the job needs a defined conversation, a managed set of instructions, specific knowledge sources or a connection into a workflow.
Configuration choice
Some agents can stay close to existing Microsoft 365 content and instructions. Others need a more custom build with tools, connectors, data sources and stronger testing.
Variable knowledge work where a person stays in control.
Drafting, summarising, meeting follow-up, analysis support and rewriting.
A narrower assistant around a repeated task, audience or knowledge source.
Internal policy guidance, client intake, onboarding support and document review preparation.
Work that starts on a trigger, needs AI for one judgement step and must stop for a person to approve.
Starts when an email arrives, a file lands or a date comes round. Uses AI to read, match or draft, acts through Power Automate, then waits for approval in Teams.
Repeatable, rule-based steps that should run without a conversation.
Approvals, reminders, notifications, task creation, file movement and status updates.
Judgement, relationship handling, final review and decisions with professional responsibility.
Client advice, matter strategy, partner sign-off, exception handling and ethical judgement.
Good use cases
A focused agent that helps staff find approved policies, templates, process notes or firm guidance without searching across old Teams channels and SharePoint folders.
A guided route that asks the right questions, explains what information is needed and prepares a structured handover for the team that owns the next step.
Starts when the day's scanned post lands in a folder. Each letter is read, matched to a client and filed with an audit line, and anything uncertain goes to a person.
Starts when the practice system's WIP report arrives. Each manager gets their own exceptions in Teams with bill now, defer and write off buttons, and every decision is logged.
A controlled way to point people to current policies, escalation routes and process guidance, especially when the source content has a named owner.
Starts when the Teams meeting ends. The file note is drafted in the firm's template with the actions, and it is filed, tasked and followed up only once the adviser approves.
Governance and security
Agent governance is not there to slow the work down. It is what lets a firm use agents without turning source content, permissions and professional judgement into avoidable risk.
An agent people talk to should only reach what the person asking can already see. An agent that runs on its own uses a service account limited to the mailboxes, folders and systems its job needs. Either way, permission hygiene is part of the product.
An agent grounded in poor content gives poor answers faster. Useful builds usually start by naming the approved sources and cleaning up the obvious gaps.
Every agent needs a business owner for scope and content, plus a technical owner for configuration, access and support.
Good agents are tested against normal questions, awkward questions, edge cases and the moments where they should say they cannot help.
Agents should be reviewed, improved or retired. If the process changes and the agent does not, trust will fade quickly.
Inside an agent build
The aim is not to build the most advanced thing possible. It is to build the smallest useful agent that has a clear job, safe boundaries and a route to being maintained.
01
We start with the work itself: who uses the process, how often it happens, what source material exists and what goes wrong today.
02
The answer might be an agent, an agent workshop, a Power Automate flow, SharePoint cleanup or a simpler process change. The tool follows the job.
03
We define what the agent can use, what it should refuse, what it should escalate and how users will understand its limits.
04
The first build is tested with real questions and real users. We look for useful answers, permission issues, vague wording and weak handoffs.
05
People need to know when to use the agent, how to ask better questions, what to check and how to report problems. That briefing is part of the handover, not a course.
06
A useful handover includes ownership, review rhythm, documentation and practical measures for whether the agent is helping.
Where to go next
Capability map
Five stages from using your first agent to running a team of agents, with a learn or build route at every stage.
Commercial build support
The service page for scoping, building and rolling out agents with clear ownership, testing and handover.
Automation route
For rule-based workflow steps, approvals, reminders and handoffs that should run without an agent conversation.
Accountancy
Six ready-made agents for the admin every practice shares, each with a trigger, one job and a point where your team approves.
Worked example
An accountancy firm's incoming post read, matched against more than 1,500 live clients and filed in about two minutes, with people handling only the exceptions.
Legal
Where agents can support knowledge, intake and review preparation without replacing legal responsibility.
Next step
A short conversation is usually enough to tell whether the work needs an agent, a flow, a workshop or a cleaner source of truth.