Question 1

Do we need Microsoft 365 Copilot licences for this?

Accepted Answer

Not necessarily. Copilot Studio has standalone licensing and consumption options; it does not require buying Microsoft 365 Copilot for every user. Microsoft 365 Copilot licences also include some agent capabilities, so an existing licence can change what needs to be purchased.

The assessment confirms what your tenant needs for the proposed users, publishing channel, actions and connected services. Licence entitlements and consumption rules change, so we check the current Microsoft terms before recommending a purchase.

Question 2

Where does our data go, and does the agent train on it?

Accepted Answer

Grounding retrieves relevant information to answer a question; it does not, by itself, train a model on your documents. For the Azure OpenAI-backed features covered by Microsoft’s privacy FAQ, customer data is not used to train Microsoft’s Copilot AI models unless the tenant opts in, and prompts and responses are not used to train third-party models.

This is cloud processing, not a promise that all data stays in one tenant or in Canada. Knowledge files can be copied into Dataverse, activity history can use Microsoft 365 storage, and connectors or external models can send data to other services under their applicable terms. The assessment checks the chosen model, data sources, permissions, regions, retention and optional data-sharing settings against current Microsoft documentation before a build is agreed.

Question 3

Can the agent respect who is allowed to see which documents?

Accepted Answer

It depends on the source. A file uploaded straight into an agent can be read by anyone who can use that agent, because the upload does not carry the file’s original permissions with it. The same document connected through SharePoint or OneDrive is checked against the person asking, and sign-in has to be switched on for that check to happen.

The assessment checks this before anything is built: which sources are used, who can reach the agent, and whether someone could receive an answer drawn from a document they could not open themselves. Where that risk exists, the fix is usually the source or the audience, not the wording of the prompt.

Question 4

Do we need to organise our documents before we start?

Accepted Answer

You do not need to tidy the whole document library before the assessment. We start with a representative sample and the questions your team asks most often, then look for missing guidance, conflicting versions and unclear ownership.

Before a build, we agree which documents the agent can use, who owns them and how outdated versions will be retired. If the source material needs work first, the assessment identifies what to fix and why.

Question 5

What access do you need to our tenant, and for how long?

Accepted Answer

The assessment starts with an administrator walkthrough and an agreed document sample. We ask for access only where that review needs it. Before any account is granted, we agree its purpose, permissions and removal date; standing Global Administrator access is not the starting point.

For a build, the scope identifies the environment, agent and data access needed for the work. Your administrator handles privileged changes. Access is limited to the agreed engagement window and removed at handover, unless a separate support arrangement explicitly extends it.

Question 6

How is this priced?

Accepted Answer

The readiness assessment is fixed scope and fixed price, agreed before it starts. A build is quoted separately once the assessment has established what it involves, against a written scope with acceptance checks.

Microsoft licensing and consumption are billed by Microsoft, not by us, and sit outside our fee. The assessment identifies what your tenant will need so the running cost is visible before you commit to a build.

Question 7

What happens when our documents change?

Accepted Answer

It depends on the source. Locally uploaded files are snapshots and need replacing. Supported SharePoint and OneDrive file connections can synchronise changes, but refresh and indexing are not instantaneous. Updating a document is not the same as proving the agent is using the right version.

The handover names a document owner and an update process, including retiring superseded versions and rerunning checks on citations and refusals. If the approved documents do not support an answer, the intended behaviour is to refuse and point to a person. Grounding reduces risk; it does not make generated answers infallible.

Question 8

What if the honest answer is that we shouldn’t build an agent?

Accepted Answer

That is a valid assessment result. The report explains why, what would need to change, and whether a simpler process, form or automation would help. You pay for a decision you can use, and there is no obligation to commission a build.

Question 9

How long does a first build take?

Accepted Answer

The build timeline is agreed after the assessment, before work starts. It depends on the document set, permissions, integrations and acceptance checks. We will not give you a delivery date before those dependencies are understood.

Question 10

Who do we need to involve on our side?

Accepted Answer

Three roles, and often one person covers more than one of them. Someone who owns the documents or the process and can say what the correct answer is. Someone with tenant administration rights for access and publishing. Someone who does the work daily and can say where it actually breaks. Progress usually stalls on the first of those rather than the technical ones.

Question 11

What do we own at the end?

Accepted Answer

You retain your tenant and data. The agreed handover includes the working solution, configuration and documentation needed to operate and maintain it. The statement of work identifies the deliverables, ownership terms and any pre-existing or third-party dependencies before the build begins. Microsoft services and third-party software remain subject to their own licences and ongoing charges.

Question 12

What if the agent gives a wrong answer once it is live?

Accepted Answer

Grounding reduces that risk; it does not remove it. The build defines what the agent should refuse as well as what it should answer, and the acceptance checks test both before go-live.

After handover you have the update process, the checks to rerun and a named owner for the source documents. A wrong answer is more often a document problem than a model problem, so the first step is finding which source produced it.

Question 13

Can our own team change the agent after handover?

Accepted Answer

That is the intention. The handover covers how the agent is put together, where its knowledge comes from and how to change it, so someone on your side can maintain it without us. Where a change would need skills the team does not have yet, we would rather say so during the assessment than leave it as a surprise after go-live.

Agents that know when to stop, answer with sources, keep work moving, and connect your tools.

Work that keeps coming back

The same questions, every day

Documents change. Answers don’t.

A good demo, a difficult production rollout

Assess, build, extend and govern

Assess

Build

Extend and govern

App Library

Turn your policies into answers people can trust.

Find the right quote without switching tools.

Keep documents moving. Bring people in when it matters.

Time Off and Request Intake

Field Safety Assistant

Client Onboarding Assistant

Agent Governance Setup

From the first call to handover

Discovery call

Assessment

Scoped build

Handover with documentation

Built around your team

Before getting started

Start with what is going wrong

Agents that know when to stop, answer with sources, keep work moving, and connect your tools.Agents thatknow when to stopanswer with sourceskeep work movingconnect your toolsknow when to stop