Artificial intelligence

Microsoft Copilot Studio

Build internal agents with Microsoft Copilot Studio

Build and test an agent that answers process questions using authorised documentation, with references, clear boundaries and a review the team can continue.

12 h · 6 × 2 hSpanish & English
Contact

Before we begin

Corporate identity and authoring access in Microsoft Copilot Studio, an authorised practice environment and administrator support. Confirm licence, features, sources, permissions and usage. The free trial supports creating and testing, not publishing; licence purchases and tenant setup are not included.

Check identity, authoring, environment, features, policies, sources and usage. A free trial does not allow publishing. Uploaded files do not demonstrate individual SharePoint permissions: that needs prepared sources and test users. Classroom material does not connect real documents or change admin policies.

See preparation and terms →

By the end you will be able to

  1. 01

    Define an internal agent’s objective, useful questions, sources and boundaries.

  2. 02

    Prepare authorised knowledge and check answers against references.

  3. 03

    Handle clarification or routing without claiming actions the agent has not performed.

  4. 04

    Test cases and gaps, and leave a handover with owners and conditions for possible publication.

What we will work on

An internal agent is useful when it knows which documentation to consult and where its task ends. Build a focused tool-and-access prototype, check references and leave a handover the team can use to continue the review.

01

Scope the agent and prepare its knowledge

Questions, owners, currency and access. Create the agent and instructions for one process.

You will work on

  • Choose recurring questions and exclude unauthorised actions.
  • Review source versions, owners and access.
  • Create instructions bounding the objective and knowledge gaps.
Practice
Create the scope and prepare a fictional guide, procedure and notes.
How you will check it
Initial agent and inventory with a version, owner and use for each source.
02

Answer and recognise limits

Knowledge, references and clarification. Prepare a reviewable request when information or approval is missing.

You will work on

  • Check each answer against its reference.
  • Request essential details for an incomplete request.
  • Leave a pending summary without claiming activation.
Practice
Answer questions and prepare a request without inventing missing details.
How you will check it
A checked answer and conversation with explicit gaps.
03

Test, improve and prepare the handover

A short set, changes and individual evidence. Distinguish a prototype from operational publication.

You will work on

  • Test answerable questions, gaps and out-of-scope requests.
  • Update a source or instruction and repeat cases.
  • Document dependencies, ownership and publication conditions.
Practice
Complete the set, resolve an individual variation and leave a handover.
How you will check it
A record and maintenance guide separating actual results from prepared examples.

Answer with a source and keep approval visible.

Classroom example with fictional data

Build a classroom agent to consult an internal tool guide and access-request process. Check references, missing details and an old instruction; prepare a summary pending review without creating accounts or granting permissions. Use fictional documentation and an agreed environment.

Fictional classroom case. Brisa has a current guide, a procedure and support notes. The agent helps find instructions and prepare requests; it does not create accounts or change permissions.

Starting point
“I need access for Atlas” does not identify the tool or its purpose.
First reading
Broad instructions assume access and confuse an old guide with the current one. This is a prepared answer for analysis, not a real execution.
What we review
Ask for the tool and purpose, consult the current source and keep approval pending. Route knowledge gaps without invention.
Result
The example leaves a reviewable summary with details or gaps, a procedure reference and no claimed activation.

Check before drawing conclusions.

Ask for the missing detail

Before review

“I need access for Atlas” becomes activation of an unspecified tool.

After review

“Which tool do you need and for what task? I will prepare a summary for review.”

The procedure requires tool, team and purpose. Clarification is not approval or execution.

Uploading a source does not validate it

A learning example, not a Studio execution. Guide v2 requires approval; v1 is archive.

Uploading a source does not validate it
CheckEvidenceDecision
CurrencyInventory: v2 current, v1 archivedDo not add v1 as valid knowledge.
AnswerReference to v2’s approval sectionCorrect an answer unsupported by its source.
AccessSource type and tested userAn uploaded file does not prove SharePoint permissions.

Swipe to see all columns.

Repeat cases after a change and record what was actually checked.

What we cannot conclude

  • Check a link against its content.
  • Publication and user access require their own preparation.

Before calling it complete

  • The current version is distinguished from the archive.
  • Essential missing information is requested or kept pending.
  • A person retains the approval decision.

Sessions to put it into practice.

Reference edition · 12 h live

Reference edition: 12 live hours across 6 sessions of 2 hours. Individual practice and short feedback take place in the sessions, without additional mandatory tasks.

  1. Session 12 h

    Define and create the prototype

    Agree task, questions, boundaries and owners; create the agent in the prepared environment.

    By the end

    Scope and initial instructions.

  2. Session 22 h

    Prepare the knowledge

    Review and add fictional sources; check the first answer and reference.

    By the end

    An inventory and checked answer.

  3. Session 32 h

    Handle gaps and changes

    Test an ambiguous question, missing details and an old instruction.

    By the end

    Revised rules and a documented correction.

  4. Session 42 h

    Prepare a reviewable request

    Collect essential details and produce a summary pending approval. Do not grant access or send messages.

    By the end

    A request with details or explicit gaps for human review.

  5. Session 52 h

    Test and check again

    Repeat cases after a change and check identity and sources where the environment permits.

    By the end

    Case-level evidence and comparison.

  6. Session 62 h

    Resolve and hand over

    Individual variation and source, ownership and boundary documentation. Feedback in the session.

    By the end

    Individual evidence and handover; publication has separate conditions.

If your team needs to go further

  • A Studio-specific programme, separate from everyday Copilot use in Microsoft 365.
  • Licences, tenant setup, complex connectors, autonomy and production deployment require a separate scope.
  • Publication and user access depend on licences, permissions, policies and their own tests.

An individual piece of work to close.

Configure a bounded prototype, check answers against sources and record an individual variation. Leave a handover with outstanding work; trainer review and feedback take place within the 12 hours.

Scope and prepare

Pass
Defines task, boundaries, current sources and owners in an authorised environment.
Needs review
Uses unreviewed documentation or leaves actions unbounded.

Answer with support

Pass
Checks references, asks for essential details and routes gaps.
Needs review
Cites an unsupported answer or invents details.

Test and correct

Pass
Records actual environment results, repeats the set and resolves an individual variation.
Needs review
Copies examples as executions or does not check the change.

Leave a handover

Pass
Documents ownership, sources, tested access and outstanding work.
Needs review
Promises deployment or permissions that were not tested.

The trainer reviews the task and explains the result within the agreed course time. Additional reassessment is agreed for each edition.

Learn with someone who develops projects for companies.

Borja Martínez

Borja Martínez

The Optimal Flow consultant

Over two years of experience developing projects for companies across industries and sizes.

We confirm the case, practice setup, dates and trainer availability when preparing each edition with your team.

Materials to practise and leave evidence.

Fictional sources and cases, blank templates and worked examples. Actual results are recorded during the edition.

  • Practice guide and timings
  • Fictional sources and identified archive
  • Blank scope and inventory
  • Initial and revised instructions
  • Questions and expectations
  • Blank test record
  • Explained review example
  • Blank handover
  • Worked handover example
  • Individual variation
  • Assessment rubric
  • Reference answers and checks

With guidance, practice and support.

Each delivery includes working materials, feedback on practical work and assessment against the agreed objectives.

See materials, support and terms

Materials you can keep

Downloadable, versioned guides, exercises and review criteria for internal use.

To prepare and complete the delivery

A short recommendation, an editable communications kit and a group learning report.

TOF’s own certificate

A certificate of achievement after the agreed assessment. It is not a vendor certification.

A month of email support

On the content covered, up to two accumulated hours for the whole group.

What does your
team need to learn?

Tell us what you want to work on, how many people would take part and where you are based. We will prepare a proposal with scope, sessions and a budget.

Would you like to ask about FUNDAE?

We check eligibility for each company and edition, subject to requirements and available training credit in Spain. Funding is neither automatic nor guaranteed. Official FUNDAE information ↗

Discuss eligibility
The Optimal FlowlearnMicrosoft Copilot Studio

Build internal agents with Microsoft Copilot Studio

Build and test an agent that answers process questions using authorised documentation, with references, clear boundaries and a review the team can continue.

12 live hours6 sessions of 2 hoursIn Spanish or English

Reference edition: 12 live hours across 6 sessions of 2 hours. Individual practice and short feedback take place in the sessions, without additional mandatory tasks.

Who it is for

People in systems, operations or business teams who understand a process and will configure an agent, with Microsoft administrator support to prepare the environment.

By the end you will be able to

  • Define an internal agent’s objective, useful questions, sources and boundaries.
  • Prepare authorised knowledge and check answers against references.
  • Handle clarification or routing without claiming actions the agent has not performed.
  • Test cases and gaps, and leave a handover with owners and conditions for possible publication.

Before we begin

Corporate identity and authoring access in Microsoft Copilot Studio, an authorised practice environment and administrator support. Confirm licence, features, sources, permissions and usage. The free trial supports creating and testing, not publishing; licence purchases and tenant setup are not included.

Check identity, authoring, environment, features, policies, sources and usage. A free trial does not allow publishing. Uploaded files do not demonstrate individual SharePoint permissions: that needs prepared sources and test users. Classroom material does not connect real documents or change admin policies.

Your trainer

Borja Martínez
The Optimal Flow consultant

Over two years of experience developing projects for companies across industries and sizes.

We confirm the case, practice setup, dates and trainer availability when preparing each edition with your team.

The Optimal FlowlearnBuild internal agents with Microsoft Copilot Studio

Programme sessions

  1. Session 1 · 2 h

    Define and create the prototype

    Agree task, questions, boundaries and owners; create the agent in the prepared environment.

    Result: Scope and initial instructions.

  2. Session 2 · 2 h

    Prepare the knowledge

    Review and add fictional sources; check the first answer and reference.

    Result: An inventory and checked answer.

  3. Session 3 · 2 h

    Handle gaps and changes

    Test an ambiguous question, missing details and an old instruction.

    Result: Revised rules and a documented correction.

  4. Session 4 · 2 h

    Prepare a reviewable request

    Collect essential details and produce a summary pending approval. Do not grant access or send messages.

    Result: A request with details or explicit gaps for human review.

  5. Session 5 · 2 h

    Test and check again

    Repeat cases after a change and check identity and sources where the environment permits.

    Result: Case-level evidence and comparison.

  6. Session 6 · 2 h

    Resolve and hand over

    Individual variation and source, ownership and boundary documentation. Feedback in the session.

    Result: Individual evidence and handover; publication has separate conditions.

A fictional classroom case

Answer with a source and keep approval visible.

Fictional classroom case. Brisa has a current guide, a procedure and support notes. The agent helps find instructions and prepare requests; it does not create accounts or change permissions.

By the end of the exercise: The example leaves a reviewable summary with details or gaps, a procedure reference and no claimed activation.

The Optimal FlowlearnBuild internal agents with Microsoft Copilot Studio

What we will work on

Scope the agent and prepare its knowledge

  • Choose recurring questions and exclude unauthorised actions.
  • Review source versions, owners and access.
  • Create instructions bounding the objective and knowledge gaps.

Practice: Create the scope and prepare a fictional guide, procedure and notes.

Answer and recognise limits

  • Check each answer against its reference.
  • Request essential details for an incomplete request.
  • Leave a pending summary without claiming activation.

Practice: Answer questions and prepare a request without inventing missing details.

Test, improve and prepare the handover

  • Test answerable questions, gaps and out-of-scope requests.
  • Update a source or instruction and repeat cases.
  • Document dependencies, ownership and publication conditions.

Practice: Complete the set, resolve an individual variation and leave a handover.

Extensions agreed separately. A Studio-specific programme, separate from everyday Copilot use in Microsoft 365. Licences, tenant setup, complex connectors, autonomy and production deployment require a separate scope. Publication and user access depend on licences, permissions, policies and their own tests.

What accompanies the programme

  • A downloadable, versioned guide, exercises and review criteria; keep a copy for internal use.
  • A short recommendation, editable kit and group learning report. Responsibilities, a prior access check, handover and recipients are agreed for each delivery.
  • TOF’s own certificate of achievement after passing the agreed assessment. It is not a vendor certification.
  • Email support on the content covered for one calendar month after the last session. A maximum of two accumulated hours for the whole group, with a response time of up to two working days.

Clients provide licences unless expressly agreed otherwise. Recordings are agreed in advance. Implementation, work carried out independently by TOF, production deployment and ongoing maintenance are quoted separately. Delivery link access is agreed per delivery. Later updates and specific support require agreement. Structured documents and text alternatives; participation needs reviewed before starting.

Full format and terms → theoptimalflow.com/en/training/how-we-teach