Who this is for
Business teams, L&D and people teams.
Start with the challenges your team faces and each person’s current skills.
Plan ChatGPT training around approved sources, practical team tasks, documented review and evidence from a second assessment. Includes a sourced Morgan Stanley example.
ChatGPT training workshops: Practise a repeatable ChatGPT workflow using task briefs, source checks and reviewed outputs.
Business teams, L&D and people teams.
Start with the challenges your team faces and each person’s current skills.
A task brief, reviewed draft and correction log.
This is an example to help scope the learning; delivery arrangements are agreed before booking.
Specify the reader, source, format and boundaries.
Produce a draft and compare its claims with the source.
Record errors and request a targeted revision.
Agree a suitable next task and reviewer.
Use these stages to discuss the learning scope. Timing, tools and exercises depend on the team’s brief.
Specify the reader, source, format and boundaries.
Produce a draft and compare its claims with the source.
Record errors and request a targeted revision.
Agree a suitable next task and reviewer.
Confirm approved account access and bring a recurring task, fictional or approved source material, and the person who reviews the output.
Estimate the time your team needs for sessions and follow-through. Change the example to fit your people; this is not a workshop package or quote.
Compare programs against your challenge, audience and available time. Confirm the actual schedule and fees with the organizer.
Find relevant programs in AISFYYou may need to sign in. Results stay on this page and are not saved or transferred to your AISFY account. Copy your plan before continuing.
Cohorts = participants ÷ cohort size, rounded up. Per-person time = session hours + weekly practice × weeks. Team time multiplies that by participants. Delivery hours count repeated sessions across cohorts; they exclude facilitator preparation.
Defaults are illustrative inputs, not benchmarks or recommended commitments. Changing or leaving the page resets the plan.
Open a topic for the full guidance, examples and references.
OpenAI reports that more than 98% of Morgan Stanley advisor teams adopted its internal AI assistant. The case describes expert evaluation and feedback. OpenAI — Morgan Stanley customer case.
Evidence limits: This is a vendor-published adoption figure for an internal system, not an independent productivity study or a result for a standard ChatGPT account.
The Morgan Stanley example makes evaluation part of the operating method. A useful training brief therefore specifies the material staff may use, the decisions they may support, and who approves the result. A demonstration that produces fluent prose does not test these responsibilities.
Use a fictional customer renewal: the notes contain a service date, an unresolved complaint and no approved discount. Ask for a draft response. The reviewer must confirm the date, preserve the unresolved status and reject any invented offer. Record factual corrections separately from tone edits; a polished response with an unauthorized discount fails.
Sources checked 20 October 2026. External examples are not AISFY customer results. Practice scenarios are editorial examples.
Participants identify seeded errors, explain corrections and recognize when the output should not be used.
The watch-item: verify facts before anything leaves the building. ChatGPT is confident by default, including when it is wrong, so review habits are part of the skill — not an optional extra.
A few clever interactions are not a capability. Teams need a simple operating system for defining work, and the five-part structure works in every case: Role (tell it who it is, within policy), Task (define the outcome), Context (what it needs to know — never sensitive data), Format (specify the output so it is easy to review), and Guardrails (what to avoid: confidential information, unsupported claims, decisions that belong to humans). Adding examples and constraints raises quality further — most weak outputs trace back to vague requests.
The training structure we use runs in three moves. Align: connect the session to real productivity goals — participants name the recurring tasks that eat their week. Equip: hands-on practice across those actual tasks — writing, planning, research, analysis, review — with data sensitivity embedded in every exercise, not appended at the end. Measure: capture confidence shift and the workflow each participant will improve, then follow up at 30, 60 and 90 days to see whether adoption held.
The single highest-leverage closing exercise: each participant identifies one recurring task, builds a reusable skill or workflow for it, and defines the review rule they will apply before sharing its output. Personal skills and use-case libraries — by role and by recurring deliverable — are what make the difference between a workshop and a capability.
Not attendance records. Ask for: confidence scores by workflow, the list of workflow improvements identified per team, evidence that review habits are being applied, and whether the skills and use-case libraries are still in use a month later. If those are missing, the training was a demo.
Use an approved account and a fictional source pack to create a management brief. Require the draft to separate facts, assumptions and questions. Participants should verify the output against the source and remove unsupported claims before it is shared.
Check the features and administrator settings available to the actual participants. Do not design an exercise around a capability available only in the facilitator’s plan. If a task involves files or connected information, confirm the organization’s permission and data requirements first.
Ask participants to repeat the method on a different source pack and explain what changed. A good result on the demonstrated example does not prove the method transfers. Keep the task brief, corrections and review decision as the useful learning record.
When comparing effort, count source preparation and review as well as drafting. Choose the next workplace task from what the team can evaluate reliably, not from the most impressive demonstration.
For workshop options, explore enterprise AI workshops. Use the preparation checklist and evaluation guide to connect the session to a practical next step.
Choose a recurring task and write down six elements: the intended reader, the decision or action needed, approved source material, output format, boundaries and acceptance checks. For a customer update, that might mean a short factual email using a service record, with no invented completion date or compensation offer.
Ask participants to produce a draft, annotate unsupported statements and revise it. Save the corrected example alongside the task brief so colleagues can see what good work looks like. A reusable brief needs an owner and a review date when source material or policy changes.
Use the same quality criteria before and after assistance: factual accuracy, completeness, clarity and suitability for the intended reader. For consequential decisions, the appropriate qualified owner retains responsibility.
Change the input: remove a key date, include a conflicting figure or add an unfamiliar exception. Ask the participant to identify the gap and decide whether to seek clarification, narrow the output or stop. This reveals whether the workflow depends on an unusually clean example.
Record drafting, checking and correction time separately. Agree a small supervised workplace task and a manager review before extending the method. Retain examples of failures as well as accepted outputs; they show colleagues where the method’s boundaries are.
Business teams working with documents, internal communication and recurring summaries can use a shared exercise to practise task definition and output review. Managers and L&D colleagues can participate to agree how learning will be supported afterwards.
Bring the team’s current experience, approved account arrangements and one recurring task to the scoping conversation. Participants do not need coding skills for the indicative activities.
Start from a fictional source pack. Describe the intended reader, output format and information boundaries, then produce a draft and check each factual statement against the source. Record omissions and unsupported claims before requesting a targeted revision.
A second scenario can test whether the participant recognizes a missing policy answer or an unauthorized commitment. The learning evidence is the reviewed draft and explanation, not simply the amount of tool activity.
Use an organization-approved account and confirm the features available to participants. Access, licences and organizational permissions are not assumed or included automatically. Fictional documents can be used when real materials are not approved.
No live sending, publication, purchases or system changes are required for these exercises. Integrations, custom development and organization-wide deployment are outside this indicative learning scope unless separately agreed.
AISFY is based in Dubai. Share your team’s location, preferred delivery format, language requirements and approximate group size when enquiring. Dates, facilitator, format, fees and final agenda are confirmed before booking.
Plan a suitable supervised task with a named reviewer after the workshop. Demonstrated learning should remain separate from claims of productivity, financial return or business-outcome attribution.
Account access and licences must be confirmed during scoping. No licence entitlement is implied by this workshop description.
No certification or OpenAI affiliation is claimed. This is an AISFY workshop using approved tools and independently developed learning exercises.
Only information specifically approved by your organization for the chosen tool may be used. Fictional material is sufficient for the indicative exercises.
The intended outcome is a bounded task brief, a source-checked draft and an explanation of review decisions. Actual progress depends on practice and is assessed separately from business results.
For hands-on practice, participants need access to the approved account and features selected for the exercise. Confirm access beforehand and use fictional or approved material. A guided demonstration can introduce a method, but an independent task is needed to assess the participant’s skill.
Choose fictional or explicitly approved inputs, remove unnecessary sensitive details and follow the organization’s information rules. An available upload feature is not permission to upload a document. Ask the responsible owner when the boundary is unclear.
Compare a representative task using the tools your organization permits. Check source fidelity, review effort, available features, accessibility and cost. Choose based on observed task fit rather than assuming one brand is best for every team.
Preview the work · illustrative resource
Inspect a deliberately flawed draft against fictional source notes. This illustrates a review method, not output from a recorded ChatGPT session.
For: Teams practising summaries with ChatGPT or Gemini
Requirements were reviewed, not approved. Testing is waiting for a sample dataset. Friday is a proposed date; the owner has not confirmed it.
Deliberately flawed draft: “Requirements are approved. Testing will finish Friday. No support is needed.” It changes the approval status, turns a proposal into a commitment and hides the dependency.
Reviewed example: “Requirements have been reviewed; approval is not confirmed. Testing awaits a sample dataset. Friday remains proposed. The project owner should confirm the date and identify who can resolve access.” Check each claim against the source before sharing.
Ask for relevant subject experience, an example exercise and the proposed review approach. Check the facilitator’s named profile and agree who will assess the work. A profile or attendance certificate alone does not demonstrate results for your team.
Confirm the audience, permitted information, tools, accessibility, format, dates, fees, practical outputs and follow-through in writing. The examples here are learning materials; they are not client case studies or fixed delivery promises.
Read the AISFY founder profileCaptioned screen walkthrough · no audio · fictional example.
Read the source. Requirements were reviewed, not approved. Testing is waiting for a sample dataset. Friday is a proposed date; the owner has not confirmed it.
Challenge the draft. Deliberately flawed draft: “Requirements are approved. Testing will finish Friday. No support is needed.” It changes the approval status, turns a proposal into a commitment and hides the dependency.
Keep the correction traceable. Reviewed example: “Requirements have been reviewed; approval is not confirmed. Testing awaits a sample dataset. Friday remains proposed. The project owner should confirm the date and identify who can resolve access.” Check each claim against the source before sharing.
Share the challenge, audience and skills you want to develop. AISFY helps you plan the learning and connect with relevant expertise.
Scope, facilitator, availability and fees are confirmed before booking.
Explore AI workshops ↗