Business teams and non-technical project teams — 1 day (7 hours), in house, on your own documents. Maximum 8 participants.
Everyday generative AI: the uses that save time, and the confidentiality rules that protect your data
At a glance
- Duration
- 1 jour (7 heures)
- Format
- On site and remote
- Group size
- 8 participants maximum
- Level
- Beginner
- Audience
- Owners and managers of small and mid-sized companies who want concrete uses rather than a demo
- Prerequisites
- No technical prerequisite. No prior knowledge of generative AI is required.
- Methods
- The day runs on the participants' real documents: no fictional case as long as real material exists
- Assessment
- Beforehand: a positioning questionnaire sent to participants (recurring tasks, tools already in use, existing internal rules), used as the comparison point at the end of the day
- Lead time
- Scheduled during framing, to suit your availability
- Pricing
- On a quote basis, depending on format, duration and number of participants. The first call is free.
- Accessibility and disability
- See the dedicated section
In most organisations generative AI is already in the building: a few people use it, with no written rule, often with documents that should never have left. Elsewhere, licences were bought and nobody touches them. This day handles both situations at once. We sort the uses that genuinely save time from the ones that waste it, we put in writing what may be submitted to a tool and what must never be, and we work on your real material: workshop notes, meeting reports, long emails, requirement statements. By the end of the day each participant leaves with one recurring task chosen, a baseline time measured, and their team's usage rules, drafted together during the session.
Learning objectives
- Tell apart the tasks where generative AI saves time from the ones where it wastes it, starting from your own recurring work
- Sort a working document into three sensitivity categories: what can be submitted as is, what must be anonymised first, what never leaves the company
- Write a prompt containing its four components — role, material, output format, constraints — and producing a usable draft
- Turn a set of raw notes into a shareable report structured into decisions, open points and assigned actions
- Check an AI output against a review grid, and spot a false statement delivered with confidence
- Draft your team's usage rules on three levels: allowed, allowed under conditions, never
- Set up the measurement of time actually saved on one recurring task, over a two-week period
Programme
01Sorting it out: what works, what does not
1 hour- What a language model actually produces: plausible text, not verified truth
- The four uses that hold up day to day: cleaning up raw notes, drafting a first version of a document, challenging your own analysis, translating between business and tech
- The uses that do not hold up yet: estimating effort, prioritising on your behalf, producing a binding deliverable without review
- Why an AI-produced estimate is plausible and wrong at the same time, which is the worst possible combination
- Round table: each participant names two recurring, time-consuming tasks from their week, which become the raw material for the day
02The confidentiality framework, before anything else
1 hour 30- Three data categories: what can be submitted as is, what must be anonymised, what never leaves the company
- Recognising personal data, customer data or a trade secret inside an ordinary working document, including when buried in a long text
- Anonymising an extract before submission: a three-pass method (identities, identifiers, cross-referenceable details) and the limits of the exercise
- Company-approved tool versus personal tool: what the distinction changes in practice, and why it is not negotiable in a regulated environment
- Who remains accountable for the output produced and for what it states
- Workshop: each participant takes a real document they brought, sorts it, then prepares it for submission
03Writing a prompt that produces a usable draft
1 hour 30- The four components of a useful prompt: expected role, material supplied, output format, constraints to respect
- Showing an example of the expected output rather than describing it: the highest-yield shortcut there is
- Iterating over two or three exchanges instead of aiming for the perfect output first time
- What makes a prompt fail: vague request, missing context, two tasks asked at once, output format left unspecified
- Workshop: each participant writes a prompt for one of their two tasks, runs it, then fixes it against the result
04The highest-yield case: raw notes into a meeting report
1 hour 15- The real gain is not typing time, it is the write-up rate: the reports that would never have been written finally get written
- Imposing the output structure: decisions taken, open points, actions with an owner and a due date
- What AI cannot guess and must be given: the roles of the people mentioned, what was at stake, what was settled and what was not
- Assessed exercise: three pages of messy notes become a shareable report, reviewed and corrected during the session
05Checking before sending out
1 hour- The risk is not a typo, it is the false statement delivered with confidence in the middle of an otherwise accurate text
- A five-point review grid: figures, proper nouns, dates, commitments made, sources cited
- What must be checked every single time: anything numeric or factual
- Spotting a rephrasing that quietly changed the meaning of a business requirement
- Workshop: cross-review of the previous module outputs, hunting down unverifiable statements
06Leaving with a two-week plan
45 minutes- Picking one single recurring, time-consuming task per participant, rather than hunting for places to put AI everywhere
- Setting the measurement before starting: how long does this task actually take today
- Deciding after two weeks on the measurement, not on an impression: extend the use, or stop it
- Collective drafting of the team's usage rules, on three levels: allowed, allowed under conditions, never
- Validation quiz and handover of the take-away materials
Who it is for
- Owners and managers of small and mid-sized companies who want concrete uses rather than a demo
- Business teams spending significant time on recurring formatting work
- Non-technical project teams: framing, meeting reports, requirement statements, testing
- Organisations under confidentiality constraints: customer data, personal data, trade secrets
- Teams already using AI with no defined framework, who want to set one before it causes trouble
What you take away
- The three-category data sorting grid, filled in on your own documents
- Your team's usage rules, drafted in session: allowed, allowed under conditions, never
- A library of reusable prompts, produced by the group on its real tasks
- A meeting report template (decisions, open points, actions) and the prompt that fills it
- The five-point review grid, on a single page
- A two-week measurement plan, with the baseline time noted before starting
- The session materials, in a readable and printable format
Prerequisites
- No technical prerequisite. No prior knowledge of generative AI is required.
- Being comfortable with a word processor and a work email client.
- Having a workstation with browser access during the session.
- Bringing real working documents (workshop notes, a meeting report, a long email, a partial requirement statement): the day runs on them. Sample documents are provided otherwise.
Teaching methods
- The day runs on the participants' real documents: no fictional case as long as real material exists
- Short alternation: roughly twenty minutes of framing, then immediate hands-on practice at the workstation
- Each participant works on their own tasks, identified in the first module, not on a generic example replayed identically for everyone
- Commented demonstrations, failures included: wrong outputs are shown too, along with the reason they are wrong
- Collective drafting of the team's usage rules during the session, not in a document written up afterwards
- Group capped at 8 participants so everyone gets hands-on time and has their work corrected
- A framing questionnaire sent to the client beforehand, to fit the examples to the trade and to the confidentiality rules already in force
Assessment of learning
- Beforehand: a positioning questionnaire sent to participants (recurring tasks, tools already in use, existing internal rules), used as the comparison point at the end of the day
- Data sorting exercise (module 2): sorting ten document extracts into the three sensitivity categories, corrected and argued in session
- Assessed exercise (module 4): producing a shareable report from raw notes, validated against three observable criteria — decisions identified, actions assigned to a person, no unverifiable statement left in the text
- Cross-review (module 5): every output is checked by another participant using the five-point grid, then reworked
- End-of-day validation quiz: fifteen questions covering uses, confidentiality and verification, corrected collectively
- Four to six weeks later: a follow-up on the task being tracked and on the time actually measured, compared with the baseline noted in session
Where what I teach comes from
Adrien Bouthet, freelance IT Project Manager and Product Owner based in Bordeaux, with eight years of experience steering digital projects. I run this day because I apply these rules myself, every day, in an environment where they are not negotiable: since February 2025 I have been steering the digital onboarding journeys of La Banque Postale — account opening, minor and adult journeys, product subscriptions — on a budget in the order of one million euros, under constant KYC and GDPR pressure, including raising team awareness of data governance within my scope. That is where the confidentiality half of this training was forged: knowing what may be submitted to a tool and what must never leave the company is not a theoretical stance, it is a constraint I work under on journeys that are genuinely live. Those figures are results from steering assignments: they say where what I teach comes from, they are not a promise made to participants. The sorting between uses that work and uses that do not comes from my own practice as a project manager, which I published in detail in the article “Generative AI in project management: what actually works” — the four uses that hold up, the three that do not, and the three rules I impose on myself. I have also been teaching as a trainer with Bordeaux schools since 2020, in programming and management: teaching forces you to clarify your thinking, and that is what makes a method transmissible in a single day.
Real example
Systematic output checking is not a classroom principle for me: I industrialised it. On my video production chain, 79 compositions are generated in code and 6 stages run end to end — rendering, AI-generated subtitles, automatic quality control, publishing — without any editing software ever being opened, and with a control stage that runs before anything is broadcast.
Read the case studyFrequently asked questions
Do participants need technical skills?+
No. The programme is written for business teams and non-technical project teams: no code, no tool configuration. If your need is about building AI applications (API integration, agents), that is a different subject, which I handle as an assignment rather than in this session.
We have no company-approved AI tool yet. Is that a blocker?+
No, and it is often the right moment. The confidentiality module exists precisely to decide what is acceptable before a tool is rolled out widely. Participants leave with written rules, which makes the later tool decision far easier to settle.
Can we really work on our own documents if they are confidential?+
That is the whole principle of the day, and it is why the data sorting module comes before any production workshop: sort first, submit second, and some documents are never submitted at all. Documents brought in remain company property and are neither kept nor used after the session.
Will AI replace the people I am training?+
From what I observe, no: it removes part of the formatting work, not the judgement. Prioritisation and trade-offs depend on stakes you will never fully put into a prompt. Decisions stay human, and that is where the value of the work sits.
Is a single day enough?+
To install a few solid uses and a confidentiality framework, yes: the day is deliberately restricted to a small number of cases, each taken all the way through. It does not cover advanced automation or application design, which call for support over a longer period.
On site or remote?+
Both. On site in Bordeaux and across Nouvelle-Aquitaine, at your premises; remotely anywhere in France, with the same workshop format and the same work on your documents. A mixed format works when part of the team is spread across several locations.
How many participants at most?+
Eight. Beyond that, not everyone gets hands-on time and correcting work in session becomes impossible to sustain within a single day. For larger headcounts we schedule several sessions.
Framework and terms
Funding
These sessions are invoiced directly to the company, out of its own skills development budget. They are not eligible for pooled funding schemes (OPCO, CPF, France Travail), which require Qualiopi certification.
Certification
This session is not a certifying course. I do not deliver any Scrum.org, Scrum Alliance, Scaled Agile (SAFe) or PeopleCert (ITIL) certification, nor any qualification registered with the RNCP or the Répertoire spécifique. The goal is operational: practices you can apply, not a diploma.
Contractual framework
Each session is covered by a professional training agreement (convention de formation professionnelle) signed before the start, setting out the objectives, the programme, the duration, the delivery arrangements and the agreed fee. The internal rules applicable to the session are handed to participants at the opening.
Certificate of attendance
A certificate of completion is handed to each participant at the end of the session. It states the objectives, the nature and the duration of the training, along with the results of the assessment of learning.
Accessibility and disability
Disability referent: Adrien Bouthet, contact@adrienbouthet.fr. Any situation requiring an adjustment (access to the premises, adaptation of the materials, of the pace or of the assessment arrangements) is examined during the framing phase, within 5 working days. If the adjustment exceeds our means, we refer to specialised resources (Agefiph, Cap emploi).
Framing how AI is used in your teams
Book a free first call. We look at your real tasks and your confidentiality constraints, and I tell you straight whether a one-day in-house session is the right format for you, or whether your need is something else.
The matching service
Generative AI integration