Organise an AI workflow

How can you integrate AI into a business process without losing control over the steps or the final quality?

1960 € per session, excl. VAT

In short

A 7-hour day, excluding breaks, to organise an AI-assisted business process: define inputs, divide tasks between the human and the assistant, link successive outputs and check the results. Participants produce a documented operational procedure, then measure its quality across several fictitious cases.

Why this training?

A succession of requests to AI does not constitute a reliable workflow: incorrectly extracted information can affect the assessment and then the final response. Defined inputs and intermediate checks make these errors detectable. The challenge is to decide where AI genuinely supports the work and where a human decision remains essential. The training links each step to an expected output, a responsible person and an acceptance rule. Running case study: The fictitious organisation Avelune Équipements receives requests to replace professional equipment. Using fictitious emails and a fictitious catalogue, participants must assess each request, identify missing information and prepare a response to be submitted to the sales manager before sending. Application at work: Over the following two weeks, revisit a process from your own role using authorised, anonymised case files, use the checklist and compare defects before and after revising the instructions. Have responsibilities approved before any routine use.

Objectives of this training

  • Map a process from receipt of a request through to the approved deliverable.
  • Assign each task to the human or to AI according to its risks.
  • Write instructions that can be linked together, with explicit inputs and outputs.
  • Run the workflow on complete and incomplete case files.
  • Measure defects and document an improvement to the instructions.

Who is it for and what are the prerequisites?

Roles and tasks concerned

  • Operational managers seeking to organise recurring case processing with AI.
  • Coordinators, support staff and subject matter specialists who already use a generative assistant.

Prior skills

  • Be able to formulate a structured request to a generative assistant and review its response.
  • Know the steps and approval rules for a process within your own role.

What you will work on

  1. Define the scope of the business process

    Objective: Establish the boundaries and the expected outcome. Trigger, recipient, available documents, pain points and success criteria.

  2. Allocate tasks and decisions

    Objective: Choose appropriate AI interventions. Extraction, comparison, drafting; decisions reserved for humans, confidentiality and stop points.

  3. Prepare handovers between steps

    Objective: Prevent information loss between successive steps. Required fields, sources, output formats, contextualised instructions and handling unknowns.

  4. Run the complete process

    Objective: Produce a usable, checked case file. Manual sequencing in the authorised tool, intermediate checks and traceability of corrections.

  5. Measure and correct defects

    Objective: Improve the workflow based on observed discrepancies. Accuracy, completeness, compliance; counting defects and distinguishing between source errors and instruction errors.

  6. Formalise the reusable procedure

    Objective: Enable a colleague to take over. Execution order, instruction versions, exceptions and approval authority.

What you take away

  • Workflow map and responsibility matrix.
  • Instruction library with handover formats between steps.
  • Processed case file: assessment, request for further information and approved response.
  • Operational procedure and completed quality measurement checklist.

Method and assessment

  • Short teaching inputs followed by immediate practice on the running case study.
  • Individual work followed by work in pairs, with peer review of outputs.
  • Guided group discussion: what works, what needs correcting and what remains a human decision.
  • Preparation: Describe a recurring process, its deliverable and two frequently encountered errors, without confidential data. Ensure you have access to a tool authorised by the organisation; the exercises use only fictitious or anonymised case files.

Observable criteria: Each step specifies its input, output and responsible person. ; Every proposed product reference matches the catalogue provided. ; Missing information triggers a request for further details, without fabrication. ; The final response includes a recorded human approval. ; Defects are counted using the same checklist before and after correction..

Frequently asked questions

Do you need to know how to programme?

No. The steps are carried out manually; no connector or technical automation is required.

Can you use your own case file?

Only after anonymisation and internal authorisation. The fictitious case study allows you to complete all the exercises.

Can AI send the final response?

This process includes human approval before anything is sent; it does not automate distribution.

How do you assess an improvement?

By comparing defects in the same case files using an unchanged checklist, then checking an additional case file.

Last updated: