Automating with RPA and AI agents: identify, cost, steer
IT heads, business units, PMOs and process owners · 1 day (7 hours) · in-house, on site or remote
At a glance
- Duration
- 1 jour (7 heures)
- Format
- On site and remote
- Group size
- 10 participants maximum
- Level
- Beginner
- Audience
- IT heads and IT managers in small and mid-size companies asked by the business to 'put a robot on it'
- Prerequisites
- No technical background required: the day contains no robot development and no tool configuration.
- Methods
- Short talks followed by workshops on the real tasks participants bring.
- Assessment
- Beforehand: a positioning questionnaire (candidate processes, known volumes, robots already in place) that adjusts the examples.
- 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
An RPA robot does exactly what it was told, at the same speed on Monday morning and Friday evening, and stops dead the day the screen changes. An AI agent tolerates the unexpected, but sometimes gets it wrong with confidence. Between the two, most companies hesitate, launch an enthusiastic pilot, then discover an orphan robot nobody knows how to maintain. This day is for the people who decide and pay: IT heads, business units, PMOs, process owners. It does not teach you to build a robot. It teaches you to spot a task worth automating, to cost honestly the days saved and the maintenance effort, to choose between a rule-based robot and an AI agent, to frame a POC that proves something, then to steer the go-live and the supervision. Participants work on their own processes and leave with a costed file for the two or three most promising tasks.
Learning objectives
- Explain what an RPA robot can do, what it cannot, and what an AI agent changes at that boundary.
- Qualify a candidate task for automation on three criteria: volume, rule stability, exception rate.
- Cost the days saved and the build and maintenance effort of a robot, and present the result on one page management can read.
- Decide, for a given task, between a rule-based robot, an AI agent, both combined, or no automation at all.
- Frame a POC with a scope, a measurable success criterion and an end date written before starting.
- Organise a robot's go-live: named owner, supervision, procedure for when it stops.
- Recognise the three most frequent traps (changing process, moving screen, orphan robot) and set a countermeasure for each.
Programme
01What a robot can do, and what it cannot
1 hour- RPA in one sentence: software that replays clicks and keystrokes on existing applications, without changing them.
- What works: high volumes, stable rules, applications without an API. What breaks: judgement, exceptions, screens that change.
- Attended or unattended robot: two uses, two levels of risk.
- Commented demonstration of a robot running, then of the same robot facing a modified screen.
- Round table: the candidate tasks brought by participants, sorted on the spot.
02Spotting an automatable task
1 h 15- The three criteria that matter: volume, rule stability, exception rate. And the fourth one everyone forgets: who owns the process.
- Describing a task step by step, with its inputs, outputs and special cases, before talking about tools.
- Warning signs: the process changes every quarter, ten people do it ten different ways, nobody knows the volume.
- Workshop: qualification grid filled in on the tasks brought in, with a verdict per task: automate, document first, leave alone.
03Costing: days saved, build effort, maintenance effort
1 h 15- Measuring current time without cheating: timing, sampling, asking the people who do the task.
- Days saved per year: volume times unit duration, minus the time spent on exceptions that stay manual.
- The cost everyone forgets: maintenance when the target application evolves, supervision, the tool licence.
- Presenting the costing on one page: estimated gain, estimated cost, assumptions, what would flip the result.
- Workshop: costing of each sub-group's most promising task, peer review of the assumptions.
04Rule-based robot or AI agent: deciding
1 hour- What an AI agent adds: reading unstructured documents, tolerance to variation, decisions on unplanned cases.
- What it costs: a non-deterministic result, a human check to plan for, a bill that depends on usage.
- The four possible answers for a task: rule-based robot alone, AI agent alone, a robot that calls an agent on exceptions, no automation.
- When not to do it: low-volume task, process not yet stabilised, sensitive data without written usage rules.
- Workshop: reasoned verdict on each task brought in.
05Framing a POC that proves something
1 h 15- A POC has a scope, a quantified success criterion and an end date, written before it starts.
- Choosing the POC task: simple enough to finish in a few weeks, real enough to convince.
- Who does what: the business as process owner, IT, a contractor if any. And who decides at the end.
- What gets measured during the POC: real time saved, observed exception rate, incidents.
- Workshop: one-page POC framing note, on the task chosen by each sub-group.
06Going live, supervising, avoiding the orphan robot
1 h 15- From POC to production: access, environment, real data, business sign-off.
- Supervising a robot: knowing it ran, knowing it stopped, knowing who restarts it.
- The three traps: the process that changes without telling the robot, the screen that moves at the next application update, the orphan robot whose owner has left.
- A robot's identity sheet: owner, process, frequency, stop procedure, contact.
- Validation quiz, corrected in the room, and individual two-week action plan.
Who it is for
- IT heads and IT managers in small and mid-size companies asked by the business to 'put a robot on it'
- Business units (back office, accounting, HR, customer service) absorbing high-volume repetitive tasks
- PMOs and portfolio managers who have to arbitrate between several automation requests
- Process owners and project managers tasked with carrying a first POC
- Teams that already have a robot in production and do not know who maintains it
What you take away
- The qualification grid filled in on your own tasks, with a verdict per task.
- The one-page costing of the most promising task: days saved, build effort, maintenance effort, assumptions.
- The decision tree: rule-based robot / AI agent / no automation.
- A one-page POC framing note, ready to present.
- The robot identity sheet and the supervision checklist.
- Blank templates for these documents and the day's material.
Prerequisites
- No technical background required: the day contains no robot development and no tool configuration.
- Knowing at least one repetitive process in your area, with an idea of its monthly volume.
- Bringing two or three candidate tasks: they are the material for the qualification and costing workshops.
Teaching methods
- Short talks followed by workshops on the real tasks participants bring.
- Commented demonstration of a robot running, including its failure in front of a modified screen.
- One-page grids filled in during the session: qualification, costing, POC framing note, robot identity sheet.
- Sub-groups of three to four participants, with peer review of the costing assumptions.
- Field stories from RPA roll-outs in banking and industrial settings: what held, what broke.
- Tool-agnostic content: the grids work whichever vendor is chosen afterwards.
Assessment of learning
- Beforehand: a positioning questionnaire (candidate processes, known volumes, robots already in place) that adjusts the examples.
- During the session: the qualification grid and the costing are assessed against criteria handed out at the opening (measured volume, identified exceptions, written assumptions).
- The POC framing note is reviewed collectively against three criteria: bounded scope, quantified success criterion, end date.
- At the end: a validation quiz covering the six modules, corrected and discussed in the room.
- A few weeks later: a remote check-in, outside training hours, on the decision taken for each costed task.
Where what I teach comes from
Adrien Bouthet, freelance IT project manager and Product Owner based in Bordeaux, eight years steering digital projects. I have rolled out RPA, with UiPath and Selenium, on assignments at La Banque Postale and at EDF. In total, those automations represented between 100 and 200 days saved. That figure is a result from steering assignments, not a promise made to participants. That practice feeds the qualification grid and the costing method of this day. At EDF I also steered a portfolio of IT and industrial IoT projects, including a demand management application handling around 20,000 tickets a year: that is where you learn that a high-volume process is not always a stable one, and that the robot you put on it has to be watched like any other production component. I have also been teaching in schools since 2020, in programming and management: that is what taught me to make a practice transmissible in a single day.
Real example
The qualification grid and the costing method of this day draw on real RPA roll-outs, with UiPath and Selenium, at La Banque Postale and at EDF: between 100 and 200 days saved in total. EDF's industrial project portfolio, with its demand management application handling around 20,000 tickets a year, completes that experience: high volume does not guarantee a stable process. Results from steering assignments, not a promise made to participants.
Read the case studyFrequently asked questions
Do I need to know how to code?+
No. No robot is built during the session. The day is for the people who decide to automate, cost and steer. Building the robot is then handed to an internal team or a contractor, and participants know what to ask of them.
Which RPA tool is used?+
Demonstrations draw on UiPath and Selenium, because those are the ones I have rolled out on assignments. But the day is not software training: the qualification, costing and supervision grids work whichever vendor is chosen afterwards.
Is this an AI course?+
Not in the sense of personal use of assistants. Here, AI is treated as an automation component: what an agent adds compared with a rule-based robot, what it costs, when to rule it out. For individual uses of generative AI within a team, another day in the catalogue exists.
On site or remote?+
Both. On site at your premises, in Bordeaux and across Nouvelle-Aquitaine, or remotely anywhere in France, with the same sub-group workshops and the same demonstrations. A mixed format is possible when part of the team sits on another site.
How many participants at most?+
Ten. Costing and POC framing are done in sub-groups on real tasks, and each sub-group goes through peer review: beyond ten, the review time no longer fits in the day.
What happens after the POC?+
The day leaves you with a framing note and a decision to make. Building and going live belong to an assignment or an internal team. If you wish, I then step in to steer the POC or the go-live, separately from the training.
Our processes change often. Is that a blocker?+
It is precisely a criterion of the qualification grid. A process that changes every quarter is a poor candidate for rule-based RPA, sometimes a good candidate for stabilisation first, sometimes a case for an AI agent. The day teaches you to say so before spending.
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).
Repetitive tasks to cost
Book a free first call. We list your candidate tasks, pick those that will serve as workshop material, and set the date.
The matching service
RPA automation and AI agents