Attendance is not capability
Develop people who can take an unclear AI idea, understand the work behind it, test the simplest sensible change and make a recommendation you can read.
Two days in a room, then applied work on a real problem in your organisation. The two days develop the capability. The applied work proves it.
A Head of Talent and an AI trainer, with twenty five years working within People teams across technology, financial services, education, sport, fashion and retail.
As an experienced AI Champion, he understands the work you do day to day and can turn it into practical examples of where AI boosts performance and improves efficiency.
He sat on the advisory board of a 15,000 member HR network, alongside senior leaders from Lego, Santander, LV and Greene King.
The person colleagues bring an idea to
An internal practitioner who can diagnose, scope, prototype, test, evidence and route work to the right specialist.
A Champion is the buffer between "somebody had an idea" and "Engineering has a ticket". They work out what the job actually costs today, whether AI is even the right answer, and what evidence would settle it.
- an engineer
- a security approver
- legal counsel
- your data protection officer
- the owner of every process they touch
- the person who says yes to AI
The sentence they leave able to say: "I do not have the authority to approve this, but I know who needs to review it and what they will need."
Rehearse safely. Then apply it to real work.
Participants first rehearse on controlled Northstar cases where the correct behaviour is known and the exercise is designed to expose failure. Only then do they move onto an approved problem from your own organisation.
Rehearse where failure is safe
Northstar Group is a fictional company with documents built for this. The cases include incomplete information, two sources that disagree, and requests the assistant should refuse. The right answer is known in advance, so somebody can get it wrong in a room and find out why.
The parts of anything that runs
What starts it, what it may read, the steps, the instructions it always follows, what it produces, where that goes, what happens when something is missing, and who signs it off. Plain words, then mapped to the tools your company already has.
Diagnose before you build
The problem, who feels it, how long it takes and how often, and a baseline that can be measured again later. No baseline, no prototype.
Their own problem, prototyped
An approved real problem from your organisation, shrunk to the smallest version that tests the idea. No reward for complexity.
Testing that looks for failure
They write down what should happen before they run anything, then test normal, difficult, incomplete and edge cases, and the ones it should refuse. Then they try to break each other’s work, fix what fails, and test it again.
Handover and the recommendation
A pack somebody else can run it from, the risks and who needs to review them, and a recommendation: Scale, Change or Stop, with the evidence attached.
Building it is only half the exercise
Participants write down what should happen before they run a test. They test normal, difficult, incomplete and edge cases, deliberately try to break each other’s work, fix what fails and retest it.
A polished AI answer is not accepted as proof that the work works. The question asked in every exercise is the same one: how do you know it is right?
AI is allowed to be the wrong answer
The goal is not to maximise the number of AI projects. A well-evidenced decision not to build something is a successful outcome.
Later, once something has been tested, the same discipline gives the recommendation: Scale, Change or Stop.
Where most training quietly ends
Capability that is never applied is capability you paid for and did not get. The follow-through is part of the programme, not an upsell.
Nine things you can read
Not a certificate and a set of slides. These are the artefacts a sponsor can pick up and check.
- A problem brief
- A baseline, with how it was measured
- A map of the current workflow
- The proposed workflow, with human checkpoints
- A test card, and the failures it found
- The fix, and the retest
- A risk and escalation map
- A handover pack
- A Scale, Change or Stop recommendation
Earned on a capstone, not on attendance
The two day programme, then 30 to 60 days applying it to a real business problem, gathering the evidence and defending the decision. That assessment is what carries the certification. The two days on their own do not.
- Define the problem, the people it affects and what is out of scope
- Establish a baseline, with at least two measures. No baseline, no prototype
- Diagnose the opportunity, including whether a simpler fix would do
- Build the smallest version that tests the hypothesis
- Test it against at least ten realistic examples, including cases it should refuse
- Write the risks, the limitations and who needs to review them
- Recommend Scale, Change or Stop, following the evidence
The programme, or the assessed pathway
Applied AI Champion Programme. Two days. Develops the capability and ends with a tested piece of work and a recommendation. Attendance is not certification.
Certified Applied AI Champion. The same two days, then 30 to 60 days applying it to a real problem, with evidence and a formal assessment. The certification comes from that assessment.
The cohort is up to twelve either way, in person or online. It stays at twelve because it is people working on their own problems with help.
Score your people against the Champion standard first. Seven competencies, four levels, ten minutes each. It will tell you whether you need the programme, and which competencies to aim it at.
Champions need somewhere for the work to go afterwards. The PROVEN Framework™ is the method around it, and TransformLayer is where the cases, decisions and evidence live.