Make your people AI-native, and prove it on their work.
Your teams learn on work shaped like their own. What they build is scored against a published rubric, and every credential is decided by a person who puts a name and a reason on the record. You watch the programme move by what people made, not by who attended.
- Role paths
- A path for each role. Every module ends in something the person builds.
- Reviewed evidence
- A rubric score on every submission. Human marks on the work that counts.
- Decisions of record
- Credentials decided by a named person, with the reason written down.
- The Program desk
- Your cohorts, certifications and people, with figures frozen each week.
Every credential sits at the end of a chain you can inspect.
A model gives the first read because it is fast and consistent against a rubric. The credential itself is always a person's decision. The steps run in the same order for everyone.
- 01
Mission
Each person works the path for their role: developer, product manager, project manager or forward-deployed engineer. A module sets the problem and names the artifact it expects.
Role paths - 02
Evidence
They submit what they built. A model scores it against the module's published rubric, 0 to 4 per criterion, and the record says which reviewer read it.
Artifact review - 03
Review
Assessed work goes to two assessors who mark it blind. A two-point gap on any criterion goes to a third. A technical and a business panellist must agree.
Assessor desk - 04
Readiness
Each credential names the work it requires, and a requirement closes only on reviewed work. When every one is backed, the claim is ready to decide.
Credential gate - 05
Decision
A named person decides and writes down the reason. Nobody decides their own claim, and the decision cannot be edited once it is recorded.
Decision of record
The Adaptive Path reads the person's latest live assessment and their reviewed work, then reorders what they should do next. It plans; it never marks anything complete. Only a reviewed artifact moves someone forward.
Your programme, on one desk.
Your facilitators run the programme from the Program desk. It reads the roster you already manage and a set of figures that freezes when each week closes, so nobody can watch a number move and work out who moved it.
A facilitator sees seats, participation and whether a baseline exists. Never a score, a level, a response or a named piece of evidence. Figures about people come from at least five people who chose aggregate reporting, and any count under five reads “fewer than five”.
The model scores. People decide.
A credential here is worth what it refuses. That only holds if the rules are written down before the first submission is read, and if a person answers for every credential.
Published rubrics
Every module shows the criteria its artifact is scored against before anyone submits. When no model can be reached, the submission is kept with no verdict. An offline check can never pass work.
Blind marks
Each assessor's mark stays sealed until both are in. A disagreement is adjudicated and recorded, never averaged away. The panel's technical and business members each record the decision independently.
A decision of record
The decision carries the name of the person who made it and a written reason. It is appended to a register the database will not let anyone rewrite. Anyone can check the credential's status and the requirements it met at /verify.
Appeals
The person can appeal within ten working days. The re-mark goes to an assessor who has seen neither them nor the original marks, and the original decision stays on the record beside the outcome.
Access to your code is written down before it is technical.
Training has no connector into client repositories and no route for taking in client source. People submit their own work as evidence, and those evidence fields are not a channel for client code. If an engagement ever needs a repository, the terms come first.
Six terms in writing before one repository.
Written scope must name every control below before any approved repository is opened. Client source and client artifacts are not reused outside that engagement without separate written permission.
- 01Processing location
- 02Authorized people and systems
- 03Retention and deletion window
- 04Egress rules
- 05Subprocessors and providers
- 06Incident and termination handling
Scores belong to the person.
Facilitators see seats, participation and whether a baseline exists. Aggregate reporting is each person’s own revocable choice, and a figure about people needs at least 5 of them.
Platform controls
- TLS on every request and security headers at the edge.
- Opaque server-side sessions in an HttpOnly cookie, and server-side role checks on protected account and organization routes.
- Regulated data such as PHI stays out by default. Working with it needs a separate agreement that sets the data terms.
- Artifacts are never used to train third-party models without the person's explicit consent.
A conversation first. Then a cohort.
Start with one cohort, in the roles where AI is already changing the work.
- 01
Tell us who, and what is in the way
The roles you want to move and the operating constraint that matters most. The form below is enough to begin.
- 02
A working session
Together we choose the role paths, the first cohort and its facilitators, and agree dates and access.
- 03
Terms in writing
Scope, commercial terms and data handling are written down and signed before anyone starts.
- 04
Invitations and placement
Each person accepts an invitation into your cohort, then sits the placement assessment so the work starts where they are. Your facilitators see the cohort on the Program desk from the first day.
Tell us the roles and the constraint.
A few lines are enough. We will come back to set up the working session.