For practitioners
Apply the framework.
A working guide for senior leaders who want to run the PolyCognitive diagnostic and curriculum inside their organisation. The reference page covers what the framework is; this page covers what to do with it.
01
Who runs it
The framework is designed to be run by the most senior leader with authority over a function's AI strategy. In practice that is one of three seats:
CEO / Founder
Enterprise-wide application. Highest leverage; the literature finds the CEO-engagement multiplier larger than any other factor in AI value capture.
CHRO / Chief People Officer
Cross-functional application focused on the human-layer moves. Pairs naturally with CIO/CTO on AI-layer moves.
Function head / VP
Function-level application. Faster to run, lower scope, useful as a proof case before enterprise-wide adoption.
The framework is not a programme run by a transformation office. It is a working practice for the leader closest to the work. Delegating it to a champion or a steering committee reproduces the Level-0 stuck-point it was designed to clear.
02
The sequence
Five steps. Run linearly the first time. After that, the cycle repeats.
The framework runs on a 90-day cycle. The five steps below are the canonical sequence; after the first cycle each step continues in parallel with the next.
1
Step 1
Diagnose
Place the organisation on the five-level scale.
Run the scorecard with the leadership team. Reconcile differences before discussing levels with the wider organisation. Differences in placement between functions are normal and useful — they map where attention is most needed.
2
Step 2
Identify the bottleneck capability
Find the capability whose absence is producing the stall.
Each level is unstuck by one specific capability. If the diagnosis places the organisation at Level 2, the capability to develop is Capability 3 — Managing AI Risk. Resist working on later capabilities in parallel; they will not move the organisation until the earlier stuck-point clears.
3
Step 3
Run on both tracks
Human-layer move and AI-layer move, in parallel.
Every capability runs on two tracks. Stand up the human-layer move (identity, accountability, change practice) and the AI-layer move (technical literacy, mapping, failure modes) at the same time, not in sequence. Running one without the other produces commitment without competence, or competence without commitment.
4
Step 4
Measure outcome, not activity
Track the metrics the function cared about before AI.
The dominant failure mode at Level 1 is measuring tool usage and calling it value. Define success in cycle time, error rate, customer outcome, decision quality — whatever the function tracked pre-AI. Tool usage is a means; outcome is the result.
5
Step 5
Re-diagnose at 90 days
Confirm the move; identify the next bottleneck.
Re-run the scorecard with the same group at 90 days. The expected outcome is that the organisation has cleared the stuck-point and a new capability has become rate-limiting. If the level has not moved, the team is running one track without the other, or has confused activity for outcome.
03
Workshop structure
Three formats, scaled by depth.
The framework can be run as a self-administered exercise with the scorecard, a half-day diagnostic workshop, or a two-day intensive followed by a 90-day cadence. Each format trades depth for speed.
Format A · Self-administered
~ 6 min
Per leader, individually.
Each member of the leadership team takes the scorecard individually. Compare results in a 60-minute discussion. Lightest-weight entry point; best as a precursor to a fuller workshop.
Format B · Half-day diagnostic
~ 4 hr
Leadership team together.
Place the organisation on the diagnostic by capability and by function. Reconcile differences. Identify the bottleneck capability and the human + AI moves required to clear it. Produces a 90-day action plan owned by named leaders.
Format C · Two-day intensive + 90-day cadence
2 days + 90d
Full curriculum.
Cohort of senior leaders works through each capability with AI in the room, against real decisions. Followed by a 90-day follow-up cadence. This is the format Robert Blaga's practitioner programme is built on. Programme details.
04
Who to involve
For enterprise-wide application, the framework requires the full senior leadership team in the room — CEO, CHRO, CIO/CTO, CFO, and the heads of revenue-bearing functions. The two-layer structure makes the pairings explicit:
| Track | Primary owner | Supporting roles |
|---|---|---|
| Human layer | CHRO / Chief People Officer | Function heads, L&D, comms |
| AI layer | CIO / CTO / Chief AI Officer | Data leaders, security, vendor management |
| Convener & tie-breaker | CEO / Founder | Board, where relevant |
For function-level application, replace these seats with their function-level equivalents: function head, function HRBP, function technologist or solution architect.
05
What to measure
The framework does not prescribe a single metric. It prescribes a measurement discipline: the metrics that mattered to the function before AI are the metrics that matter to AI value capture now. The principle is that AI activity (prompts, tokens, tool licences) is not a deliverable.
Track these
- · Cycle time on the work the function actually does
- · Error rate or rework rate
- · Customer-facing quality measure
- · Decision quality / decision speed where AI is in the loop
- · Headcount-adjusted output
- · Number of AI-incident reports (a leading risk indicator)
Stop tracking these
- · Number of AI tool licences deployed
- · Tokens consumed / prompts run per employee
- · Self-reported time savings
- · Engagement with AI training modules
- · Number of AI champions or ambassadors
- · Pilot count without graduation rate
06
Common pitfalls
- ◆Running only the human layer. Change-management without AI literacy produces committed teams who cannot tell good AI work from bad. Both tracks or neither.
- ◆Running only the AI layer. Technical literacy without identity work produces accurate maps that nobody follows. The team can describe the right moves and refuse to make them.
- ◆Skipping levels. Trying to address Capability 4 (workflow redesign) while still stuck at Level 1 (adoption without value) produces sophisticated redesigns of work nobody is doing well yet. Levels are ordinal for a reason.
- ◆Delegating to a transformation office. The framework reproduces the Level-0 stuck-point when it becomes a programme. It is meant to be a working practice of the leader, not an artifact of a PMO.
- ◆Treating the scorecard as a one-off. The scorecard is a diagnostic instrument, not a maturity certification. Re-run it quarterly.
- ◆Mistaking enthusiasm for adoption. The CEO posting about AI on LinkedIn does not move the organisation past Level 0. Senior engagement in this framework means making decisions with AI in the room, on the timescales the rest of the organisation watches.
07
When to hand off
The framework is open and designed to be runnable without external support. It is consolidated from 200+ first-person interviews with CHROs and CEOs and from prior engagements where earlier versions of these capabilities and levels were applied. This is its first published edition; self-administered case data will accumulate as practitioners run it.
The strongest opening move is training the senior team on the framework before the first scorecard, so that diagnosis and language are shared from day one. Self-administration works; trained-administration compounds faster.
Even without that, there are two situations in which bringing in the framework's originator makes the work materially faster:
- ·Two or more functions placed at different levels. Reconciling differences across a senior team is harder than placing the organisation on the scale.
- ·The organisation has re-run the scorecard twice and the level has not moved. The likely cause is running one track without the other. A facilitator can shorten the diagnostic loop.
Robert Blaga's practitioner training is the formal hand-off point — a two-day cohort programme followed by a 90-day cadence, designed around this framework.