The 4Ps AI Governance Framework

Compound business value with practical AI governance.

Primed, Principled, Practised, Protected. The practical AI governance framework that harnesses AI instead of caging it, by Segun Osu, founder of Teamsmiths. Distilled from 25+ team transformations and governance delivery at Haleon, GSK and Ogier Group, and grounded in the Oxford and Wharton business school programmes. It comes with a free 5-minute assessment to start.

No sign-up required to run the assessment. Your data stays in your browser.

The 4Ps AI Governance wheelA wheel showing four governance dimensions arranged clockwise from the top right: Primed (AI literacy, data readiness, use-case clarity); Principled (stated principles, decision rights, human-in-the-loop); Practised (use-case lifecycle, operating cadence, 6Ps team capability); Protected (EU AI Act and NIST alignment, risk inventory, model cards, audit). At the centre sits the wordmark "4Ps AI Governance, by Segun Osu".4PsAI GOVERNANCEby Segun OsuPRIMEDAI literacyData readinessUse-case clarityPRINCIPLEDStated principlesDecision rightsHuman-in-the-loopPRACTISEDUse-case lifecycleOperating cadence6Ps team capabilityPROTECTEDEU AI Act / NISTRisk inventoryModel cards · Audit
Engagement historyHaleonGSKGartnerOgier GroupPhilipsReutersBain (coordinated)
The problem

Most AI governance is written for the regulator. Not the operator. Not for business value.

Three patterns show up in nearly every stalled programme. Each one is a cage where there should be a harness.

Governance theatre

A binder of principles no team can recite, with no decision attached. Looks good on Friday, fails on Monday.

AI without literacy

Pilots green-lit by leaders who cannot read a model card. Platforms bought before use cases. Decisions taken on confidence, not evidence.

Controls without cadence

Risk inventory built for one audit, never refreshed. Incident playbook opened only on the day it is needed. Audit trails technically present, not retrievable.

The pattern no one governs

Compute is treated as a utility. It is a supply-constrained input.

Most teams buy AI like software and assume it will always be there. It will not. The capability runs on hardware the vendor does not fully control, so when compute runs short your usage is rationed. That is an outage you did not cause and cannot fix. The 4Ps governs this under Protected, as Supply resilience: tier what must not fail, hold a fallback you control, and rehearse the day the tokens stop.

The framework

Four dimensions. One operating system.

Read them in order. Each P assumes the one before it. Each P is testable. Each P maps to specific work a leader can commission this quarter. Think of it as a harness for AI: it turns raw model power into work you can steer, with a line for when it bolts.

Anatomy of the harnessThe four Ps shown as four straps of a harness running from a central node of raw AI power. Primed is the fitting, Principled is the reins, Practised is the pull, Protected is the safety line. Each strap carries the question its dimension answers.Anatomy of the harnessFour straps turn raw AI power into governed, daily work.AIRAW POWERPRIMEDThe fittingIs the org built to carry AI?PRINCIPLEDThe reinsWho steers, and who can stop it?PROTECTEDThe safety lineWhen it bolts, what catches you?PRACTISEDThe pullIs the power doing work, daily?Govern AI like a harness, not a cage.Fit it. Take the reins. Put it to work. Rig the safety line.

PRIMED

AI literacy and readiness across leadership, workforce, data, and platforms.

  • Leadership literacy
  • Workforce literacy
  • Data readiness
  • Infrastructure readiness
  • Use-case clarity

PRINCIPLED

Ethics, accountability, and governance written down and lived.

  • Stated principles
  • Decision rights (RACI)
  • Ethics review process
  • Transparency standards
  • Human-in-the-loop rules

PRACTISED

AI operationalised through disciplined practices day to day.

  • Use-case lifecycle
  • Model lifecycle
  • Operating cadence
  • Team capability (6Ps)
  • Vendor discipline

PROTECTED

Risk, compliance, audit, and security. The dimension that holds when challenged.

  • Risk inventory
  • EU AI Act / NIST / ISO 42001
  • Model cards
  • Audit trail
  • Incident response
Built for the loop-based company

For companies running on closed AI loops.

Self-improving companies build sense-decide-act-learn loops around a shared brain. The 4Ps governs that case better than any framework written for a static system: it is practical, it carries the value-creation frame, and it adds three sharpenings that loops specifically need.

Anatomy of a self-improving loopA closed loop with four stages arranged in a square: Sense at top-left, Decide at top-right, Act at bottom-right, Learn at bottom-left. Arrows run clockwise from Sense to Decide to Act to Learn and back to Sense. A dashed central panel labelled Company brain sits in the middle. Each stage is labelled with the P that governs it: Primed fits, Principled steers, Practised pulls, Protected catches.Anatomy of a self-improving loopSense. Decide. Act. Learn. Compounded by governance.SENSECapture signal from the work.Data, events, transcripts,customer signal, telemetry.DECIDEApply principles and policy.Decision rights, ethics,human-in-the-loop rules.ACTDo the work at the authority class.Read-only, recommend,act-with-reversal, autonomous.LEARNRecord, score, improve.Audit trail, drift detection,feedback integrity.Company brainsearchable. shared. recorded.PRIMED FITSPRINCIPLED STEERSPRACTISED PULLSPROTECTED CATCHESClosed loops compound. Governed loops compound safely.Each stage carries its P. The brain is the shared memory all four feed.

Loop as an artefact

Every loop names six things on day one: an owner, an authority class (read-only, recommend, act-with-reversal, autonomous), a feedback channel, a kill switch, a review cadence, and a retirement criterion. Most loops should not be autonomous, and the framework makes that the default.

Feedback integrity

Loops that train on their own output decay quietly. The control is mundane: an input ledger that flags AI-generated content, a clean evaluation set held in escrow, and continuous drift monitoring. Model collapse becomes a named risk class, not a surprise.

Surveillance guardrail

A company brain that records every action is one inch from worker surveillance. The framework draws the line: legibility serves the work, not the worker. Transparency standards, labour-law in the regulatory map, and the harness rule kept hard.

How it works

Three steps. Thirty days. A framework that holds.

01

Assess

Run the 4Ps Self-Assessment. Twenty questions, four dimensions, a score out of one hundred and a colour-graded read.

02

Diagnose

Read the radar and the heat map. See exactly where the gap sits and which artefacts you are missing.

03

Plan

Use the 30-day rollout: one P per week, with the artefacts pre-loaded and a monthly cadence that locks in compounding.

Where it comes from

Earned in delivery. Sharpened by study.

The 4Ps distils enterprise delivery at Haleon and GSK, a corporate and data governance framework at Ogier Group across multi-jurisdiction legal and financial services, and formal study through Oxford AI Governance and Wharton AI Strategy and Governance. The numbers below are the delivery track record behind it.

OGIER GROUP · Legal and financial services

30% performance uplift in 45 days, governance frameworks delivered ahead of schedule

Stood up the Programme Office, Programme Board and a Scaled Agile framework, and embedded agile ways of working across product, engineering and data. Delivered enterprise risk management and corporate governance frameworks ahead of schedule.

Programme Manager and Agile Coach
GSK PLC · Pharma, global, pre-Haleon de-merger

40% performance lift across product, engineering, and data teams

Same operating template, applied earlier. Co-ordinated EY and BCG consultancies alongside internal teams. 10+ large-scale planning events. Delivered under budget and ahead of schedule in a regulated environment.

Agile Coach and Programme Manager
HALEON PLC · Consumer health, regulated

45% performance lift across product, engineering, and data teams

Six months. 15 per cent cost reduction. ~99 per cent delivery predictability. Coordinated Bain on an AI-powered smart retail shelf rollout in parallel. Received a Global Recognition Award the same year.

Agile Coach and Programme Manager
45%
Performance lift at Haleon PLC, 15% cost reduction in six months
40%
Performance lift at GSK PLC across product, engineering, data
~99%
Delivery predictability across coached teams
$3M-$10M
EMEA digital media portfolio growth at Gartner, 5 years
£25M
Accounts receivable situation cleared inside a year at Philips Italy
53%
Supplier cost saving on a previous engagement
5
Production agentic AI platforms shipped
20+
Years building teams that compound
Credentials

Trained where the discipline is made.

OXFORD

Oxford AI Governance

Saïd Business School. 100% grade. December 2025.

WHARTON

Wharton AI Strategy and Governance

100% grade. March 2025. Part of the AI for Business Specialization.

WHARTON

Wharton AI for Business (full)

AI Fundamentals, AI Applications in Marketing and Finance, AI Applications in People Management. Four certificates.

MBA

MBA, University of Sheffield

The Management School. Operating model and strategy depth.

B.ENG

B.Eng (Hons), University of Westminster

The engineering instinct that underwrites the operator voice.

OXFORD

Oxford Blockchain Strategy

Saïd Business School. The lens that anchors the governance work in distributed-trust thinking.

AGILE

Agile leadership

Leading SAFe (Scaled Agile), PSM (Scrum.Org), ICP-ACC (ICAgile), Colorado Agile Leadership.

LANG

Multilingual reach

English (native), Italian (fluent, Università per Stranieri), Yoruba (conversational).

PHIL

Practical Philosophy

Years of weekly study at the School of Practical Philosophy. The Principled instinct sits here.

Companion book

Team Performance Uplift

The 6Ps and AI playbook for teams that compound. The 6Ps is the operating model. The 4Ps is the governance frame. Together, a single rollout.

Read the book
By Segun Osu
13,500 words
Engage

Three ways in.

Free assessment

Run the 4Ps self-assessment in under thirty minutes. Get a radar, a colour-graded read, and a written diagnostic of the largest gap.

  • 20-question self-assessment
  • Per-P scores and radar
  • Largest-gap diagnostic
  • Email of the result

Pilot engagement

A 30-day rollout for one team. Baseline. Principles and RACI. Model lifecycle wired in. Go-live and monthly cadence locked.

  • Day 1 baseline workshop
  • Principles + RACI design
  • Model lifecycle + risk inventory
  • Live model card + audit trail
  • Monthly cadence handover

Enterprise programme

A multi-team rollout that holds. Three teams per quarter. Governance operating model. C-suite cadence. Build/buy guidance on platforms.

  • Programme office and governance
  • 3 teams per quarter rollout
  • C-suite cadence and reporting
  • Build vs buy advisory
  • Coaching for senior leaders
For AI vendors

Selling AI to enterprises?

The Enterprise-Ready Governance Pack: everything an enterprise security review asks of an AI vendor, built with you in 14 days. Fixed price. Fixed scope.

The Governance Pack ›
Founder

Segun Osu

Founder of Teamsmiths. Enterprise AI Transformation Architect and Advisor. Twenty plus years in engineering, consulting, product, and the boardroom across three continents. Author of two frameworks: the 6Ps & AI (Team Performance Uplift) and the 4Ps AI Governance.

Trained at Westminster, Sheffield, Oxford, Wharton, and Colorado. Fluent in English and Italian, conversational in Yoruba. Speaker at Surrey Chamber of Commerce, University of Sheffield with Forrester analysts, and the Agile Sherpa webinar series.