Strategy · Wardley
Know where to build, buy or wait
We map your AI value chain, show which parts are already commodity, and set a roadmap your board can fund.
You keepWardley maps, build-or-buy research, roadmap
Counsel for AI in regulated institutions
Sit, listen, then decide.
The Majlis of AI helps banks, telcos and governments in the GCC make AI decisions they can defend: to the board, to the regulator and to their customers. It is an advisory practice, an academy, a set of books and an open conversation.
1AI advisory
We help you make the AI decision. Then your teams own it.
Architecture and operating-model advice for banks, telcos and governments. We work beside your architects and your risk teams, and we leave the method with them.
Operating model
Give every AI decision one owner
Workflow-first design from your own SOPs, with Team Topologies, decision rights and human-in-the-loop rules in DMN.
You keepTeam design, decision register, activation roadmap
Architecture
Draw the target state in your industry's shape
Target architecture on BIAN, TM Forum ODA or TOGAF, modelled in ArchiMate, with each decision recorded.
You keepArchiMate views, ADRs, reference-model maps
Governance-as-code
Prove every control to your regulator
We turn CBUAE, TDRA, PDPL or EU AI Act obligations into owned, testable rules in Git, with a release gate for every model and agent.
You keepGovernance pack, traceability matrix, autonomy tiers
How we work
- ListenSit with the people who own the decision. Write down the real question.
- MapMap the obligations, the current state and the value chain.
- DecideCompare the options. Record the decision, its owner and its evidence.one decision, one owner, one record
- BuildPair with your teams on the first governed release.
- Hand overYour teams run the method. The Academy closes the gaps.
Ways to engage
| Shape | Use it when | You get |
|---|---|---|
| Decision sprint | One decision blocks a programme | A recommendation and a decision record, defended to your board |
| Programme design | You start or reset an AI programme | Operating model, target architecture and roadmap |
| Design authority | You run AI in production | A seat on your architecture board and a review of each release |
Sectors and frameworks
- BankingBIAN · CBUAE
- TelcoTM Forum ODA · eTOM · SID · TDRA
- GovernmentUAE AI Charter · sovereign cloud
AI transformation · Open Agile Architecture™
From AI proofs of concept to AI at enterprise scale
Most AI work stops at the proof of concept. We use Open Agile Architecture, a standard of The Open Group, to shape the operating model, data platform and governance that take AI to production across the enterprise.
Intentional architecture sets the guardrails. Emergent design keeps the teams fast.
AI strategy and operating modelDual transformation · value streams · jobs to be done
Intentional AI foundations
- Central AI platform and MLOps
- Enterprise data fabric
- Pace-layered guardrails
Federated governance and continuous feedback
Emergent product execution
- Cross-functional AI squads
- Rapid AI MVPs and prototypes
- Context-driven refactoring
A · Strategy and operating model
Turn AI projects into long-lived AI products
Map AI use cases to value streams and customer journeys. Shape cross-functional AI product teams with the Inverse Conway Maneuver.
You keepValue-stream map, team design, product operating model
B · AI architectural runway
Build the platform once, let squads experiment
A central AI platform: vector stores, model registries, CI/CD for ML and observability. ADRs set the line between central standards and squad freedom.
You keepRunway blueprint, ADRs, team API contracts
C · Pace-layered governance
Govern each AI system by its criticality
- Systems of recordStrict
- Systems of differentiationBalanced
- Systems of innovationLightweight
You keepPace-layer risk matrix
D · Continuous refactoring
Keep AI right after go-live
Feed production telemetry back into the architecture backlog. Make red-teaming, toxicity filters and hallucination benchmarks sprint-level quality criteria.
You keepFeedback loops, automated guardrails
Packaged engagements
- 2–3 wkAI-ready agile architecture assessmentWe review architecture maturity, team topologies and how your AI assets are decoupled. You get a gap analysis against the O-AA axioms.
- 4–6 wkAI platform runway blueprintWe define the minimum viable architecture for AI: data pipelines, model orchestration and gateway patterns. You get ADR, team API and pace-layer templates.
- QuarterlyProduct-centric AI transformation retainerWe coach your architects and product owners on sprint-based architecture backlogs, and align value streams for multi-agent systems and model migrations.
What each leader gets
| Leader | Problem | What changes |
|---|---|---|
| CIO · CTO | AI work splits into silos or stalls at a rigid architecture board | A federated architecture practice: central platforms let squads build compliant AI without waiting |
| CAIO · CDO | Proofs of concept fail on the way to production | Data and MLOps sit inside the value streams, on a reusable runway |
| Product and business | AI takes months and misses the business metric | Journey mapping, jobs to be done and outcome-driven design |
| CISO | Shadow AI and unvetted APIs leak IP and data | Security and compliance from the start, through automated guardrails |
Works with your frameworks
- TOGAFBaseline and target
- SAFe · ScrumAgile execution
- O-AAAdaptation between them
2The academy
We build AI capability inside your institution.
Start with the flagship course, or bring one of our formats to your teams. Everything we teach uses your platforms, your controls and your regulator's expectations.
Flagship course · Edition 1 · Core
AI Regulated by Design
Mastering governance-as-code. Build one traceable governance pack for an AI solution. Keep it in Git. Defend it to a review board.
Coming soonInstructor-led
7+1Modules and a capstone
8 wkAbout 7 hours a week, live cohort
12–20Learners per cohort
1Governance pack each
| Wk | Stage | Module | What you produce |
|---|---|---|---|
| W1 | Set up | M0Harness anatomy | Pinned repository, project plan |
| W2 | Govern | M1Govern the intent | Principles, stakeholders, risk, business case |
| W3 | Understand | M2Needs and data | Requirements with IDs, data model, DPIA draft |
| W4 | Decide | M3Research and strategy | Build-or-buy research, Wardley map, roadmap |
| W5 | Design | M4Design and decide | ADRs, diagrams as code, design review |
| W6 | Design · AI | M5Govern AI and agents | AI assessment, agent pack, autonomy tiers |
| W7 | Assure | M6Assure and extend | Traceability matrix, own recipe, docs site |
| W8 | All stages | CPCapstone | Complete pack, defended to a review board |
Set the rules, understand the need, decide, design, assure. Gaps and drift feed the next cycle. Traceability IDs link every stage.
Lab case
The Lyra project
A synthetic agentic service assistant for a fictional regulated organisation. It answers with retrieval and does one customer action.
Synthetic data only. Never client data.
AIPersonal dataAutonomyNo sector
Capstone · Week 8
Defend three decisions to a mock board
Govern your own AI solution from an empty repository, with your own recipe. Then face a 20-minute board: CISO, DPO and board chair.
Full spine, a recipe and two templates, a review log, ten checked citations, a docs site and a ten-slide deck.
Who it is for
- Design
- Solution and enterprise architects, AI engineers
- Review
- Tech leads, risk and compliance, architecture boards
- Bring
- Git, a terminal, architecture basics and an AI coding assistant on a paid plan
The toolkit
GovKit, the enterprise governance harness
The same harness runs in Claude Code, Codex, Gemini, OpenCode, Copilot or a CLI. Nine overlays add sector and jurisdiction rules, from the UAE and the EU to UK Finance.
- Assistants call toolsEvery write, edit and command passes through the harness.
- The harness governs each callHooks give context, check access, block what breaks a rule and observe the rest.
- Traceable artefacts come out128Doc-types on one traceability graph, from driver to test.
How to join
| Instructor-ledComing soon | Self-pacedComing soon | |
|---|---|---|
| Format | On-site or live online | Online, on demand |
| Time | 8 weeks, about 7 h a week | About 40 h, own pace |
| Support | Instructor and peers | CI checks, forum |
| Certificate | Of achievement | Of completion |
How we assess
Pass mark for the live course: 60% overall, and at least 2 of 4 in each of the six rubric dimensions.
Planned editions
One core. Four sectors.
The core course comes first. Industry editions build on it, so what you learn in the core carries over.
Coming soonGovernmentUAE AI CharterAutonomy tiersSovereign cloud
Coming soonFinanceBIAN domainsCBUAE rulesModel risk
Coming soonTelcoTM Forum ODAeTOM · SID · TAMTDRA rules
Coming soonHealthPatient dataClinical AI safetyHealth regulators
Edition 1 · The coreGovern · Understand · Decide · Design · AssureComing soon
Each industry edition adds
Sector modelsMap designs to the sector's reference architecture.
Regulator overlayAdd the sector's rules and controls to each artefact.
Sector case labBuild your pack on a realistic sector case.
Formats for your institution
Bring the method in-house, from a board briefing to an accelerator track.
| Ref | Format | For | You leave with |
|---|---|---|---|
| AC-01 | Executive briefing | Board, ExCo, CTO and CIO | A shared view of AI value and risk, and a list of the decisions only you can make. |
| AC-02 | Technical bootcamp | Engineers and architects | Hands-on practice with DevOps, MLOps and LLMOps standards on a regulated platform. |
| AC-03 | Capstone programme | Cross-functional teams | One real use case, from problem statement to a reviewed ArchiMate architecture. |
| AC-04 | Hackathon | Delivery teams | Working prototypes, judged against your own controls. |
| AC-05 | Accelerator track | Platform and product teams | A capability your team runs without us after the track ends. |
We teach with
BIANTM Forum ODAeTOM · SID · TAMTOGAFWell-ArchitectedArchiMateWardley MappingSystems thinkingASD-STE100
capability stays when we leave
3Books
It started as a novel about the room where AI is decided.
Published on Amazon
The Majlis of AI is a tech novel of the new geopolitical council, where innovation, enterprise and politics meet in the room where AI is decided.
The story gave the practice its name and its method. Everyone sits as equals, everyone listens first, and the decision is written down.
From one book to a practice
- NovelThe story and the method
- DialoguesArticles and the newsletter
- AcademyBriefings to bootcamps
- AdvisoryArchitecture and operating model
- LabsAI services
Collaborations · with Márcio Parente
The Regulated AI Operating Model
A series of five books, co-written with Márcio Parente, for the people who must run AI inside a regulated institution.
4Articles
Name the framework, draw its real shape.
Short pieces for architects and executives in banking, telco and government. Each one starts from a framework you already know.
5Newsletter
One issue a month. One decision in each.
Majlis Dialogues is a short letter for people who must sign off on AI in a regulated institution. Read it in ten minutes. Bring it to your next steering committee.
In each issue
- One framework, drawnA BIAN, ODA or TOGAF view with the one change that matters marked.
- One regulatory signalWhat changed for AI in GCC financial services, telco or government, in plain words.
- One question for the boardA question to put on the agenda, with the answer we would expect to hear.
6Contact
Bring us the question your institution must decide.
Write to us about advisory work, the Academy, or a briefing for your board.