L2 — the domain layer

Where a Knowledge Verification Architecture should carry the work

L2 is domain knowledge — what a specialist in tax, telecoms, ERP, or compliance would carry in their head after years in that field. Here is how I actually carry it: as maps, not as contents. Twenty-five years in telecoms gives me the map of how a telco works — where the regulatory boundaries sit, where the operating norms bend, where the commercial lines run. It does not give me every regulation by heart.

For those (FIRB rules, transfer pricing, ERP system documentation, industry regulations) I have always gone back to the books. What is new is that I can now go instead to the right AI setup — one that pulls specialised rules and documents into its working context on demand. In our language, a Knowledge Verification Architecture.

The maps I carry

Telecoms and ICT

Twenty-five years across carrier, ICT services, and infrastructure. I know where the regulatory boundaries sit and where the operating norms bend.

M&A integration and divestment

The shape of integration decisions — degree, pace, earnout protection, divestment triggers. I can read a deal document and tell what is missing.

Capital portfolio management

Investment cases, allocation trade-offs, the financial-versus-delivery framing. Nine-figure portfolios. I know what reporting actually tells you.

Reading financials critically

Seeing what is missing or wrongly presented in a P&L or balance sheet. Fully-allocated costing. Activity-based costing. Valuation methodologies — EBITDA multiples, comparable transactions.

FIRB · transfer pricing · valuations

The quiet rules behind moving parts of companies. I know where to look and what questions to ask — the rules themselves are all documented.

ERP and enterprise systems

Large SAP and cross-system implementations. Common failure modes and what "custom vs configured" really costs over a decade.

AI-augmented work

Two years building a multi-capability operating environment for a single person. I know what the stack can and cannot yet do.

Full-company IT architecture

Running IT across a whole organisation is different from running one system. The meta-map of holding ERP, CAD, network, cyber, identity, HR systems, finance integration, engineering tools, and compliance simultaneously — knowing which decisions cascade and where the critical path sits.

Managed services and professional services industry

Years inside network managed services, security managed services, Microsoft partner channels, and cross-sell ICT services. The industry-specific patterns of how these businesses run — pricing, margin, SLA structure, revenue-versus-people tensions.

Where this leaves me: I carry the maps, but I have not wired the detail back to them. The rules and standards are all documented somewhere — regulations, vendor docs, industry bodies. The gap is not knowing what is there; it is that I am still flipping back to the books manually instead of having a retrieval setup that does it for me. Standing one up for my own domains is the next big codification move.

Between L1 and L2, technology now does most of the work. L1 lives in foundation models. L2 lives in retrieval architectures and documented rules. The human’s job in the AI age is above — externalising at L3a, and bringing the L3b judgement that cannot be commoditised.