Insurance · SAMA · KSA
Najm Insurance
6,000+
Daily cases · 40+ cities
Constrained Identities
The model is the easy part. The plumbing is what kills you.
70% of AI pilots die before production — at the governance layer, not the model layer. I build the layer that ships: constrained identities, attributable actions, human gates and kill switches for agents operating under real regulators.
Book a 20-minute call
One conversation. No deck theatre.
Built for
Banking
Insurance
Pharma
Energy
Government
The receipts are the moat
Four regulated systems, in production.
Most governance consultants have zero real deployments. These four did the persuading before this page did. Attribution is fixed — figures never move between cases.
Government · India
Govt of India / NMML
€10M+
180+ endpoints · 99.9% SLA
Attributable Actions
Public transit · Belgium
De Lijn
5,000+
FTE in scope · board-approved
Human-in-the-Loop Gates
Hospitality intel · USA
US Restaurant Intelligence
200 → 3
FTE to agents · ~90% cost cut
Kill Threshold Monitoring
The method · the Trust Engine
Four non-negotiables decide whether AI ships.
Not principles on a slide — controls in the runtime. Each pillar is anchored to one of the four deployments above.
Intelligence is rented. Piping is owned.
01
Constrained Identities
Agents run least-privilege service accounts. Destructive actions are blocked at the keyring, not by instruction — zero default write permissions.
PreventsBlast radius when an agent misbehaves.
02
Attributable Actions
Every decision, query and citation is logged on a read-only ledger. Any action is traceable — and reversible — in under 30 seconds.
Prevents“We don’t know why it did that.”
03
Human-in-the-Loop Gates
High-stakes actions pause and wait for manual authorization. The agent cannot execute financial or legal tasks on its own.
PreventsIrreversible autonomous action with no human in the decision.
04
Kill Threshold Monitoring
Continuous telemetry watches speed, spend and error volume. On breach the agent auto-suspends — before damage, not after.
PreventsRunaway loops and cost spirals running unchecked.
Instructions are a wish. The keyring is a guardrail.
How to engage
Three tiers. One conversation.
Read-only audit, full implementation, or end-to-end orchestration. Scoped to the size of the system, not the hours.
Tier 1
Governance Assessment
Read-only audit, EU AI Act classification and a risk register. The artifact a board can act on.
Tier 2
Trust Engine Implementation
The four pillars built into production systems — constrained identities, attributable actions, gates and kill thresholds, live.
Tier 3
Orchestration
Multi-system, multi-client governance across the full lifecycle — the operating layer for an enterprise agent estate.
Read it, then install it
The method in long form — and in your runtime.
The book
The AI Plumber
The four pillars written long-form, each chapter anchored to one verified deployment. The 70% pilot-death problem, and the four non-negotiables that decide whether AI ships.
Read the framework →Open source
Kleiber
The open-source layer that enforces the four pillars in your agent runtime. aiplumber.dev is the consultancy and the method; Kleiber is the installable proof.
Install the layer →
Koen Van Lysebetten
AI Architect & Governance Advisor · DevGap
About →
Regulatory currency: EU AI Act high-risk (Annex III) obligations apply from 2 August 2026 (operative date; subject to pending Digital Omnibus revisions — verify current status before relying on it).