Contact us

01 · WM.AI

Intelligence that sits inside the work — not beside it.

We put AI where the process already breaks — with owners, logs and kill switches. Strategy, private knowledge systems, agents and AI-assisted integration for government, infrastructure, enterprise and growth organisations that cannot afford pilot theatre.

Book an AI readiness conversation ↗ See AI Receptionist Product proof

AI Receptionist · live Council of Self · SaaS Signal · scheduled · bounded

The anti-brief

If you want a ChatGPT wrapper, hang up.

We build operational intelligence inside real systems — with boundaries, logs and owners — not strategy theatre and not a model wrapper with a logo on it.

What we refuse

  • AI strategy that ends as a deck and a steering group with no production path
  • Public live-AI gimmicks on the marketing site — intelligence is scheduled, scoped and owned
  • Model cosplay: “we use [vendor]” is not differentiation. Models are commodity; systems are not
  • Pilots that never touch money, service delivery or compliance

Running intelligence

Scheduled. Cached. Human-owned.

Outputs you can open — not a live model on the marketing page.

WM Signal · today AI agents are becoming practical, secure and enterprise-ready. 22 Aug 2026 · Curated seed Clara · latest Automation Dashboards Hide the Real Work Transfers 21 Aug 2026 Labs Agent ops · Signal pipeline Bench, not production cosplay

scheduled · cached · human-owned

Practice modes

Where we put intelligence to work

Eight modes — same craft. Select a practice; the stage shows the boundary as well as the promise.

01 · Practice

AI readiness

Honest assessment of data access, process ownership, risk classes and one expensive decision chain — before you buy another pilot.

  • Where the process already breaks under load
  • What data is reachable without theatre
  • Who owns the outcome on Friday afternoon
Discuss readiness ↗

Decision stack

Assess · govern · build · assure

A vertical control path — not a service menu of equal cards. Scroll; the active layer holds focus.

01

Assess

Readiness without theatre: data that is reachable, process ownership, risk classes, and the single decision chain where error is expensive.

  • Shadow AI and fragmented systems of record
  • What must stay deterministic (money, identity, safety)
  • Metric the business already believes
02

Govern

Governance is decision rights with a kill switch — policies mapped to system behaviour, not a responsible-AI PDF.

  • Named owner of outcome and P&L per agent class
  • Escalation and override before the first incident
  • Audit trails operations can run without a vendor workshop
03

Build

Agents, automation and private knowledge on top of systems that already hold the work — registers, PWAs, APIs, clinic and commercial platforms.

  • Thin slice that touches a real system of record
  • Scoped tools · logging · human handover
  • Products as proof: Receptionist, Council of Self, Inspect substrate
04

Assure

Evaluate under load — escalation rate, error cost, time-to-handover, override rate — then scale or stop. Kill losers fast.

  • Task success under operational conditions
  • Explicit keep / kill decision
  • No multi-year AI roadmap theatre

Decision chains

Where error is expensive

Six chains we design around. Select one — static control map. No model call on click.

AI-assisted integration

When fixed rules break, intelligence assists — carefully.

Legacy landscapes produce unstructured data, exceptions and semantic mismatches that brittle point-to-point integrations cannot resolve. AI can assist with understanding, classification, mapping and orchestration — while deterministic controls remain for critical transactions, security, validation and auditability.

Sources → policy → scoped tools + model → validators → human escalate → audit

Control matrix

  • DET Money, identity, safety — never silent mutation
  • AI Classification, mapping, exception triage
  • HUMAN Ownership of outcomes and overrides
  • LOG Every agent action leaves a trail

No public live AI gimmicks on this site. Scheduled, bounded processes for content intelligence. Show the output — not slogans about “using AI.”

Proof stack

Proof is a product, not a pilot certificate

Same craft as client systems — visible, demoable, constrained.

Available

AI Receptionist

Calls → intents → bookings against clinic systems → human handover. Not a website chat widget.

Boundary: staff own clinical outcomes View product → SaaS · live

Council of Self

Private multi-persona debate people can use today — orchestration without enterprise theatre.

Boundary: private sessions · Google sign-in View product → Available

Inspect

Registers, quarantine, audit trails — the spine intelligence must attach to, not replace.

Boundary: compliance decisions stay human View product → Available

TenderChain

Structured commercial process — AI belongs in classification and review assist, not silent bid decisions.

Boundary: award decisions stay owned View product →

Substrate (systems AI needs): AnonSurvey · CouncilBins · delivery under load

How we engage

Three moves. No multi-year AI roadmap theatre.

01 · Readiness

Honest assessment

Data access, process ownership, risk classes, one expensive decision chain.

02 · Thin slice

Production-shaped pilot

Touches a real system; human escalation; metric the business already believes.

03 · Scale or stop

Explicit kill / keep

Only what moves the metric scales. Losers die without ceremony.

Next step

Bring one expensive decision chain.

We’ll design around that — readiness, thin slice, or a product walkthrough.

Start a readiness conversation ↗ AI Receptionist demo Read Clara