Sagentivum Book a conversation
Product · Decision architecture diagnostic

WorkLattice. The structural read on whether your AI programme can land.

A graph-based decision architecture diagnostic for AI implementation risk. It reads your governing documents as one connected system and tells you whether your organisation can absorb the decisions your AI generates. Built on twenty years in safety-critical engineering, senior consulting, and institutional execution.

Book a diagnostic call Where it applies

AI didn't create the gap between what your organisation knows and what it can do. It removed the slack that used to hide it.

Recommendations now move faster than authority can ratify them. Information moves faster than accountability can follow. Models produce answers faster than the organisation can test whether those answers are usable, legitimate, or safe to act on. That gap is what kills AI initiatives after readiness has passed, and fixing the model won't close it, because the model was never the problem.

Two operating models

Every organisation runs two operating models. WorkLattice tests the gap between them.

Model A

The one in the governing documents.

Your policies, standards, delegations, and procedures. Assumed tidy, coherent, and designed - until you read them as a single connected system.

Model B

The one that actually runs.

The decisions, escalations, and overrides that emerge from people, history, and political weight.

WorkLattice reads your governing documents as a directed graph - authority, obligation, escalation, record - and produces two readings at once. First, where the documents contradict themselves: authority assigned twice or not at all, obligations one document delegates and another never enacts, records that dead-end before they reach a decision. Second, where the documented organisation has parted company from the one that actually runs - decision rights that exist on paper but not in practice, shared terms that mean different things to different functions, rules that are formally specified but routinely overridden when political weight says otherwise. The first reading comes from the documents alone. The second is confirmed against practice.

Most AI initiatives don't fail because the design was wrong. They fail because the conditions that had to hold for the design to work weren't there in the first place, and nobody had a way to see that before the build.

Sagentivum is an independent Sydney-based advisory practice that diagnoses AI implementation risk through this graph-based decision architecture analysis. The same engine runs behind buyer-specific reads, including a pre-deal structural read for private equity. The practice is led by Paul Dorrell, founder and Managing Director.

What the test finds

Three of the conditions it tests for.

01

Who actually holds the decision

Are authority boundaries clear, or does the real decision depend on who's in the room? Where rights are ambiguous, AI doesn't resolve the ambiguity: it encodes it, then runs it at speed. The test shows where rights are clean enough to build on and where they have to be settled first.

02

Whether everyone means the same thing

Do the words that matter, qualified, approved, complete, escalated, mean the same thing across the functions that depend on them? Divergent definitions coexist quietly for years. When an AI imposes one, the disagreement surfaces at the execution layer and looks like the system is broken.

03

Whether the rules actually hold

When a rule says X must happen before Y, does it, or does political weight routinely override it? Enforceability is the hardest condition to see from documentation and the most reliably predictive of whether the programme lands.

Where it applies

Three moments where the structural read changes the outcome.

  1. Application 01

    Mid-flight AI programmes that aren't landing.

    Pilot succeeded, scale is stalling, and the explanation is drifting toward "adoption problem." WorkLattice diagnoses the structural cause, identifies what is recoverable, and tells you which parts of the design need to change before the next sprint. The output is a decision a sponsor can act on.

  2. Application 02

    Before a scaling decision is signed off.

    Where a board paper is being prepared and the sponsor privately suspects the organisation can't absorb the rollout. The diagnostic identifies the right starting boundary, the conditions that must hold for the boundary to expand, and the specific structural risks that will manifest if scale proceeds without remediation.

  3. Application 03

    Operating model and structural review.

    Where a restructure or transformation isn't landing. The diagnostic distinguishes what's working from what is structurally preventing delivery, before the next redesign is built on top of the same constraints.

The deliverable

What you receive.

  • A decision architecture graph built from your governing documents - the documented structure, and where it parts company from the one that runs.
  • A structural findings report: each finding graded for confidence and chained to the verbatim clause it rests on, with its implication for the AI programme.
  • A sponsor-grade presentation deck.
  • Optional: validation against practice, and a remediation roadmap covering what must change before scale, what runs in parallel, and what the AI implementation should be reshaped around.

See it run: a worked example on a stacked-framework governance suite →

Two to four weeks, fixed scope, fixed fee. Typically one to three percent of the AI programme cost it is testing.

Considering a structural read?

A first conversation establishes whether WorkLattice fits your situation, the structural questions worth answering, and how the work would be scoped. Thirty minutes.

Book a diagnostic call About Sagentivum