Research & advisory · for the era of AI transformation
Everyone measures the machine. We measure whether it lands.
What we build
Most enterprise GenAI pilots deliver no measurable P&L impact.¹ The few that succeed differ in one thing: the organisation around the AI, not the model. Divergent Kind is the research and advisory practice that builds that organisation – we find where the value leaks, fix the layer the AI runs on, and make the change hold.
The call: 30 minutes · your context, our read · no deck.
A research & advisory practice · discipline: coherence architecture – the layer between AI systems and the people they serve
The coherence layer between AI and work
Everyone measures the machine. We measure whether it lands.
Most enterprise GenAI pilots deliver no measurable P&L impact.¹ The few that succeed differ in one thing: the organisation around the AI, not the model. Divergent Kind is the research and advisory practice that builds that organisation – we find where the value leaks, fix the layer the AI runs on, and make the change hold.
A research & advisory practice · discipline: coherence architecture – the layer between AI systems and the people they serve
The coherence layer between AI and work
Everyone measures the machine. We measure whether it lands.
Most enterprise GenAI pilots deliver no measurable P&L impact.¹ The few that succeed differ in one thing: the organisation around the AI, not the model. Divergent Kind is the research and advisory practice that builds that organisation – we find where the value leaks, fix the layer the AI runs on, and make the change hold.
The instrument · four dimensions
- 01Agency
- 02Reciprocity
- 03Alignment
- 04Signal Integrity
A research & advisory practice · discipline: coherence architecture – the layer between AI systems and the people they serve
Who we are
What we do, in one sentence.
We read whether your organisation – and the AI going into it – holds together for the people in it – then architect the conditions so it does.
That read is the centre of the practice. Not model benchmarks, not adoption dashboards: whether the system lands for the people who live and work inside it. Everything we measure, publish and fix starts there.
For builders, businesses and institutions – current engagements with scale-ups, PE-backed portfolio companies and boards.
A living demonstration · the site adapts to you
Tell us who you are. Watch the site meet you.
Everything we claim about systems that adapt to people – this page does it, live. The mechanism has a name: Dynamic Transposition™ (patent pending) – one truth, transposed to the person receiving it. Pick who you are, add your own words if you like, and watch it happen. We never guess. The truth never bends.
WHO ARE YOU?
WHAT ARE YOU TRYING TO MAKE WORK? · OPTIONAL
The transposition runs here on the page. Nothing sent, nothing saved, no profile.
From the demonstration to the work
This page adapts to its reader. Inside organisations, that signal usually stays buried: where value leaks, where trust breaks, where the system stops landing for its people. Reading that layer is the work.
Making the invisible visible. See what we find ↓
You might recognise this
The tools work. The transformation doesn’t.
“We’ve poured money into AI and I still can’t tell the board what we got for it.”
Chief executive
“It works in the demo. On the floor, people quietly went back to the old way.”
Chief operating officer
“Every initiative stalls in the same place, and no one will say why.”
Founder
“The model passes every test, and the people who use it still don’t trust it.”
Chief product officer
“Management says we’re on track. I have no independent way to know if that’s true.”
Board director
“The AI was meant to take the load off the team. It’s burning them out instead.”
Chief people officer
Different words for one problem: the business isn’t organised to use the AI it’s paying for. That problem has a name – and, for the first time, a measure.
The name for it
Coherence, defined.
A system is coherent when it still makes sense to the people inside it – decisions land, effort returns value, actions match intent, and the signal can be trusted. We read it on four dimensions:
Can people choose, and act – or does the system quietly decide for them?
Does the system return value in proportion to what it asks of people?
Does what the organisation does match what it says – and intends?
Can people trust what the system tells them – and act on it safely?
What sets us apart
Emerging measurement. Most organisations have none. We have one instrument – four named dimensions.
Latency, accuracy, throughput, evaluations, deployment volume – the AI industry can measure all of it. None of it tells you whether the output actually works for the person who has to act on it. That is the measurement that decides whether AI pays off, and almost nobody takes it. We built a named receiver-side instrument that does – reading whether AI-enabled work holds for the people who must act on it: agency, reciprocity, alignment, signal integrity.
The instrument is ours: authored, tested in deployment, and held – the proprietary core beneath an open method. You see the read, not the engine.
Current focus
AI-era transformation – before, during and after the change.
The principle is permanent; the focus is current. Right now, coherence matters most where AI is reshaping how organisations run – so that is where the practice is pointed. One discipline, applied at three moments:
Before you commit: the free Snapshot, then the Coherence Read – is the organisation coherent enough to hold what you are about to build? Small first, evidence early.
While it ships: the Coherence Audit, Compass design and operational install – friction found, costed and fixed in the structure, while the work keeps running.
Once it lands: Guided Evolution – coherence measured continuously, so the change holds after the programme closes, and after we leave.
One sequence underneath: Read → Audit → Compass → Install → Evolution · when the AI era gives way to the next shift, the focus moves – the principle does not.
The instrument
We read what KPIs miss.
There is no continuous, receiver-side, coherence-specific read of the human–AI boundary at the organisational layer. The Lens is how the instrument reads it.
Can people choose, and act?
Is value flowing both ways?
Does behaviour match intent?
Does it hold up under variance?
Participant Coherence
Does the signal land in the human's hands?
The evidence · why most AI investment stalls
Where AI actually pays off: the organisation around it.
Evidence under confidentiality – what travels between engagements is the pattern, not the client.
of leaders call their company mature in AI deployment²
of companies have yet to show tangible value from AI³
of employers’ businesses expected to be transformed by AI by 2030⁴
McKinsey 2025 · BCG 2024 · WEF 2025
The shift
Two conditions arrived.
The third was never built.
01 · Intelligence is infrastructural.
Capability is now a utility. The model is no longer the scarce input – access to it is universal and cheapening by the quarter.
02 · Judgement is industrialised.
Analysis, drafting, synthesis – the work that used to be the bottleneck – is produced at machine speed and volume.
03 · The layer with no instrument.
The layer between human awareness and built systems was never constructed – and no continuous, receiver-side instrument reads whether it holds. Outputs hold while systems fragment.
The gap
Outputs intact. Systems fragmenting.
Activity is up while coherence is down. AI is shipping while decisions are not landing. The signal degrades somewhere between intent and the hands meant to act on it – and nothing reads that degradation.
A century of system design assumed people adapt to systems. At machine speed, that default reverses.
The thesis · why now
Intelligence is becoming abundant. Coherence is becoming scarce.
That line is what this practice is built on – the category, in argument form. Read the thesis → · See the shift →
What you are not measuring
AI is amplifying the conditions you are not measuring.
What you already measure
- Revenue
- NPS
- Latency
- Throughput
- Accuracy
- Deployment volume
Mature instruments. Dashboards work.
What decides whether it holds
The Canonical Coherence Stack
Everyone is building Layers 1–4.
A fleet of well-governed agents inside an incoherent organisation compounds dysfunction at machine speed. Agent governance answers what an agent may do. Coherence OS answers whether the system around it holds.
The coherence layer (Layer 5) is the organisational layer that determines whether everything built below it produces coherent outcomes for the people it is meant to serve. Divergent Kind names it, builds it, and measures it.
One substrate, two configurations
An incoherent organisation cannot ship a coherent product.
Upstream
The Operating Model
- Signal routing
- Decision rails
- The coherence instrument, installed
- Where intent becomes structure
Downstream
The AI Product
- Receiver-adaptive delivery as design principle
- Participant Coherence: Signal Landing
- Meaning Formation · Action Coherence
- Trust Continuity
Same substrate, two configurations · one six-month engagement can deploy both
Where demand meets the discipline
Three currents. One layer. Three doors.
The field's own research keeps arriving at the same place: AI success is mostly the operating model, not the model; a large share of agentic projects will be cancelled for organisational reasons; and human-impact assessment of government AI is now mandatory, with dates attached. Three currents, one underlying layer – the organisation around the AI. That is the ground this practice works.
Door 01 · Organisations
For the organisations doing the work.
High-growth scale-ups and PE-backed portfolio companies hitting the structural complexity that arrives with scale – and boards that need the AI question answered coherently. Sponsored at founder, COO, CPO or director level. This is where the practice works today: the Coherence Audit in three to four weeks, then bounded change architecture that holds after we leave.
The shape of the engagement
- Fixed fee · 3–4 weeks · partner-led · board-ready
- Friction map · the Coherence Tax in dollars · 3–5 owned moves
- An honest decision gate: proceed, pause, or stop
- Expansion only where the read says it will hold
Door 02 · Government & public institutions
For government: engageable this quarter, not next panel cycle.
Agencies may engage Divergent Kind directly as an Australian SME – under CPR Appendix A Exemption 17 for eligible procurements to $500,000 federally, and NSW direct-negotiation provisions to $250,000 – subject to value for money and agency procurement requirements; no open tender needed for eligible engagements. The principal’s record includes work with Austrade (June 2019 – March 2020), the Commonwealth’s own trade and investment commission, alongside two decades inside Australia’s most regulated industries: aviation, banking, health insurance.
Specialist · time-bound · capability-transferring – aligned to the APS Strategic Commissioning Framework and the NSW Core Work policy. An Australian SME, ABN 30 693 755 672.
Door 03 · The AI-policy wave
The assessments are now mandatory. The capability to run them is scarce.
The Commonwealth’s AI policy v2.0 is already in effect, with mandatory requirements phasing in through 2026 – by 15 December 2026, in-scope Commonwealth AI use cases require pre-deployment impact assessment, alongside use-case registers, accountable officials and adoption plans. NSW’s assessment framework is already mandatory. The demand is policy; the pool of suppliers who have actually run receiver-side assessment of AI in production is small. We bring a ready qualitative method, facilitate the assessments with your people, and hand the method over – so the capability stays when we leave.
Now → 15 Dec 2026The Commonwealth compliance window: most agencies must produce artefacts they have never produced before, this calendar year.
Cth AI policy v2.0
Mandatory nowNSW’s AI Assessment Framework applies to all agency AI use, including procurement – and suppliers are evaluated on their fluency in it.
NSW AIAF
4–6 weeks, handed overFacilitated impact assessment, use-case register stood up, method transferred – built so it keeps working after we leave. The dependency never forms; that’s the deliverable.
The same receiver-side method serves enterprise AI teams shipping products under rising assurance expectations – one instrument, public and commercial registers.
Try the Lens · free, two minutes
Which dimension is most strained?
Four questions. A qualitative read – no scores, no scale.
Coherence Snapshot
When a decision is made at the top, what happens by the time it reaches the work?
Illustrative · qualitative only · no scoring logic exposed
How we work
Audit → Compass → Install → Evolution.
The instrument applied once, end to end – where coherence holds and where it fragments.
The operating architecture that resolves the friction the Audit names.
Decision rails, signal routing and instrumentation built into how the work runs.
Coherence measured continuously, not annually – the layer kept honest over time.
The sequence is the moat – principle → framework → protocol → vehicle. We don't build vehicles before the principle holds. The engagement, in full →
Where to start · the Coherence Audit
Three to four weeks. You leave with:
Most AI initiatives stall for organisational reasons, not technical ones. In three to four weeks, for a fixed fee, we show you where yours is most likely to stall – and the three to five moves that most improve its chance of holding.
You do not buy a dashboard. You leave with the read, the map, the decision gate, and the moves.
A friction map of your operation – where decisions degrade, where the frontline absorbs what the system should, where AI output stalls.
The Coherence Tax – what incoherence is costing you, counted in dollars from real events. Not estimated from vibes.
The 3–5 changes that recover the most, prioritised and owned – architecture, not a workshop.
Board-ready, with an honest decision gate: proceed, pause, or stop. We don't build before the read says it will hold.
Where to start
Four rungs – lowest commitment first.
The four-question Coherence Snapshot. A qualitative read of where your system strains. → you learn which dimension strains first.
Run the Snapshot ↑A focused read of a single artefact – a decision flow, a spec, a governance framework. → you learn whether it helps or hurts coherence.
Apply for a Read →The productised entry: Baseline · Friction Map · Priority Moves · Board-Ready Readout · Decision Gate. → you know what incoherence costs and the moves that recover it.
Book a Coherence Call →When it opens: the instrument deployed under licence, run by certified assessors under a published code of practice. → your team runs the read, the standard stays honest. Until the evidence is complete, we don't sell it – that discipline is the standard.
See what's coming →The standard
For the people building the systems.
If you are building AI into how an organisation works – a portfolio operator, an enterprise AI team, a governance body – you will be able to build on the standard directly. Divergent Kind will license the Coherence Lens and certify assessors to run it, under a published code of practice. You ship the systems. We keep the standard honest.
Status: the certification programme opens once the instrument’s reliability evidence is complete. The code of practice publishes first. Honesty about that sequence is the standard.
What you can build on
- The Coherence Lens, under licence
- Certified-assessor training & the quality floor
- A published code of practice
- Complementary to ISO 42001 procurement
The cost of incoherence
Incoherence has a price you can measure.
The Coherence Tax is counted from real events – margin that leaked, cost that was avoidable, decisions that stalled and what the stall cost. Not estimated from vibes. The same patterns repeat across organisations.
Evidence
What it looks like when the layer works.
The record behind the practice
Two decades inside the organisations everyone knows.
Divergent Kind is principal-led, and the practice draws directly on the founder’s record – design, product and operating-model leadership earned in roles and engagements at:
Roles and engagements held by the founder across his career. Listed organisations are part of the principal’s professional record – not clients or endorsers of Divergent Kind Pty Ltd.
What we stand for
Four rules the work runs on.
Coherence through Difference · Difference by Design
Cognitive difference is the substrate the design is built from – not the exception it accommodates after the fact.
Coherence is made structural in how the system is built, not asserted in a document no one reads.
The layer is instrumented and read over time – not surveyed once a year and filed.
We facilitate conditions for agency. We disown uses that extract from or distort the people inside the system.
The author layer
A discipline, not a service line.
Coherence architecture is a named discipline with a thesis, a canon, and an instrument – authored, defended, and published. Divergent Kind is its source. The thinking is open; the measurement is ours. Read the thesis, then decide whether you want it read into your organisation.
Led by Richard Lipp
Founder & Principal.
Founder & Principal · Operating Model Architect
Two decades across product, experience and operating-model design – Apple, Jetstar, the NAB Innovation Lab, Qantas, and Virgin Australia, where he built the airline's first internal product and experience team and contributed a step-change of more than $100 million in digital-channel revenue in a single year. A late neurodivergent diagnosis at forty became the catalyst for the practice.
Difference is not what you accommodate after the design. Difference is what the design starts from.
Read Richard's full story →Divergent Kind facilitates conditions for agency and coherence. We do not diagnose, treat, or cure.
The boundary is not a disclaimer – it is an architectural feature.
¹ MIT NANDA, The GenAI Divide: State of AI in Business 2025 – drawing on 150 executive interviews, a survey of 350 employees and 300 public AI deployments, the study found that about 95% of enterprise generative-AI pilots deliver no measurable P&L impact; only around 5% achieve rapid value. The report attributes the divide to integration, learning and workflow fit – not model quality.
² McKinsey & Company, Superagency in the workplace (January 2025, mckinsey.com) – almost all companies invest in AI, but about 1% of leaders describe their company as mature in AI deployment.
³ Boston Consulting Group, Where’s the Value in AI? (October 2024, bcg.com) – 74% of companies surveyed had yet to show tangible value from their AI investments.
⁴ World Economic Forum, Future of Jobs Report 2025 (January 2025, weforum.org) – 86% of employers expect AI and information-processing technologies to transform their business by 2030. All figures attributed and dated; they will be replaced by our own field measurements as engagements clear publication.