// A six-playbook methodology
Transformation in the AI era.
Six playbooks for the work of taking an organisation from AI ambition to operating reality. Diagnose the maturity. Set the OKRs. Stand up the portfolio. Frame the programme. Pick the partner. Land the change. Read together, the six also take a position — about what's actually being transformed, who does the transforming, and how the language we use about it shapes how the change feels for the people living through it. Updated for Bain's AI value paradox — value first, tools last, and a way out of the use-case swirl.
· one body of work
· four phases
· Halcyon · Tallowood
· applied throughout
Where this starts.
A small phenomenon, in hospitalsIn hospitals, the difference between "the diabetic in bed four" and "Maria, who manages her diabetes around her shift schedule" shows up in measurable ways — trust, adherence, the quality of care she receives. Same person. Same medicine. Same diagnosis. The language is doing real work.
AI-era transformation has the same property. The way we describe the people doing the work shapes how the work feels to do — and how the change feels to live through.
Most transformation prose makes AI the subject. "The CX tool transforms how your team works." "AI accelerates productivity for your front line." "Agents will redefine customer service." It sounds neutral. It isn't. A sentence with the tool at the front and the person somewhere in the background tells the person on the receiving end something specific — whose work is being talked about, and whose isn't.
This methodology is built on the inverse. Maya triages tickets using the CX tool. Tom holds the harder conversations in 1:1s. The engineering team prototypes faster with AI-assisted tooling. Where there's a choice about who acts in a sentence, the person wins.
It isn't a stylistic preference. It's the same thesis the six playbooks argue at different scales: the people doing the work are protagonists, not background. Treat them as protagonists in the language, and the rest of the architecture of the change starts to follow.
Where you start.
Six stages · six dimensions · honest baselineMost organisations don't start at zero — but they don't start at AI-native either. The journey from pre-AI to AI-enabled to AI-native is uneven: strong in one dimension, absent in another. The grid below names where you actually are, per dimension, so the work goes where the gap is. Six stages across, six dimensions down. Score each dimension independently. Most organisations sit in stages two to four, with at least one row a stage behind the others. The unevenness is the diagnosis. Playbook 01 is the playbook for running this assessment honestly.
The six playbooks.
Six weeks each · four phasesEach playbook follows the same skeleton — six weeks, four phases — and runs the same worked example. Halcyon Financial, a regulated digital financial services platform building out a transformation program. Same cast across the six. Different problem each time. Same discipline. Read in order, the six cover the first twelve months of a transformation lead's tenure — from honest diagnosis to landing the change.
Running the maturity assessment.
A six-week playbook for honest diagnosis before the transformation starts. Score against the maturity model to surface where the organisation actually sits. The output is the input to Playbook 02.
Open playbook →Strategic planning · setting honest OKRs.
A six-week playbook for turning the maturity diagnosis into three to five OKRs. Calibrated confidence, backcasted narratives, a premortem on the slate. An OKR is a bet, not a promise, and the difference shows up at year end.
Open playbook →Building a continuous improvement centre of excellence.
A six-week playbook for standing up a CoE that captures ideas at scale, scores them honestly, and ships the right ones. The operating engine the rest of the track stands on. Maya Chen's KYC-notification idea is the worked example.
Open playbook →Setting up an "AI-Native" program.
A six-week playbook for turning a strategic ambition into execution. Decomposed into workstreams, the first three scoped with stop conditions, operated on a cadence that doesn't bury the team.The CX tool pilot starts here.
Open playbook →Selecting a transformation partner.
A six-week playbook for selecting and onboarding a delivery partner without losing optionality. Frame, shortlist, evaluate, onboard. Lighthouse AI is the partner Halcyon selects — the worked example walks the full discipline.
Open playbook →Landing the change honestly.
A six-week playbook for the change-readiness work that surrounds launch — without making promises the programme can't keep. The CX tool goes live with twenty CX officers; Maya is one of them. The most workforce-sensitive of the six.
Open playbook →Getting past the swirl.
Aligned to Bain & Company · Sep 2026Bain names a pattern most transformation leads will recognise: frenetic AI activity, flat earnings. Hundreds of ideas. Dozens of pilots. A copilot on every desk. And no line on the P&L that moved. They call the root of it use-case swirl — ideas generated faster than they can be prioritised, with resources spread across so many experiments that none gets enough attention to matter. Activity isn't value. This section sets out how the six playbooks answer the five obstacles Bain identifies, the three amendments that tighten the methodology against them, and a second worked example — an organisation deep in the swirl.
Five ways the swirl holds.
Value first. Tools last.
Start with the value plan.
Isolate the three to five opportunities already in the strategy. Diagnose against those — not against the use-case register.
Playbooks 01 · 02Redesign before tool selection.
Rebuild the priority workflow as if starting from scratch with AI in hand. With the people who run it today.
Playbook 03Blueprint and sequence.
Turn the redesign into requirements. Prototype to validate. Then concentrate resources on production — not pilots forever.
Playbooks 04 · 05Manage the change.
Product discipline and organisational change, together. Scaling is a people problem before it's a platform problem.
Playbook 06- Anchor. Every key result names the line on the value plan it moves. If it can't, it's a micro-productivity result — useful, but not a transformation objective.
- Order. Workflow redesign precedes tool selection. Playbook 05 doesn't open until a Playbook 03 or 04 blueprint exists.
- Stopping. Every pilot carries a stop condition and a production decision date. A pilot with neither is part of the swirl.
Eighty-seven use cases. Nothing on the results slide.
Tallowood Bank is an ASX-listed regional bank, built from a regional bank and two customer-owned banks — around 3,400 people and 1.2 million customers. The AI register holds 87 use cases. Twenty-three pilots are live. Four tools are licensed. A copilot sits on every desk.
At the board, the Chair asks which of the strategy's value drivers any of it moves. No one can answer. AI sits with the CIO. The CEO attends the quarterly showcase and leaves.
Start from the value drivers, not the register. No private equity owner, so no value creation plan — the anchor is the investor presentation: cost-to-income, time to yes on home loans, customer retention. The maturity assessment scored against those — Data & Architecture came in at Stage 02, with customer data on three core banking platforms.
The CoE forum re-scored all 87 against the drivers. Home lending specialists redesigned the loan workflow end-to-end before any vendor was briefed. The CEO took the sponsor seat. The CIO took the sceptic's.
87 use cases became 4 funded bets — home loan document checks, a customer retention early warning, supplier consolidation and fee-leakage detection. Nineteen pilots stopped. Four licences became one.
Time to yes on a standard home loan is targeted from nine days to three. Every key result names its value driver. The showcase became a quarterly decision log: what shipped, what stopped, why.
Moved from attending the showcase to owning the four bets.
Asks one question every board: which driver, what evidence?
Owned AI by default. Now owns the platform — and the stop conditions.
Redesigned the loan workflow with four peers. Knew where the nine days went.
Does this hold for a listed company?
Mostly. The diagnosis travels. The forcing function doesn't.
- The data isn't PE-only. Bain's CEO survey covers chief executives across the economy. The swirl is an organisational pattern, not an ownership one.
- The five obstacles are ownership-agnostic. Prioritise to a few, redesign before tools, fund production over pilots — all of it applies wherever there's a P&L.
- Sponsorship matters more, not less. In a larger, federated company, AI handed to IT gets more diffuse, not less.
- No value creation plan. A listed company has a strategy, guidance and the drivers analysts model — broader, slower-moving, contested across business units.
- No exit clock. A finite hold forces the cut to four. A listed company has half-yearly or quarterly reporting and a perpetual horizon — pressure to show AI activity each period, and room to fund longer bets. Both can feed the swirl.
- No operating partner. Ownership is dispersed. Executive incentives (STI, LTI, TSR) tie less directly to EBITDA from any single programme.
- Disclosure risk. AI claims reach the market. Continuous disclosure applies, and regulators have started acting on "AI washing" — the SEC's first cases came in March 2024.
- Swap the anchor. Replace the value plan with the three to five value drivers in the investor presentation. Hold AI work to the same line of sight.
- Make the board the operating partner. A standing question at every meeting: which driver, what evidence, what stopped.
- Say what's in production, not what's in pilot. Route AI claims through disclosure review before they reach an announcement.
- Use the longer horizon on purpose. Fund the Data & Architecture gap as a named bet, not a hidden cost.
The thinking underneath.
Four durable ideas · borrowed honestlyTransformation in the AI era runs on ideas that pre-date AI but matter more now than ever. The technology has changed faster than the practices around it. The four ideas below aren't novel — they come from people who've thought about complex work for decades — but they're the load-bearing beliefs the methodology rests on. Each playbook puts them into operational mechanics. They're surfaced here so the work is honest about its sources.
The people doing the work are protagonists.
Most transformation prose makes the technology the subject of the sentence. The methodology inverts that — at the grammar level, in the assessment, and in the design of the work itself. Maya triages tickets using the CX tool, not the CX tool triages tickets for Maya. The order matters because it tells the person on the receiving end whose work is being talked about.
// Source — the position above, articulated in full in Section 07.Not every idea is the same kind of problem.
Best-practice playbooks work in known territory and actively mislead in complex territory. Most AI-era improvement work is genuinely complex — cause and effect can only be understood in retrospect — which means you probe and adapt rather than plan and execute. Sorting the problem before solving it is the discipline that protects effort from being routed at the wrong tool.
// Source — Dave Snowden's Cynefin framework. Used in Playbook 03 at the assess phase.A good call can still have a bad result.
When a high-scored initiative fails, the question isn't "was the call wrong?" — it's "was the reasoning sound given what we knew?" AI tools generate confident outputs that may be wrong, which makes calibrated confidence and the discipline of separating decision quality from outcome quality more valuable now, not less. The prioritisation forum's job is accuracy, not agreement.
// Source — Annie Duke, Thinking in Bets. Used in Playbooks 02 and 03.Most frameworks assume teams, instead of building them.
Popular Agile and Lean approaches grant an implied teamwork waiver — they describe the way teams work, not how they work. AI amplifies complexity, which means the human side of the work — distributed leadership, real teaming, psychological safety — becomes more important, not less. Build the team capability deliberately; don't assume it.
// Source — Turner & Thurlow, The Flow System. Used across all six playbooks.Measuring what changed.
Three layers · resist measuring moreMost measurement frameworks in this space collapse under their own weight — thirty metrics, half of them vanity, none of them changing a decision. The bar is simple. If a number isn't going to change what you do, don't track it. The methodology runs on three layers: a baseline captured once, a small set of leading indicators watched during the work, and lagging outcomes reviewed at 90 and 180 days. That's it. Everything else is noise.
What you measure once, before the work starts.
Four numbers, captured imperfectly. The point isn't precision — it's having something to compare against in ninety days. Baseline is also the moment to capture the maturity score from Section 02, which is Playbook 01's primary output.
- Idea velocity (per month)
- Time-to-decision (median days)
- Throughput (per quarter)
- Active WIP (per team)
What you watch during the work.
Signs of life. Tracked weekly or monthly. These tell you the system is operating — not yet whether it's working. If these flatline, intervene before the lagging numbers catch up.
- Submissions per week
- SLA hit rate (5-day ack)
- Decision-log entries
- Forum attendance (5–8)
What you measure at 90 and 180 days.
The evidence the system is working. Compared against baseline, not against an aspiration. Add one qualitative pulse: do people trust where the work is going? A single yes/no question, asked of a representative cross-section. The number itself isn't the point — the trend is.
- Velocity vs baseline (3× target)
- Time-to-decision (<15 days)
- Throughput vs baseline (2×)
- Decline ratio (30–50%)
How we write about people.
Three small habits · and whyWhen a leader says "tough decisions had to be made," everyone in the room hears that no one wants to own them. When the same leader says "I decided to close the Brisbane office," they hear something different — even if the outcome is identical. Grammar carries who's responsible. Three small habits run through this methodology. None of them are about being clever with words. All of them are about the person on the receiving end — what they hear, what they trust, what they have to read between the lines for.
Put the person at the front.
When the AI sits at the front of a sentence, the person doing the work disappears into the background. The person reading it notices. "The CX tool triages tickets for Maya" tells Maya she's the recipient of automation. "Maya triages tickets using the CX tool" tells her she's doing the work — with a tool. Both are true. Only the second is honest about who's doing what.
"The CX tool triages tickets for Maya."
Name people as people.
In 2026, the word agent reads as AI by default. A CX officer reading "twenty agents will use the CX tool" has to do quiet double work to figure out whether that's AI agents using a CX tool, or human agents using one. The ambiguity isn't fatal. It's just constantly tiring. A small editorial choice — officers, engineers, team leads, or simply people — saves them the tax. People are named or role-titled. AI systems are named for what they are.
"Twenty AI agents and twenty agents will use the CX tool."
Transformation is what the organisation does.
AI transformation is the phrase the market uses. It implies the organisation is being transformed by AI — the vendor's product moves to the centre of the sentence; the organisation moves to the side. Transformation in the AI era puts the noun where the work actually lives — what the organisation is doing, with AI as one of the conditions of the era it's doing it in. A small reframe. Costs nothing to adopt. Stops centring whoever sold the AI.
"Halcyon's AI transformation."
What this becomes.
Three things · each more visible than the lastFor a Transformation Lead reading this methodology for tactics, the language work above is a small detail. Useful for clarity, not the centre of the work. For an organisation that adopts it at the leadership level, it becomes something larger — three things that build on each other, each more visible than the last.
A small habit, every day.
CX officers instead of CX agents. The CX tool instead of the agent. The kind of choice one team can adopt this afternoon and one editor can review on Friday. Low-cost, low-ceremony, immediate in effect. The reading-friction it removes is small per sentence and large over a quarter.
Trust, made visible.
When the same habit is adopted in change comms, system documentation, vendor briefs, and leadership statements, something quiet happens. Staff can read the way the organisation thinks about them. They don't have to be told they're valued — they can check. Anyone reading any piece of internal comms can ask the same question — am I the subject of this sentence, or the object? — and find an honest answer.
This is the structural answer to the problem most transformation programmes get stuck on: how to communicate respect during a time of change you can't fully promise about. The usual move is repeating the unenforceable — no one will lose their job, we value our people, this won't change anything important. Staff read those statements once and discount them. The alternative is to put the respect into the architecture of the language itself, where it can be seen and verified day to day.
How the organisation is known.
"We're an organisation that writes our people as the subject of their work." The same habit, scaled, becomes a cultural marker. Visible in job ads. Visible in how vendors are briefed. Visible in board papers. Visible to prospective hires and to staff who stay. It isn't a slogan — it can't be — because it's verifiable in any document anyone picks up. Most consultancies don't take a position on this; they use whatever framing the vendor's marketing arrived with. Most organisations don't either. The asymmetry is the opportunity.
People read what we mean, not what we say.
Using this in practice.
A methodology, not a prescriptionA methodology, not a prescription.
Each playbook is written for a Transformation Lead inside an organisation reading what's actually in front of them — not for a consultancy selling a method. The shape travels; the specifics don't. Halcyon Financial is one shape. Tallowood, stuck in the swirl, is another. Yours will be different. The discipline will be the same.
The cast carries across the six — Maya Chen, Tom Nguyen, Priya Nair, Sam Patel, Anna Petrović, Rachel Doyle, James Wong. They're composites, not real people. They're there because transformation work is people work, and writing about it without specific humans in the room produces the kind of advice that doesn't land in the room.
If you're standing one of these up and want to talk through where it's getting stuck, I'm happy to.