// Playbook 06 — Change readiness & risk
Landing the change honestly.
A six-week playbook for the change-readiness and risk work that surrounds the launch of a programme in the AI era — without making promises the programme can't keep. Listen first. Frame what's changing and what isn't. Equip what you can. Measure trust as carefully as you measure adoption. Worked example throughout: Halcyon Financial's CX tool pilot going live with twenty CX officers.
around launch
the pilot cohort
· adoption, value, trust
promises made
The launch.
Worked exampleA CX tool pilot meets the people who'll use it.
Halcyon Financial is fourteen weeks into the transformation programme set up in Playbook 04. The Lighthouse AI partnership from Playbook 05 has built the CX tool. It's about to go live with twenty CX officers. Maya Chen is one of the twenty.
Underneath the launch, the question every CX officer is privately turning over: what happens to us. The programme can't honestly answer that — no one at Halcyon yet knows what the workforce shape will be in eighteen months. What the programme can do is design the structure of the change so that the unanswerable doesn't break trust in the answerable.
The Transformation Lead has six weeks of change-readiness work that runs alongside and after the launch — starting two weeks before go-live. Listen first. Frame what's actually changing. Equip the cohort with capability, channels, and recognition. Measure trust as a leading indicator, not a lagging one.
programme
· including Maya
· not just the manager
· scheduled before launch
// And the people you'll meet
The Transformation Lead
You, the readerAccountable for nothing. Responsible for the change-readiness work alongside the launch. The role this playbook is written for.
Maya Chen
CX Officer · pilot cohortOne of the twenty CX officers in the CX tool pilot. Her KYC-notification idea — captured in Playbook 03 — is now publicly attributed to her in the CoE portfolio.
Priya Nair
Head of CXOwns the change at her function level. Holds the line on what's promised — and what isn't — when leadership pressure mounts to over-claim.
James Wong
Head of People & CultureJoins the cast. Owns the capability-uplift layer, the manager-coaching work, and the workforce-shape conversation that doesn't sit with Engineering or Compliance.
Sam Patel
Engineering DirectorAccountable for the CX tool from Playbook 05. Hands off from Lighthouse AI to the internal team at week 14 — the moment the change becomes Halcyon's to operate.
Anna Petrović
Head of ComplianceHolds the regulatory-risk line on what the CX tool is — and isn't — allowed to do once it's live. Sign-off on any customer-facing change.
Four phases. Listen, frame, equip, measure.
The framework// Change programmes don't fail at communication. They fail at credibility.
Listen
Weeks 1–2Diagnose what the workforce actually believes is happening. Map the trust state, not the org chart. Surface the questions no one's asking aloud.
Frame
Weeks 2–3Name what's changing — and what isn't. Sort what you can promise from what you can't. Design the structural answer to the unpromisable.
Equip
Weeks 3–4Capability that's applied, not just attended. Multiple feedback channels. Recognition that's portable, named, visible. Managers equipped to hold the change.
Measure
Weeks 5–6Adoption and value tracked separately. Trust as a leading indicator. The honest workforce-shape conversation, quarterly.
Phase one · Listen.
Weeks 1–2The goal of the first phase is honest current-state diagnosis. Not the surveyed state — the actual state. The workforce has already formed views, mostly from things leadership didn't intend as signals. Surface those before the launch, not after, or the change conversation will be a response to assumptions the programme didn't know it was making.
Diagnose the actual state, not the surveyed state.
Engagement surveys tell you what people will write when their manager will see it. Real diagnosis is anonymous, short, and asks questions people are already privately answering. Three open questions. One closed question on confidence. Aggregate, never attribute.
Map the trust state, not the org chart.
Who do staff actually believe when leadership communicates? Often it isn't leadership. It's a team lead two levels down, a long-tenured peer, an external voice in a Slack community. The trust map is who carries the signal — and it isn't always who the all-staff names.
Surface the questions no one's asking aloud.
Anonymous channel, two weeks open, single prompt: "What's the question about this you'd ask if no one could trace it back to you?" The answers are the diagnosis. Most will repeat — patterns matter more than uniqueness. The unspoken questions are the change-readiness inventory.
Three weeks before launch, a quiet inventory.
The Transformation Lead and James Wong open an anonymous question channel two weeks before launch. Single prompt. Twelve days. Two hundred and seventy-three CX staff in scope. One hundred and eighty-one responses. No identifiers, aggregated to themes by James.
Four patterns emerge. The dominant theme — present in 58% of responses — is the workforce-shape question, phrased thirty different ways. The second pattern (39%) is operational: how does this change my day. The third (24%) is fairness: who gets onto the pilot, why these twenty. The fourth (18%) is technical scepticism: will it actually work.
The trust map runs in parallel. Five-question anonymous survey to a stratified sample. Tom Nguyen (CX Team Lead) and two other named team leads carry the highest credibility. The most-trusted external voice is a Substack newsletter on AI-in-customer-service.
- ChannelAnonymous question form · 12 days · 181 responses out of 273 in scope
- Top theme"What happens to us" · 58% of responses · 30 different phrasings
- Trust mapTeam leads carry the signal · all-staff carries less than expected
- Read-outTo exec sponsor at the steering forum · unedited themes, no attribution
The first instinct after seeing the workforce-shape responses is to write a leadership statement reassuring staff. Don't. The statement will be unenforceable, and Maya — and every Maya — will read it as the moment the programme stopped being honest. Take the responses to Phase 2 and design the structural answer instead.
Phase two · Frame.
Weeks 2–3By Phase 2 you have the inventory. The next discipline is the cut: sort what you can honestly promise from what you can't, and design the structural answer to the unpromisable. The cut isn't just the right thing to do — it's what makes everything that comes later credible.
Name what's changing — and what isn't.
Two lists. Be specific. "CX officers now triage with CX tool assistance" is changing. "Final customer-facing decisions still sit with the CX officer, not the model" is not changing. The "what isn't changing" list is the more important one — it's the stable ground people stand on while the rest moves.
Identify the AI-specific risks — by category.
Four categories worth naming: model (does it work as expected), adoption (do people use it), workforce (how does the work shape change), regulatory (does it pass scrutiny). Each gets a one-line risk statement and an owner. Without categories, risk lists become wallpaper.
Sort what you can promise from what you can't.
Two columns. The "can't" column is sacred — anything in it is something the programme will be silent on, not loud. The "can" column is anything specific, falsifiable, and within the org's gift to deliver. Capability access. Pathways. Recognition mechanics. If you can't enforce it, it goes in the silent column.
Design the structural answer to the unpromisable.
For each thing in the "can't promise" column, design what you can offer structurally. The workforce-shape question gets a portable-recognition system, multiple feedback channels, capability access. The design replaces the promise. It's slower-acting and more credible — and once it exists, the silence on the unpromisable reads as restraint, not avoidance.
- What the workforce shape will be in 18 months.
- That no role at Halcyon will be affected as the work evolves.
- That every current task remains in its current shape.
- The exact career path for every officer in the pilot cohort.
- That AI tools won't change managerial decisions about team structure.
- Portable recognition — every contribution attributed by name, exportable, owned by the contributor.
- Multiple feedback channels — not just the line manager. Three routes for ideas and concerns.
- Capability access — every officer in the cohort funded for 16 hours of AI-adjacent training over six months.
- Honest reporting — adoption, value, and trust metrics published quarterly. Including when they're bad.
- A quarterly workforce conversation — chaired by P&C, including cohort representation. Visible, on the calendar.
A two-column document, signed at the leadership table.
The Transformation Lead, James Wong, and Priya Nair spend two sessions building the matrix. The first session generates a long "can promise" list. The second session strips it back — anything that can't be enforced inside the next twelve months gets moved to the "can't promise" side. The list gets shorter and more honest with each pass.
The exec sponsor signs the matrix at the steering forum. Two pages. Both sides of the line. The programme's posture from this point onward: silent on the left column, loud on the right.
Risks are mapped by category alongside. Model risk owned by Sam Patel. Adoption risk owned by Priya Nair. Workforce risk owned by James Wong. Regulatory risk owned by Anna Petrović. One owner per risk. No shared accountability.
Pressure from leadership or comms to "say something reassuring" about workforce impact in the launch message. The reassurance will almost always be unenforceable. The matrix is the answer. Point to it. The structural commitments on the right are the reassurance — they just don't sound like one.
Phase three · Equip.
Weeks 3–4Phase 3 builds the operating layer. Capability that's applied, not just attended. Feedback that has multiple routes, not just the manager. Recognition that travels with the contributor. Each of these is a small system. Done together, they're the structural answer the promise matrix points to.
Build capability that's applied, not just attended.
Training-as-checkbox is the most common adoption failure. Equip means time scheduled for use, not just for learning. Sixteen hours of training over six months, paired with sixteen hours of dedicated application time on the officer's own work. Without the second half, the first half evaporates inside two weeks.
Open multiple feedback channels — not just the manager.
If the line manager is the only route, anything sensitive doesn't travel. Three channels at minimum: direct to the programme (named or anonymous), peer-led (an embedded cohort lead from the pilot group), and P&C-confidential (the formal route). Make the routes visible and the SLAs honest.
Make recognition portable, named, visible.
Every contribution gets attributed in the public CoE log. Every officer in the cohort has an exportable record of what they shipped, what they learned, what they taught. The recognition belongs to the contributor, not the org. If they stay, it grows. If they leave, it goes with them. This is the load-bearing structural commitment.
Equip managers to hold the change.
Line managers carry most of the change conversation in 1:1s — and most of them have had no preparation. Two short coaching sessions before launch. Topics: how to acknowledge the workforce-shape question without faking an answer, how to surface the structural commitments, how to escalate concerns through the multiple channels. Without this, managers default to either over-promising or going silent.
Three channels, one portable record, and Maya's name in lights.
The Transformation Lead and James Wong set up three feedback channels two weeks before launch. Direct-to-programme (Slack form, named or anonymous). Peer-led (Tom Nguyen as the embedded cohort lead). P&C-confidential (existing channel, refreshed messaging). Each channel publishes a response SLA: 5 business days for direct, 48 hours for peer, 10 days for confidential.
Capability uplift is sixteen hours, paired with sixteen hours of application time on the officer's own work. Both halves are funded out of the programme budget — not absorbed into "personal development time" which never gets used.
Recognition is the structural commitment that takes the most resolve to actually ship. The CoE portfolio now publicly attributes Maya Chen as the originator of the KYC-notification idea — the workstream that landed three weeks ago and saved an estimated 80 CX tickets per week. Her name is in the post-mortem, the lessons-learned library, and the public dashboard. If she leaves Halcyon, the recognition leaves with her. That's the design.
Line manager coaching runs across two ninety-minute sessions in week 4. Eight team leads in the room. James Wong leads. The hardest part isn't the content — it's getting managers to admit they don't have the answers, and to be okay with that being okay.
Portable recognition is the most counter-intuitive commitment in the matrix. Leadership often resists it — "why are we giving them recognition they can take to a competitor?" The answer: because that's exactly what makes the recognition credible. A non-portable recognition program is an HR programme. A portable one is a covenant.
Phase four · Measure.
Weeks 5–6Most change programmes measure adoption and call it a day. Adoption is necessary but insufficient. The full picture needs three metrics moving in parallel: adoption (are people using it), value (is it producing outcomes), and trust (is the relationship between programme and workforce holding). Trust is the leading indicator — adoption and value follow from it, not the other way around.
Track adoption and value separately.
Same discipline as Playbook 04. Adoption is leading; value is lagging. Adoption tells you whether the equip-phase landed. Value tells you whether the work is actually better. Combining them lets you celebrate adoption that hasn't translated to value yet — and miss the gap.
Treat trust as a leading indicator.
Trust is measured by short anonymous pulse — three questions, repeated monthly. "Has the programme done what it said it would?" "Is the change conversation honest?" "Would you recommend the cohort experience to a peer?" Trend matters more than absolute score. A trust score that falls before adoption falls is the early warning system most programmes don't have.
Hold the honest workforce-shape conversation, quarterly.
Once a quarter, chaired by P&C, with cohort representation. Public agenda, published outputs. Not a forum for promises — a forum for showing what's known and what isn't. The conversation that the launch message can't be becomes the conversation the quarterly review is. The honesty is the credibility.
Three dashboards. One quarterly conversation.
The Transformation Lead publishes the first dashboard at week 4 after launch. Three metrics, three trends, three colours. Adoption is climbing — 35% to 60% to 75% across weeks 4, 8, 12. Value is trending — median time-to-resolution down 28% by week 12. Trust moves more slowly — 52% to 61% to 64%, with a known dip in week 6 after a Lighthouse-built component had a brief outage that affected three agents.
The first quarterly workforce conversation is held in week 14. James Wong chairs. Two officers from the cohort — Maya is one — sit at the table alongside Priya Nair, Sam Patel, and Anna Petrović. The agenda is public. The minutes are public. The promise matrix from Phase 2 is on the wall. The conversation reviews what's been delivered against the "can" column, and the structural answers to the "can't" column. No new promises are made.
The trust score dip in week 6 is discussed openly — what happened, what was done, what the score moved to by week 8. The discussion of the dip is what builds the trust back up — more than any reassurance could have.
The most counter-intuitive measurement choice is publishing trust scores when they're bad. Almost no programme does this. The ones that do see their trust scores recover faster than the ones that hide them — because workforce credibility is built by what gets published in the bad weeks, not in the good ones.
What you can't promise, you have to design.
Three cadences. Weekly, monthly, quarterly.
Operating rhythm// The change work has its own rhythm — separate from delivery, separate from steering.
The cohort check.
- Embedded cohort lead reports
- Adoption blockers surfaced
- Feedback channel volume
- Escalations to monthly
The trust pulse.
- Pulse-survey read-out
- Adoption vs value gap analysis
- Manager-coaching follow-up
- Channel-by-channel feedback themes
The workforce review.
- Public agenda, cohort at the table
- Promise matrix reviewed line by line
- Trust trend over the quarter
- Honest unknowns named
Four ways this fails.
Common pitfalls// Every change programme that lost the workforce failed in one of these four ways. Each has a structural counterpart — the fix is design, not communication.
The glossy disclaimer.
A leadership statement reassures staff in language they can see through. "We're using AI to make work better, not replace people." Within twenty-four hours, half the workforce has decoded it as the moment the programme started lying. Trust collapses before adoption begins.
The promise matrix. Quiet on what you can't honestly commit to. Loud on what you can. Frontline staff prefer no promise to one they can read through in a single glance.
Training as checkbox.
Sixteen hours of training delivered, completion rates 90%, applied-use rates 4%. The capability evaporates inside two weeks because no time was scheduled to actually use it.
Learn-and-apply pairing. Every training hour matched with a funded application hour on the officer's own work. Both halves come out of the programme budget — neither absorbed into "personal development time."
The single channel.
The line manager is the only route for feedback. Anything sensitive doesn't travel. Anything contradictory of the manager's view never reaches the programme. The diagnosis stops being honest in week three.
Three channels minimum, each with a published SLA. Direct to programme. Peer-led cohort representation. P&C-confidential. The manager remains one route — not the only one.
Trust as a lagging metric.
Trust gets measured once a year, in an engagement survey, after the change has already failed. By the time leadership sees the score, the cohort has either disengaged or quietly started planning to leave. The signal arrived too late to act on.
Monthly anonymous trust pulse. Three questions. Published — including when scores are bad. Trust as leading indicator beats trust as autopsy.
By week six, you should have.
Starter checklist// Twelve items. If you can tick all twelve, the change-readiness layer is operational. Trust will follow the structure, not the statements.
Using this in practice.
CloserCredibility is what's left after the message.
Most change programmes are remembered for what they promised. The good ones are remembered for what they designed. When the workforce can't be honestly reassured, structure is what's left — portable recognition, multiple channels, capability access, honest measurement. None of these say "trust us." Together, they make the trust earnable.
What travels is the shape — listen, frame, equip, measure — and the discipline of the cut: what you can promise, what you can't, what you have to design instead. Halcyon Financial's launch is one shape. Yours will be different. The posture should be the same.
If you're standing one of these up and want to talk through where it's getting stuck, I'm happy to.