A capability assessment is the easiest piece of work in any transformation to finish and the hardest to land. Finishing it is a defined job: pick a model, score the organisation against it, argue about the scores in a steering committee, publish. Landing it is a different job, and nothing in the assessment phase prepares anyone for it. Six months later the same organisation commissions a similar exercise, because nobody could point at a budget line the first one moved.
I worked on a digital readiness review at a European composite insurance group running several distribution labels, and the page I still go back to is not one of the scored ones. It is the operating plan for what happens after the scoring. Four phases, each with a short list of activities and a named artefact at the end. The first three carry a week count. Four weeks of assessment, already complete when the plan was written. Six weeks to widen it. Six weeks to convert it into a roadmap. Then execution in increments of three to six months.
Sixteen weeks of that is analysis, and four of the sixteen are the assessment everybody thinks the engagement is.
Four of sixteen weeks are the part people buy
The arithmetic inverts how these engagements are usually scoped. Phase 0 ran four weeks: assess capability with IT stakeholders, define hypotheses about where the gaps sit, form an initial view of the initiatives that would close them. Its artefact was the first capability heatmap, a complete and presentable piece of work, and in most organisations it is where the engagement stops.
Phase 1 then ran six weeks and Phase 2 another six. Both are longer than the assessment, and neither produces a new score. Phase 1 widens the evidence base and settles what the organisation is aiming at. Phase 2 turns the gap between those two things into a costed sequence of work.
Scoring is the visible deliverable and the cheap part. Three quarters of the calendar sits after it, doing work nobody puts in a proposal because it has no photogenic output.
The first assessment was scoped to whoever would answer
Phase 0 is described plainly in the plan as an assessment with IT stakeholders. Not with the business. Not across the value chain. With IT.
That is an honest constraint rather than a mistake. IT stakeholders are reachable in four weeks, hold a view across the whole estate, and will answer a structured survey. Claims, underwriting, marketing and the back office of several separate brands take longer than the phase had.
The consequence shows up in the evidence itself. The capabilities furthest from IT, the process and culture ones, came back on visibly thinner samples than the technology capabilities did, and the review says so in its own commentary rather than hiding it. An assessment inherits the shape of whoever answered it, and the domains where IT has least visibility are exactly the ones where a technology-sourced reading is least reliable.
The transferable part is what it does next. It does not extrapolate, and it does not caveat the first heatmap and move on. It books a second six-week phase to run the assessment again across the full retail property and casualty value chain, separately for each label and separately for the back office. The first heatmap is treated as a hypothesis about the organisation, and the second is treated as the baseline. The word the plan uses for its Phase 1 artefact is validated, and validated here means re-collected from a different population, not reviewed by a wider circle.
Anyone running an AI readiness assessment in the next quarter is about to make the same first move for the same reasons, and most will not book the second phase.
The sign-off gate sits before the prioritisation
The other Phase 1 activity has nothing to do with measurement. Alongside the widened assessment, the phase aligns the digital aspiration with corporate strategy and gets a group-level framing document signed off at the executive strategy setting.
That ordering is the part I would defend hardest in a scoping conversation. If direction is still open when prioritisation starts, every ranking argument in the roadmap phase becomes a proxy argument about strategy, conducted by people with no authority to settle it. Six weeks disappears into that. Settling the aspiration first leaves the roadmap phase arguing about sequence, dependency and what fits in the first increments, with the signed framing document as the tiebreaker.
The deliverable of the roadmap phase is not a roadmap. It is a change to a document the assessment team does not own, cannot write alone, and has to be let into.
The roadmap phase delivers an edit to somebody else's document
Phase 2 has four activities: determine the key capability gaps, inventory the digital initiatives already running, define and prioritise the initiatives that close the gaps against the signed framing document, then integrate the result into the overall roadmap and lock it into the division business plans.
The gap arithmetic and the portfolio inventory are substantial pieces of work in their own right, and I have written elsewhere about what happens when you credit every funded programme in full and measure what survives, in Ten Capability Gaps Survived Crediting Every Funded Programme in Full. The structural point in this phase is the last activity, and specifically the object it acts on.
Every other phase in this plan produces something the review team can write: a heatmap, a validated heatmap, an alignment. Phase 2 does not. Its named artefact is the roadmap integrated into the division business plans, plus detailed project plans and business cases for the first increments. That business plan is not the assessment team's document. It has an owner, an annual cycle, a finance process attached and a queue of people already trying to get into it. Naming it as the deliverable turns the last six weeks from a writing task into an access problem, to be solved while the analysis is still warm and the sponsors still care.
That is a different discipline from naming an output for every phase, which is ordinary good planning. This names an output the team cannot produce unilaterally. The plan is deliberately underwriting a negotiation.
- 4 weeks
- Assessment phase, already complete when the plan was written
- 6 weeks
- Widening the assessment across the business and settling the aspiration
- 6 weeks
- Gap analysis, prioritisation and entry into the division business plans
- 3 to 6 months
- Length of each execution increment that follows
What the source does not say
Set the tense honestly. Phase 0 was complete and marked as such on the page. The three phases after it were scheduled, with dates, and the page exists because the follow-up still had to be agreed. Nothing in the source records that any of them ran, and no delivered outcome can be read out of it.
The word funded, in particular, does not belong on this page. Phase 2 produces business cases for the first increments, and a business case is an argument for money rather than money. The plan books the writing of them and says nothing about approval. The honest description of the target state is a roadmap sitting inside the division business plans with costed cases attached, which is further than most assessments ever reach and still one committee short of funded.
Phase 3 is sketched rather than specified: governance, indicators and a programme office, two organisational capabilities on the critical path, and the first initiatives launched in increments of three to six months. The argument for why digital and agentic work needs its own delivery lane, running beside the standard change process rather than inside it, is made in Why a Quick Win That Ends in a Demonstration Never Reaches the Business Case. What matters here is only that the plan puts organisational build in a phase of its own with its own clock, rather than assuming the existing process absorbs the work.
What I would copy
Scope the assessment to the population you can reach inside the window, and name that population on the page where the scores appear. A readiness score sourced from platform and data teams is a real reading of the platform and a weak reading of the operating model.
Book the second collection before you present the first. If widening the evidence base is a phase with a week count, it happens. If it is a recommendation in the closing section, it does not.
Settle direction before you rank anything. Get the aspiration signed at the level that can sign it, then let prioritisation be about sequence.
Name the document your roadmap has to enter, by its real name, at kickoff. For most organisations that is the divisional or business unit plan and its budget cycle, and the cycle has a date on it that no amount of good analysis will move.
And write the phase lengths down. Four, six and six is not a magic sequence, but a number on the twelve weeks after the assessment stops them being absorbed into whatever comes next.
Source: the follow-up phasing page of a digital readiness review at a European composite insurance group, with its per-phase activities, week counts and named deliverables as recorded. Phase 0 was complete when the plan was written; the three phases after it were scheduled, not delivered, and no outcome is claimed for any of them. The plan names business cases as a deliverable and records no approval, so this piece does not describe the roadmap as funded. Reading the phase structure as a mechanism for landing an assessment, and the business-plan integration as an access problem rather than a writing task, is ours.
“The deliverable of the roadmap phase is not a roadmap. It is a change to a document the assessment team does not own, cannot write alone, and has to be let into.”
Get in touch
Put RealAI’s applied-AI team on your hardest data problem.
We help enterprises move from pilots to production: sovereign models, governed data, and agents you can audit. Start with a value-first assessment.