How we work

Four stages. One accountable party across all four.

Diagnose, recommend, deploy, adopt. Most partners stop when the roadmap is handed over. That handover is the point where value usually leaks out, so it is not where we stop.

Stage 01

Diagnose

Two weeksOne functionOne sponsor

Read the function along both cuts — how it is organized, and how work actually flows through it. Score every task. Find where the hours go and what AI can actually take.

Stage 02

Recommend

In sequenceROI in rangesAssumptions named

Turn the evidence into resolved decisions: what changes first, what each move is worth in a range, what stays human on purpose, and what has to be true before it can run.

Stage 03

Deploy

Built to the designPartners namedChange communicated

The redesigned workflow gets built and rolled out against the design, with implementation specialists co-selected with you. The change is communicated from the same evidence.

Stage 04

Adopt

Capability where it movedAdoption tracked

The redesigned work becomes the way the organization operates — capability built where the work now needs it, and adoption tracked rather than assumed.

The claim is one accountable party through all four stages, not that we perform every specialist task ourselves. The evidence behind diagnose and recommend is strong today. The evidence behind deploy and adopt is structural — it is how the engagement is built, not a delivery record.

The entry point

A pilot, on one function, in two weeks.

One function. One executive sponsor. At the end of it the sponsor can make a real decision about what comes next, on evidence rather than on a proposal.

A full structural assessment of the function, with a viability verdict for every team in it.
One or two of the most critical workflows mapped end to end, current state and future state.
Three to five role blueprints for the people who operate those workflows.

Week one

Function head and senior leader interviews.
The starting picture of the function, team by team.
The one or two workflows you cannot afford to get wrong, named.
Preliminary verdict on every team.

Week two

Interviews with the people who actually run those workflows.
Current state and future state, step by step, scored.
Team verdicts finalised in the last two or three days, after the workflow evidence lands.
Role blueprints started only once the verdicts hold.

Day fourteen

A decision the sponsor can actually make about what comes next — on evidence, not on a proposal.

The structural picture is finished last, not first. That order is deliberate: it stops the role work resting on assumptions the workflow evidence would have revised.

Scope and timing for the full engagement follow the pilot, because they are a function of what the pilot surfaces.

Data

A redesigned workflow that depends on information you cannot produce reliably will not run.

So data is a named part of the work, with a clear division of labour: we assess what each redesigned workflow needs from your data, we tell you what blocks it and what that block is costing you, and we bring a specialist for the build.

We deliver

Whether the information each workflow depends on exists, connects, and is reliable enough to act on.
The data each future-state workflow rests on, specified.
Gaps ranked by the workflow they block and what that workflow is worth.

Partners deliver

Pipeline build, integration, migration, data engineering.
Data governance design, model risk management, security.

What is different about this

Three things a buyer can check.

Read at task grain, not at framework grain

Most diagnostics sample a few interviews and map the result onto a framework. We decompose the work itself — every task in the roles that run the workflow, scored — so the finding survives someone senior pulling on it.

A resolved decision, not a map

A live map never resolves. What you get is a small set of decisions in the order they have to be made, each carrying its verdict, the bets it rests on, and the evidence path down to the task it came from.

One accountable party across all four stages

Not the most people in the building — the one who carries the consequence. Deployment partners, training partners and the data specialist are named, not hidden. And we do not ask for a fee we have not first quantified against a loss you are already carrying.

Deploy and adopt

The value is not created when the analysis is finished.

It is created months later, when the redesigned work is simply how the place operates. That is several handovers past where a diagnostic engagement ends, and every handover is where the value leaks. So we do not hand over.

The change is communicated from the evidence

The change narrative for the function, briefing material at team and role level drawn from the same analysis, and honest answers to the questions people will actually ask — including whether their role survives it. We are not writing reassurance. We are telling people what the work becomes.

Capability is built where the work moved

The role blueprints already say which human capabilities become more valuable as AI absorbs the routine volume. Learning is designed against that, for the people on teams that are staying, delivered through training partners.

Adoption is tracked, not assumed

Whether the redesigned workflow is actually being run the way it was designed, where it is drifting back, and what that costs. A feedback route from the manager conversations back into the work.

Why this sits with us rather than a communications agency or a training vendor: the answers are only as good as the task-level evidence behind them, and we hold that evidence.

The diagnosis is where you start. Staying through adoption is why it is worth starting with us.

Common questions

The questions we actually get asked.

Q01How is this different from hiring a consulting firm?

A firm sends a team, samples a set of interviews, and hands you a roadmap. We decompose the work itself — every task in the roles that run the workflow, scored — and we stay past the handover into delivery and adoption, because the handover is where the value usually leaks. On coverage of a large multi-year programme they out-resource us and we say so. On whether the decision holds up when your CFO pulls on it, the task-level evidence is the difference.

Q02How long does it take?

The pilot is two weeks on one function, with one executive sponsor. At the end of it you can make a real decision about what comes next. The full engagement is scoped after the pilot, because its size is a function of what the pilot surfaces.

Q03What does it cost?

The fee is bounded by the loss the redesign avoids, and we scope it once we understand your work. If we cannot quantify a loss you are carrying, we say so and we do not take the engagement. Training programmes are priced separately and never blended into the same figure.

Q04Do we need our data in order before you start?

No. Data readiness is part of the work, not a prerequisite for it. We assess what each redesigned workflow needs from your data, tell you what blocks it and what that block is costing you, and bring a specialist for the build. What we do not do is design a workflow that depends on information you cannot produce.

Q05What happens to people in this?

Our work is framed at the level of capability and work, never at the level of an individual. We name which capabilities now take precedence and which are no longer worth selecting for, and every hour AI frees carries a named destination. People inside a transformation ask whether their role survives it — we answer that question rather than stepping around it, at the level of the work.

Q06Who owns what you produce?

You do. The blueprints, the workflow designs, the assumptions behind every number. They are built to be re-read by your own people a year later and to be updated as AI capability moves, which is the point of writing them down in a structured form rather than as slides.

Q07How do you keep from inventing numbers?

Every score uses a published rubric with worked examples at each level. Every ROI figure is a range with its assumptions on the surface and upside kept separate from the base case. No consequential verdict rests on one source. And where something is a design estimate rather than a measured figure, it says so.

Book 30 minutes