Transformation roadmap

Move by evidence, not enthusiasm.

Agentic transformation is not a race to deploy. It is a sequence of decisions that earns the right to delegate more consequential work.

01 · Orient

Map current AI use, shadow activity, critical workflows and material exposure.

Work of the stage

Inventory agents, copilots and shadow use
Identify material workflows and exposure
Name an executive sponsor and working group

Evidence to carry forward

A current-state map, issue register and clear mandate.

Leadership gate: Do we know where AI already influences or performs work?

02 · Focus

Choose a small portfolio tied to consequential enterprise outcomes.

Work of the stage

Select one or two consequential domains
Define the outcome and baseline
Choose work bounded enough to test safely

Evidence to carry forward

A focused portfolio with accountable business owners.

Leadership gate: Is the work important enough to redesign and bounded enough to test?

03 · Design

Specify authority, workflow, context, controls, ownership and success evidence.

Work of the stage

Redesign the workflow end to end
Set authority and escalation conditions
Specify context, controls and success evidence

Evidence to carry forward

A complete six-system operating design.

Leadership gate: Can leaders explain how the system should behave when conditions change?

04 · Prove

Operate in controlled conditions and test normal work, exceptions and intervention.

Work of the stage

Run normal, edge and failure scenarios
Test intervention and rollback
Compare outcomes with the agreed baseline

Evidence to carry forward

A decision record to expand, revise or stop.

Leadership gate: Did it produce acceptable outcomes within agreed boundaries?

05 · Scale

Reuse proven controls, interfaces, data services and operating patterns.

Work of the stage

Turn proven patterns into shared services
Standardize identity, policy and evidence
Fund adoption as operating capability

Evidence to carry forward

Reusable paths that make governed delivery easier.

Leadership gate: Can adoption expand without weakening accountability?

06 · Evolve

Review evidence, incidents and changed capabilities; revise or retire designs.

Work of the stage

Review outcomes, incidents and capability shifts
Reauthorize material changes
Retire controls or systems that no longer fit

Evidence to carry forward

A living operating model with visible revision history.

Leadership gate: What should now be expanded, constrained or stopped?

A practical first 90 days

Establish the right to prove.

Days 1–30 · Orient

Make existing activity visible. Choose one enterprise outcome, document its current workflow and identify where AI already shapes decisions.

Days 31–60 · Focus and design

Define the operating envelope: authority, accountable owner, context, controls, exception paths and evidence of acceptable performance.

Days 61–90 · Prove

Run controlled work. Exercise failure and intervention—not only the happy path. End with an explicit expand, revise or stop decision.

Assess your starting point