Rewriting the Build
How the delivery lifecycle changes, from product and design through QA and operations.
With-AI operating model
Turn scattered AI activity into an operating model that works.
For leaders whose organizations already use AI in many places without an enterprise plan. Our With-AI operating model ties that activity to business objectives, sets how much AI can do based on evidence, and puts governance where the work happens. We start from what's actually running, and we implement what we recommend.
The problem
Most organizations have AI activity, not an AI operating model.
Teams build agents, skills, and tools on whatever models they choose. Few organizations can say what is running, what it touches, who is accountable for it, or whether it moves a business objective.
The approach
We start from what's running.
The operating model is built from your real AI estate, not from a strategy deck. We look at the AI work teams already rely on, then set direction, decision rights, autonomy, roles, and controls around it.
The six components
The operating model has six components.
01
The AI estate.
A working inventory of the agents, skills, tools, and models in use: what each one does, which data and systems it touches, and who owns it.
02
Direction and portfolio.
AI work is tied to business objectives. Leaders use one view of the estate to decide what to fund, what to scale, and what to stop.
03
Decision rights.
Clear answers to who can build, approve, deploy, change, and retire AI work, and who is accountable for it.
04
Earned Autonomy.
How much AI can do on a task is set by evidence. AI takes on more only when test results and review history support it.
Accountable decisions stay at level 2 or below
LEVEL 1
Assist
AI suggests or answers questions. The person does the work.
LEVEL 2
Draft and approve
AI prepares the work. The person reviews and approves every item.
LEVEL 3
Review by exception
AI completes routine items. The person reviews flagged items and a sample of the rest.
LEVEL 4
Act and audit
AI handles defined routine work. The person audits results and owns the outcome.
Person AI Share of the work, illustrative
A task moves up a level only when evaluation results and override rates support it.
Decisions that carry legal or mission accountability stay at level 2 or below.
05
People and routines.
Roles and routines change to match the work, from product and design through engineering, QA, and operations.
- Role and ceremony redesign: role by role, we sort tasks into what AI drafts, what a person reviews, and what stays with people, then change the handoffs and meetings to match.
- AI literacy: hands-on practice drafting, reviewing, and approving AI output on the team's own work.
- Playbooks: written guidance on what AI may draft, who approves it, and what gets logged.
06
Written policy, working control.
Policy is written down and built into the tools teams already use, as review steps, logged approvals, and tests. Governance sits where the work happens.
How an engagement starts
Engagements begin with an Estate Review.
- 01
Estate Review
A short review of the AI work already running: what it is, what it touches, who owns it, and which objectives it serves.
- 02
Pilot
We apply the operating model to one or two real use cases chosen from the review.
- 03
Implementation
We extend the model across the organization and put the decision rights, routines, and controls in place. We implement what we recommend.
What you get from the Estate Review
- An inventory of the AI work already running: what each item does, what it touches, and who owns it.
- Each item mapped to the business objective it serves, or flagged if it serves none.
- A starting Earned Autonomy level for each use case, based on the evidence available.
- A recommended pilot: one or two use cases, with the decision rights and controls they need.
Our own practice
Our own delivery teams already work this way: AI drafts code, tests, and documents, and engineers review.
Research
It draws on our Ground Truth research.
Rewiring the Role
How coordination roles change.
Earning Autonomy
Coming next.