We build and run the AI team. You run the business.
An agreed engagement, not an application: a function, a team or a whole operating unit, run agent-first through the GenWA methodology in a controlled environment, with a small human core, live in twelve weeks. You pay per outcome delivered, with a cap you set. If you are founder-led and want senior AI leadership time instead, see the Accelerator.
An agreed engagement, no application · twelve weeks from definition to live · outcomes verified weekly from week seven
of agentic AI projects will be cancelled by 2027. Ours are checked weekly from week seven and stopped within a month if they are not working.
of organisations have had agents exceed their intended permissions. Every incubated agent has its own identity, a spend cap and a defined scope.
of internally built AI tools ever reach production. Ours follow the GenWA methodology, with the log and evidence packs already live, metering real client spend.
of UK AI adopters already report a cost reduction from AI. Inside the incubator that saving is priced and paid for outcome by outcome.
A function, a team, or an organisation.
Accounts payable. Customer communications. KYC and onboarding. Contact-centre wrap-up. Marketing operations. One function, its agents, its rules and its owner, run inside the incubator from week seven.
Comms · ContactA sales development team. A compliance monitoring team. A research desk. A team of agents with a team lead in your business, working to targets you set and outcomes you sign off.
GrowthA new venture, business line or back office run agent-first from day one, with a small human core and every agent run through the GenWA methodology. What a 2026 start-up looks like, built inside an established business.
Full suite · Managed OperationsTwelve weeks. Four stages. One record throughout.
Everything the platform gives you, applied to an organisation we run.
Every agent and every job is run through the GenWA methodology: defined jobs, named owners, enforced rules, and a check on every outcome. The same control layer as the rest of your business applies: identity, rules, data protection, model choice, metering, record.
Each agent has its own identity, scoped credentials, a spend cap and a scope of action. Destructive or customer-facing actions are held for a named human. There is a kill-switch.
New models, new agents and changes to either are tested against your data in the sandbox before they touch production. A model change is a governed event with a test and a record.
The log records every request, decision and outcome. Evidence packs map it to UK GDPR, the DUAA safeguards, the ICO code and ISO/IEC 42001. What you would show a regulator already exists.
A named person in your business owns every workflow and signs off every outcome. Our people run the agents; your people run the business. The IP, the prompts and the record are yours.
A set-up fee, a platform fee for the control layer, then a price per verified outcome with a monthly cap you set. Token cost sits with us, which is why we route it down.
It is yours. Take it when it holds.
The point of an incubator is that things leave it. When the function has delivered its outcomes for a quarter, it can stay in our cloud under Managed Operations, move to your cloud on the platform, or hand over to your team with the agents, the rules, the prompts and the full record. There is no lock-in, because the control layer is the same in every case, for the Incubator and the Accelerator alike.
Managed Operations and support tiersAn agreed engagement, not an application.
There is no application form and no selection panel. The Incubator is another way of working with GenWA, in a controlled environment we run to the GenWA methodology. It starts with a conversation, then a scoping session, then a written agreement. If you would rather have senior AI leadership time, see the Accelerator, which is by application only.
What we agree together before work starts
- The function you want run, and the outcomes it is measured on. We suggest the AI health check first, so the starting point is clear.
- A named person in your business will own each workflow and sign off each outcome.
- Your board or owner is content to run AI on that function.
- The work is not a decision on credit, hiring or medical matters, which are not permitted for AI by default.
- Your data can be reached securely, in your cloud or ours, as agreed.
- The fee structure and a monthly cap, agreed in writing.
- Operations are in the UK, where our regulation mapping applies.
Keep reading
About the Incubator.
What is the GenWA Incubator?
The Incubator designs, builds and runs an AI-operated function, team or operating unit for you, through the GenWA methodology with a small human core. The target is live in twelve weeks. You pay for verified outcomes, with a cap you set. It is an agreed engagement in a controlled environment, with no application.
How does the twelve-week Incubator timeline work?
Weeks 1 to 2 define the function and outcomes. Weeks 3 to 6 build the agents and controls. Weeks 7 to 12 run it live, with outcomes checked weekly against agreed definitions. From month 4 the function graduates to its long-term home. Timings are confirmed per engagement.
How is Incubator pricing structured?
A set-up fee, a platform fee, then a price per verified outcome with a monthly cap you set. Amounts are agreed in writing before you commit. Outcome prices and rules for held, blocked or reversed outcomes are to be confirmed with John White.
What if the Incubator does not deliver?
You would know within the first month. Outcomes are checked weekly against agreed definitions, and if the function is not delivering we stop. You pay for set-up and what was delivered. The reversal window and outcome price are to be confirmed.
Are we locked in after graduation?
No. At graduation the function can stay in our cloud under Managed Operations, move to your cloud or be handed to your team with the agents, rules, prompts and record. You own the IP, prompts and record. The control layer is the same in each case.
How does the Incubator start and what do you need from us?
It starts with a conversation, not an application. We scope the function you want run, and we suggest the AI health check first so the starting point is clear. In weeks 1 to 2 we define the outcomes, owner and data access, and agree the engagement in writing. You will need a named owner and a few hours a week.
The other four ways to work with us.
Most businesses start with one and add the others. Each has its own page. It is built for businesses with no in-house AI team.
Independent AI audit, policy and controls, training, strategy and planning. Fixed fees, credited against year one if you go on to the products.
See advisory services02ProductsAnalyse · Guard · Evidence · Automate · GrowYour rules, any model. Then agents on top of them.The control layer between your business and every AI lab, and a trained agent team deployed on it. Modular, priced as a range configured to your requirements, run in your cloud or ours. Built for a 10-person business as much as a 5,000-person one.
Explore the products04AcceleratorBy application only · Ten places a yearWe invest in you. You pay in equity, or by project, retainer or licence.Senior AI leadership time, advisory and platform access put behind a founder-led business. Paid for by equity, a project fee, a retainer or a licence, agreed in writing.
See the Accelerator05BuildScoped and priced in writingA senior engineer and an agentic team, in control.Bespoke software for a business-critical mission with a date that will not move. A senior engineer who specialises in your domain leads, agents do the heavy engineering, and control and human oversight are built in.
See BuildTell us the function. We will tell you what it would take.
Seven questions · an indicative scope and investment on the spot · no obligation