How we work
A delivery lifecycle designed for systems that go to production.
Every engagement follows the same seven stages. For agentic systems we add a gate: supervised first, evaluate, then earn autonomy.
Lifecycle
Seven stages, from discovery to continuous improvement
- 01
Discover
Understand the business process, users, systems, constraints, and the outcome you want.
- 02
Map
Document the current workflow and identify where automation and AI genuinely help.
- 03
Design
Define architecture, UX, data flow, agent boundaries, permissions, and human approval points.
- 04
Build
Develop in iterative, production-focused sprints with working software reviewed every cycle.
- 05
Test & evaluate
Test workflows, edge cases, model behaviour, permissions, and failure modes.
- 06
Deploy
Launch with monitoring, a controlled rollout, and a rollback plan.
- 07
Improve
Track performance, cost, reliability, and user feedback — and act on it.
The autonomy gate
Supervised first → evaluate → earn autonomy.
Agents do not start autonomous. Autonomy is granted per task, based on evidence, and can be withdrawn.
Gate 1
Supervised first
Every action the agent proposes is reviewed by a person. We learn where it is reliable and where it is not.
Gate 2
Evaluate
Behaviour is measured against real cases and agreed criteria, not judged by impression.
Gate 3
Earn autonomy
Specific, low-risk actions move to unattended execution once they consistently meet the bar. High-impact actions stay gated.
Engagement models
Ways to work with us
Strategy & discovery
A fixed-scope engagement to map workflows, prioritise use cases, and produce a build plan.
Focused first build
One AI employee, agent, or product slice, taken from design to supervised production.
Product delivery
A cross-functional team building an application or SaaS product over several months.
Ongoing engineering
Continuous monitoring, improvement, and new capabilities for systems already in production.
FAQ
Frequently asked questions
01How long does a typical project take?
Strategy work usually takes a few weeks. A focused first build is typically measured in weeks to a few months, depending on integrations. Larger product builds are planned in phases. We give a timeline after discovery rather than before.
02How do you price work?
Discovery and strategy are usually fixed-scope. Builds are scoped and priced per phase. Ongoing engineering is a monthly engagement. Details are agreed in a proposal after the first conversation.
03How involved does our team need to be?
We need a sponsor, access to the people who do the work today, and someone to review progress each sprint. Reviewing AI output during the supervised phase is essential.
04What do we receive at the end?
Working software, documentation, architecture notes, evaluation suites, and runbooks — in repositories and environments you control.
Next step
Bring us the process your team is tired of doing manually.
A strategy call is a working conversation about one workflow: what it involves today, what could be automated, and what a first build would look like.
What happens next
- 1Tell us about one workflow
- 2We review it before replying
- 3A working session, then a first-build proposal