Choose end-to-end delivery, hands-on support for your team, or focused expert advice. Each engagement starts with a clear problem, an accountable owner and an agreed outcome.
Discuss your projectFrom discovery and architecture to a working system, release and handover.
End-to-end delivery. Gunnar owns the engagement.
Hands-on engineering, architecture and delivery support inside your team.
Your delivery lead keeps accountability.
Clear recommendations on AI opportunities, knowledge quality and technical direction.
A named decision-maker receives the findings.
Most work mixes the three. An architecture assignment often involves advice and a period of embedded support, so we agree the balance rather than forcing the work into one mode.
For each one: the situation it helps with, what is delivered, and how we judge whether it worked.
AI is expected somewhere, but nobody has agreed where it would pay.
A mapped process, a shortlist of candidate use cases and a recommended first move.
You can defend the choice of first use case and the measures attached to it.
The team is building, but the requirement keeps moving underneath them.
Written intent, acceptance criteria and the open questions that still need answers.
Developers and stakeholders describe done in the same words.
Answers exist somewhere in documents, systems and people's heads.
Approved sources, retrieval with citations, permissions and an evaluation set.
Answers cite a source someone owns, and the evaluation set improves over time.
A knowledge or search tool is live and people do not trust what it returns.
A source map with owners, a record of gaps and risks, and a prioritised plan.
The next three pieces of work are agreed, with owners against each.
A repetitive process spans several systems and still needs judgment in places.
Connected tools, explicit operating limits, approval points and observability.
Tasks complete without supervision inside their limits, and escalate outside them.
The capability has to reach real users as a product, not a demo.
A working application, tests, documentation and a release you can operate.
People use it for the task it was built for, and your team can change it.
Your developers use AI tools, with uneven results and no shared review standard.
Task specification patterns, review gates and working agreements from our ADLC.
Agent-assisted changes arrive small, reviewed and traceable to an intent.
Decisions are being deferred, and each delay narrows the options.
Boundaries, interfaces and recorded architecture decisions with their trade-offs.
A new person can read why the system is shaped this way and argue with it.
Something works but feels fragile, and you want a second opinion before scaling it.
Findings, risks ranked by consequence, and remediation options with effort.
You know which risks you are accepting on purpose.
You have the plan and need capable hands inside the team to deliver it.
Implementation, evaluations, integration and review alongside your engineers.
The work continues after we leave, without a knowledge gap.
Data sensitivity, cost or deployment constraints rule out a single cloud model.
Model placement by task, the controls around it, and measured quality and cost.
Placement is a documented decision, not an accident of what was easiest.
Pilots multiply and nobody owns evaluation, approval or the running cost.
Named responsibilities, approval limits, evaluation practice and cost visibility.
Everyone can say who approves what, and on what evidence.
Small slices, written intent and evidence a person has reviewed. Delivery follows our Agentic Development Life Cycle.
Collaboration, adoption and organisational change matter on a delivery programme. A focused architecture review does not carry a change-management workstream, so we scope those parts to the engagement rather than promising them everywhere.
Advisory engagements are scoped individually; not every one produces the same artefacts.
We talk about your current process, constraints and desired outcome. Where discovery is needed, we agree its scope, deliverables and duration before work begins.
Discuss your project