Built before the flow was understood
Engineering starts on a feature nobody mapped, so the product solves a step that was never the bottleneck.
Services/03 For any organisation
AI Product Development
We map the userflow, find the bottleneck that actually costs money, design the interface around it, engineer the agents underneath and launch the product. One team from first sketch to live system.
What we see
Teams get stuck between a demo that works and a product people use.
Most organisations can prototype something in a weekend. Turning that into a product with a flow people understand, wired into real systems, with evaluations, monitoring and someone on call, is a different sport. That gap is where budgets and momentum quietly die.
Demo works, production does not
The prototype impresses in a meeting, then falls over on real data, real volume and real edge cases.
No evals, no confidence
Nobody can say how good it actually is. Every change is a leap of faith instead of a measured decision.
No owner after launch
Nobody watches cost, drift or failure modes. The first incident becomes the last release.
What you get
Concrete deliverables.
Userflows mapped end to end
Every step, handoff and exception written down, with the people and systems involved. The map everyone works from.
Bottleneck analysis
Where time, money and errors actually pile up, quantified. We build against those numbers, not against opinions.
AI-native design mockups
Clickable interface designs for the new flow, so the team sees and judges the product before we build it.
Agentic engineering
The agents, tools and integrations underneath. Versioned, observable and safely rolled back inside your existing stack.
Evals, monitoring and guardrails
Golden datasets, regression tests, tracing, cost dashboards and human-in-the-loop where the risk sits.
Launch and handover
Rollout with your users, documentation, runbooks and training so your team owns the product after go-live.
How we deliver
From ambition to agents in production.
01
Session
Kickoff
Half a day to walk the userflow, agree success metrics and pick the shortest path to a working v1.
02
Scan
Discovery
Two to four weeks of flow mapping, bottleneck analysis and design mockups you can react to.
03
Build
Delivery
Six to twelve weeks to design, engineer and launch the product, including evals and monitoring.
04
Squad
Ongoing
Monthly retainer for new features, tuning, new flows or a second product in parallel.
Timeline
6 to 12 weeks per product
Investment
Scoped per project, typical builds €25k to €120k
Fit
Who this is for. And who it is not.
Fits you if
Not the fit if
- 01You have a process or product idea, not yet a working system
- 02Off-the-shelf SaaS does not fit the shape of your problem
- 03You want design and engineering in one team, not two vendors
- 04There are throughput, accuracy or cost numbers you want to move
- 01You want a generic chatbot wrapped around your FAQ
- 02No appetite for evaluation, monitoring or maintenance
- 03You need the full company transformation - look at AI Program
Agentic systems we build.
Production-grade systems we design, build and deploy for commercial teams. Each one tailored to a single high-value workflow.
Your directAI partners.
01Derk Disselhoff
Founder & Director
02Elko Lemiso
AI Engineer