Each step ends with something written. Each step can be the last one.
We don’t sell a transformation. We sell one use case at a time, with a written result at every step and a fixed scope before any build starts. You can stop after any step and keep everything produced so far.
Assess
Two days on site, then analysis.
We look at your data, systems and daily work, talk to the people doing it, and rank the possible use cases by value, feasibility and effort, with reasons.
Result
A written roadmap you can act on, with or without us.
Plan
For the one use case you select.
We define who uses the system, which data and systems connect to it, and what has to demonstrably work at the end. Architecture, responsibilities, acceptance criteria and rollout are agreed together. The plan is yours whether or not you go ahead.
Result
An agreed implementation plan with stages, responsibilities and a fixed scope.
Build
Fixed scope, based on the plan.
We develop against real data and real workflows and involve the eventual users early. You get a running version to try at regular intervals. Before we start, it’s settled who owns the source code and the infrastructure, and which data external models are allowed to see.
Result
A working system in active testing, with the first users onboarded.
Run
With us, or inside your company.
We monitor, test and update the system and watch the data flows and outputs. When something looks wrong, a person who knows the system is reachable. Operation can stay with us or move to a trained owner on your side.
Result
A system in reliable operation with named responsibility and agreed response times.
Hand over
On the live system, with real cases.
Users learn on their own actual work, not on slides. One or two owners additionally learn to maintain content, judge errors and administer the system.
Result
Users and owners who can run the system without us.
Next
When the first one is reliably running.
We pick up the next item from the roadmap: another automation, an agent, a new workflow.
Result
The next use case in production, with its own scope and acceptance.
What “production” means to us.
Real data from week one.
No synthetic demo data. If the real data is messy, that’s the first finding, not an excuse.
Every output traceable.
Answers cite sources. Decisions are logged. Model versions are recorded.
Failure is designed, not discovered.
Fallbacks, escalation to a person, and monitoring are in the plan before the model is.
Languages are not an afterthought.
Hebrew and Arabic are tested as first-class, including mixed-language input and names.
You can leave.
Code, infrastructure and documentation are yours. Nothing is locked behind our accounts.
We say no early.
If a use case won’t work, you hear it in planning, in writing, before it costs you a build.