Article
The barefoot shoemaker? How we run AI inside our own processes
As a company that helps organisations embed AI into their business processes, we hold ourselves to the same standard.
There is a fair question any AI consultancy should be able to answer: do you run your own business the way you advise your clients to run theirs?
For us the answer is a practical one. Our internal operations run on the same AI-native thinking we bring to clients — because if we ask an enterprise to trust AI in production, we should be trusting it first.
Where AI sits in our own work
- Proposal and document drafting from our own structured knowledge base.
- Research and market analysis as a first pass, always reviewed by a person.
- Delivery artefacts — architecture drafts, test datasets, documentation — generated and then curated.
- Internal knowledge retrieval, so what one team learned is available to the next engagement.
- Operational routine: summaries, follow-ups, reporting and status consolidation.
What running it ourselves teaches us
Using AI daily exposes the parts that are never in a vendor deck: where quality drifts, where a human review step is not optional, how quickly cost accumulates, and how much of the value comes from clean internal knowledge rather than from the model.
It also keeps us honest about limits. We recommend automation where we have seen it hold up, and human ownership where we have seen it needed.
Why it matters to a client
Advice that comes from operating experience is different from advice that comes from a methodology slide. When we describe what breaks on the way to production, it is because we have hit it in our own workflow first.
It is also a governance statement: the same guardrails we ask clients to adopt — access control, audit trails, human review on consequential output — apply to us.
We are not consultants who talk about AI. We work inside it, and that is what makes the guidance worth something.
This article is the English version of a post originally published in Hebrew on our LinkedIn page.