Data
Is the data reachable, orderly, and current? Which source is treated as true?
Kudret Keskin — Forward-Deployed AI Architect
I move stuck AI pilots into production.
A good demo is not a working product. Ten questions show where your pilot is stuck, and the first thing worth fixing.
What the score measures
Four areas. One priority.
Pilots rarely stall on the model. Data, a process, ownership, or an old system path is usually what blocks them. The score rates those four apart.
Process
Which workflows repeat and can be measured? Who absorbs the exceptions?
Team
Who will own the outcome, and who will use it every day?
Systems
Can ERP, POS, e-document, and timekeeping connect, or is data still carried by hand?
After the score
Ten days from a pilot to a production plan.
The AI Production Plan Sprint works the first bottleneck the score names, inside your real systems. You leave with an owner, a schedule, and a cost — not a deck. You decide whether to continue.
Days 1–2
Discovery
We read the score again with your team and make an inventory of the systems and the data.
Days 3–4
Connection
We connect the most critical data source through a single read layer.
Days 5–7
A flow on real data
We run one workflow on your data, not on demo data.
Days 8–9
Measurement and risk
We write down what held, what did not, and what it costs.
Day 10
Production plan
Who, when, and on which system: a written plan and a decision meeting.
Measure your score first
The sprint is a conversation that happens after the score.
Selected work
Not demos. Products with real users.
From fintech to EV charging, from field operations to multi-branch companies: each one runs with real users and real systems.
A note from the founder
I started as a product designer. For about twenty years the work has been product, system design, and fintech. I am not an ML engineer. I do not train models. I make AI run inside a real operation. That is why I started Firmam.ai. It is in pilot with three companies, and about 100 people use it in the actual work. I will give you an honest score first. Sometimes the right answer is “not yet.”
Kudret
Kudret Keskin · Founder, Firmam.ai
How the work is sequenced
Three steps. Each one has to earn the next.
AI Readiness Score
Free
Ten questions, about ten minutes. A score across four areas, and the first step I would take.
Start the scoreAI Readiness Diagnosis
Settled in a conversation
I check the score inside your systems and write down the bottleneck and what it costs.
Plan a conversation10-day AI Production Plan Sprint
Settled in a conversation
We run one workflow on your real data and finish with a production plan that has an owner and a schedule.
Talk about the sprintQuestions that come up
Is the score actually free?
Yes. Your score appears in the browser right away. No sales call follows.
Are you an ML engineer?
No. I don’t build models. I connect the models you already have to your systems and processes so they hold up in production.
Who is this for?
Multi-branch SMEs, and fintech or enterprise product teams that are running an AI pilot.
What happens to my data?
The score uses only the answers you give. In a sprint, which data we can touch is agreed in writing, up front, with your approval.