Approach
Start small, then build it properly.
Every engagement runs in short, fixed-price steps, and every step delivers something that works on your own data. You always know what you are getting, what it costs and when it arrives.
How an engagement runs
A working conversation
Twenty minutes on what you run today, where the hours go, and what a good result is worth. You leave with a clear view of where a layer pays for itself fastest.
A fixed-price scoping step
We read the real system: the code, the data, the awkward exceptions. You receive a written plan with milestones, costs, dates and risks, ready to act on.
One layer, built in small tested pieces
Each milestone runs on a staging copy against your own data. You see it working before the next one begins.
Reads first, acts when you say
A new layer begins by reading and suggesting. Permissions to act are added one at a time, on your instruction, with a log of everything it does.
A complete handover
Code, prompts, configuration and documentation are delivered into your repository and your accounts. Ongoing monitoring, tuning and new layers are available on a monthly retainer.
Ground rules
Six principles behind every build.
Your systems stay the source of truth
The layer reads from the tools you already trust and writes back only where you allow it. There is no second copy of your business to keep in step.
Narrow beats clever
One layer, one job, done reliably. A system that answers one kind of question correctly is worth more than one that attempts everything.
People handle the exceptions
Anything outside the layer's remit goes to a person with the facts attached. We design the handover before we design the automation.
Costs you can see
You know what each answer, ticket or order costs to process, in cents, before you run it at volume.
Proven on staging first
Every change is demonstrated on a staging copy with your data. Production is always the last step.
You own all of it
The code, the prompts, the evaluation sets and the logs are yours. Your team has everything it needs to run and extend the layer.
Questions
What clients ask first.
Do we have to change the software we use?
No. The layer works with what you already run: your shop platform, help desk, database and file stores. The access it needs, such as an API key or a read-only database user, is listed in the scoping plan.
What does a first project cost?
The scoping step is a fixed price, agreed before it begins. The build that follows is priced per milestone, also fixed, so the total is known up front.
Where does our data go?
By default, to a leading cloud model through its business API, under terms that exclude it from model training. Where data must stay on your premises, we install the layer on your own hardware with no outside connection. Details are on the Trust page.
How do you keep the AI accurate?
Each layer ships with a test set of real examples from your business, so accuracy is a measured number. Layers begin in draft mode, cases outside their remit go to your staff, and every action is logged.
How long before we see something working?
The first milestone runs on a staging copy on a date fixed in the scoping plan, and each milestone is delivered and signed off before the next begins.
Can you work with an old or unusual system?
Yes. Long-lived platforms and large databases are our home ground, and several of our case studies began with exactly that.
Do you support the layer after launch?
Yes. Monitoring, tuning and extension are available on a monthly retainer, and new layers can be added to the same foundation.
Next step
Start with the scoping step.
Fixed in price, short, and it produces a written plan with milestones, costs and dates.