Fixed scope before talking about the price of building.
Custom software for a company that has a business rule of its own that no system serves. First AI Start closes the technical specification, with the fixed price of the build and the committed deadline inside it. If the build goes ahead, the cost of the specification is credited in full.
“I have tried four systems. None of them does the calculation the way we do it, and in the end everything goes back to the spreadsheet.”
And the quote you asked for to fix it came priced by the hour, or came with a round number and not one page saying what exactly would be built.
This suits few companies, on purpose.
Dedicated software only pays off when the rule is your own and the volume is real. In every other case AI Start says it does not fit, and says so in the first conversation.
- The business rule is yours and no SaaS serves it
- The volume already justifies a dedicated system, not a spreadsheet
- Someone decides the budget and owns the rules
- It has to live alongside the systems already running
- You want to test an idea whose rules are not defined yet
- You are looking for the cheapest hourly rate
- You want a ready-made SaaS with your logo on it
- You still do not know where AI fits in the operation
Two phases, and the first decides the second
Nobody commits to a fixed build price without having written down what they are going to build. That is why the specification comes first, and the number comes out of it.
Technical specification
What you receive
- Data model of the system
- Unambiguous written rules, with numeric examples of the edge cases
- Verifiable acceptance criteria, approved before building
- Integration design, tested against the real API
- Fixed price of the build, with a committed deadline
The credit
If the build goes ahead with AI Start, the cost of the specification is credited in full. It is a prerequisite of the build, so charging for both would mean charging twice for the same work.
Build
What you receive
- The system, with the rule that was specified
- Audit trail of everything that changes
- Access profiles by role
- Import of the data that already exists
- Technical and operational documentation
The warranty
60 days after acceptance, running against the criteria approved in the specification. It is not a conversation about what was expected: the criteria were written down before the build started.
What is not included here
Diagnosis of the operation. This track assumes you already know what the rule is and where the system is missing. If the question is still where AI fits in your company and what pays for itself first, the path is Growth Tech, which runs the return calculation before building anything.
The generic part is already built and running in production.
Legal entities, roles, row-level permissions, import and audit trail are not reinvented on every project. That foundation is Celeridade OS, which AI Start runs every day. What gets specified is only the rule that is yours, and that is why a price can be committed.
tables in production on Celeridade OS
row-level access policies, live
days of warranty after acceptance
integrations priced without being tested
Frequently asked questions
The price of the specification depends on the size of the business rule, and comes out in conversation. The price of the build comes out of the specification.
Ask on WhatsAppBecause it is the work. Modelling the data, writing the rule without ambiguity, surfacing the edge cases and testing the integration against the real API is what consumes engineering time. Anyone handing that over for free is charging by the hour during the build, and then the scope never closes. If the build goes ahead with AI Start, the cost of the specification is credited in full: you do not pay twice for the same work.
The specification is yours. It is technical and self-contained: data model, rules, acceptance criteria and integration design. You can build it with your internal team or with another supplier, and you can ask several of them to quote against the same document.
Fixed because the generic part is already solved. Legal entities, roles and row-level permissions, import, audit trail: that foundation runs in production on Celeridade OS and is not reinvented on every project. What gets specified is only the rule that is yours. And no integration enters the price without having been tested against the real API.
The specification takes 2 to 3 weeks. The build deadline comes out of the specification itself, committed alongside the fixed price. Before that, any deadline anyone gives you is a guess.
By the acceptance criteria written in the specification and approved before the build started. Each one is verifiable. The 60-day warranty after acceptance runs against those same criteria, not against a conversation about what was expected.
No, and it is worth not confusing the two. Custom engineering solves a business rule you already know and that no system serves. Growth Tech is for when you still do not know where AI fits in the operation, and need to find out with the return calculation up front.
Bring us the rule no system serves.
In the first conversation we decide whether dedicated software is worth it. If it is not, I will say so, and you save the specification.