AI goes into your operation where it pays for itself.
Three months of fixed scope. Month 1 maps the operation and prioritises use cases with the projected return of each one. Months 2 and 3 build one workstream per month, with written acceptance criteria and before-and-after measurement.
“Everyone talks about AI and I still have no idea where it fits in my company. I tried it, it did not stick, and nobody explains why.”
The tool worked. What was missing was the written process, the organised data, and someone on the team responsible for keeping it standing after the vendor left.
A workstream that does not pay for itself does not get built.
Month 1 projects the return of every use case before a single line is written. If one of them does not pay for itself, that becomes a document and the workstream leaves the queue.
“Should the month 1 diagnosis conclude that the projected return of a workstream does not exceed the cost of building and maintaining it, AI START will formalise that conclusion in writing and the workstream will be replaced by the next one on the priority list, with no change to the price.”
Standard contract v1 · clause 4.5
AI Start works as your external technology department.
Map the process, redesign it on the method, and implement whatever survived the return calculation. That is what the contract says, and that is what happens across the 3 months.
- A course or tool training
- Consulting that hands over slides and leaves
- A chatbot installed on top of a broken process
- Custom software (that is the other page)
- Mapping of the real operation, by observation and interview
- Bottleneck matrix ranked by impact and effort
- Two workstreams built and integrated into what you already use
- A trained team, with recurring use as the criterion
Three months, one workstream per month
A workstream is a prioritised use case turned into a delivery: a system, an automation, an agent or an AI-assisted routine. The pace is one per month from month 2 onwards, and it is in the contract.
Diagnosis and design
- Mapping of the current process, by observation and interview across the areas
- Bottleneck matrix, ranked by impact and effort
- Design of the new process structure
- Map of AI use cases, with projected return per workstream
- Roadmap delivered within 3 business days of the kickoff
- Acceptance criteria for the following months, approved in writing
First workstream built
- Construction of priority workstream no. 1 defined in month 1
- Integrations with the tools the company already uses
- Guardrails and human review on every AI output
- Organisational memory base of the operation established
Second workstream, team and measurement
- Construction of priority workstream no. 2
- Training with recurring use as the completion criterion
- Before-and-after measurement, on the indicators agreed in month 1
- Documentation, knowledge transfer and closing report
What is not included in the 3 months
A second workstream inside the same month, and any service outside the scope of the clauses above. It requires a written amendment, with its own price and deadline. Fixed scope is only fixed if that line exists.
Software with a business rule of your own is not included here either. That is custom engineering, with its own scope, price and warranty.
How we measure that it worked
At the end of month 1 both parties approve the acceptance criteria in writing, and each one has to be verifiable by an objective indicator or an observable behaviour. At the end, the assessment compares what was delivered exclusively against those criteria.
AI Start does not promise revenue figures or cost reduction figures. The obligation is one of result as to the deliverables, and that is written down too.
Not us talking. It is someone who lived through Growth Tech.
A services client that operated with no documented process. Diagnosis, redesign and AI implemented inside the project. The person telling you what changed is the client.
And when the problem is that no system does it the way your company does?
Then it is not diagnosis, it is engineering. We close the technical specification before talking about the price of building, and the cost of the specification is credited in full if you go ahead.
See custom engineeringFrequently asked questions
The price is fixed for the 3 months and does not vary by hours worked. How much it is, we discuss directly.
Ask on WhatsAppA 1-hour kickoff to define the essential workflow, the point of contact and the access. After that, the month 1 mapping is done by observation and interview inside the areas involved, at the pace of the operation. The contract also covers the opposite: a delay in feedback, in naming the point of contact or in granting access shifts that month's schedule, day for day.
It does not get built. It is in clause 4.5: if the projected return of a workstream does not exceed the cost of building and maintaining it, AI Start formalises that conclusion in writing and the workstream is replaced by the next one on the priority list, with no change to the price. You do not pay to implement something the diagnosis already showed will not pay for itself.
No, and it is written in the contract. AI Start's obligation is one of result as to the deliverables and one of means as to the business outcome. No revenue figure, cost reduction figure or commercial performance figure is promised. Anyone promising a revenue number before seeing your operation is selling something else.
At the end of month 1 both parties define and approve in writing the acceptance criteria for the following workstreams, and each criterion has to be verifiable by an objective indicator or an observable behaviour. At the end of the project the assessment compares what was delivered exclusively against those criteria. It is not a conversation about whether it turned out well.
No. The completion criterion for training is recurring use of what was implemented, not attendance. If the team is not using it, it is not complete.
The project ends. It is fixed scope, lasting 3 months, and it does not turn into an automatic subscription. You keep the closing report: what changed in the operation, what the data revealed, the backlog that was not executed and the recommendation for next steps.
Then Growth Tech is not the path, and we say so in the first conversation. Software with a business rule of your own has a different scope, price and warranty, and it is on the custom engineering page.
The method does not depend on the industry. It depends on there being a process running to be mapped and someone with the authority to decide what changes. An operation that is still being invented has nothing to diagnose.
Where does AI fit in your operation?
A 15-minute conversation to find out whether a workstream pays for itself. If there is none, I will say so on the spot.