Offer 01
One business outcome, automated end to end, in four to six weeks.
You name the outcome. We write the scope, build it into your existing stack, put evaluations and a spend cap around it, and support it for 30 days after it goes live. One price, agreed before we start, half of it up front.
- Price
- €15,000–30,000 fixed
- Duration
- 4–6 weeks
- trigger an email, a document, a schedule
- classify + enrich against your own records
- decide deterministic path
- draft the action
- your team approves
- write to your system
- log + evaluate every run
◆ marks a point where a person decides. Nothing passes it on its own.
What you get
Three outcomes we package
We build one of these per engagement — one outcome, written down and signed. Fixed price only survives fixed scope, so we would rather do one properly than three badly.
- Inbound triage
- Email, forms and chat classified, enriched from your own records, routed to the right owner and answered with a draft — with a person approving before anything reaches the customer.
- Document to system
- Quotes, invoices, contracts or claims read, extracted and pushed into your CRM or ERP, with an exception queue for anything the extraction is not confident about.
- Reporting that writes itself
- A recurring operational report assembled from your own systems, with the narrative written and the anomalies flagged, arriving before the meeting instead of during it.
The terms
The commercial terms
- Price
- €15,000–30,000 fixed. 50% on signature, 50% on delivery.
- Duration
- 4–6 weeks from the discovery workshop to production.
- Includes
- Discovery workshop · build · integration with your existing stack · evaluations, observability and an automated cost cap · 30 days of post-deployment support · handover documentation.
- Excludes
- Open-ended scope. One outcome per engagement. A second outcome is a second engagement, quoted the same way.
- Cross-border
- One clean price. Export of services from Thailand carries 0% Thai VAT, and you withhold no Thai tax. Invoiced as software development.
- After the 30 days
- Either you run it, or we do, on a monthly retainer. That choice is yours to make after it works, not before.
The sequence
How the six weeks go
The sequence matters: nothing is built before the outcome is written down, and nothing goes live before it is measured.
- 01
Discovery workshop
Half a day with the people who do the work today. We map the process as it actually runs, including the parts nobody documented, and pick the one outcome worth automating first.
- 02
Written scope, one page ◆ your call
The outcome, the systems it touches, the test that says it works, and what is explicitly out. This is the document the fixed price attaches to.
- 03
Build and integrate
Deterministic orchestration for the path, a language model for the judgement calls, and a semantic layer over your existing systems with field-level permissions.
- 04
Evaluations and cost cap
A fixed test set with a measured success rate, traces on every run, and a hard ceiling on token spend that stops the workflow rather than surprising you on an invoice.
- 05
You approve going live ◆ your call
You see the evaluation results and the approval gates before anything runs against real customers or real money.
- 06
30 days of support, then handover
We watch it in production, fix what the real world breaks, and hand over documentation written for whoever inherits it.
Fit
Who this is for
- A company of 10 to 200 people, in Europe or Thailand.
- A named process that costs your team hours every week and has a clear right answer.
- A budget owner who can sign without a six-month committee.
- Systems with an API, or a database we can read.
Who it is not for
- A team that already has this covered. If a developer with an AI assistant is shipping it for you, that is the cheaper answer and we will say so.
- "Let's explore AI" with no named outcome. We will decline until there is one.
- Anything where the process itself is undecided. Automating an argument makes it faster, not better.
Objections
The questions we get asked
The market rate for six weeks of this work is $60,000–90,000. Why are you a third of that?
Because we have no reference case for this offer yet, and the first engagements buy the case study rather than the margin. Our cost base is Thai, our delivery is agent-amplified, and we would rather have your logo than your last euro. That price moves once the references exist.
Independent research says 95% of enterprise AI pilots deliver no measurable impact. Why would this be different?
Those are the numbers we quote you, so it is a fair question. Pilots fail for reasons that are now well documented: no fixed scope, no measurement, a language model improvising steps that should be deterministic, and no owner after launch. Every one of those is addressed above, and you can hold us to the evaluation results before you let it run.
What happens if the fixed price turns out to be wrong for you?
That is our risk, and it is priced in. The protection for both of us is the one-page scope: one outcome, written down, signed. If you want a second outcome mid-build, that is a second engagement rather than an argument.
You are in Thailand and we are in the EU. What does that cost us in tax and admin?
One clean price and one invoice. Export of services from Thailand is zero-rated for Thai VAT, and there is no Thai withholding for you to handle. The work is invoiced as software development. Your own local VAT treatment is unchanged from any other non-EU supplier.
What do we actually have at the end?
A running automation in your own stack, its evaluation set and success rate, the traces and monitoring around it, an automated cost cap, and handover documentation written for the person who maintains it next.
How do you deliver a production build in six weeks?
By keeping the scope to one outcome, and by putting a fleet of agents on the repeatable engineering — the CRUD, the adapters, the test scaffolding, the migration dry-runs — with an engineer reviewing every change. It is the same system that runs this company's own sales, marketing and delivery. On the call we will open the repository and show you the dispatcher, the approval gates and an evaluation run.
Bring one process to the call.
Thirty minutes. Describe the process and who does it today. We will tell you whether it is a four-week build, a six-week build, or something you should not automate at all.