How we work with you
Four stages. You can stop after any of them — many clients do the first one and implement it themselves. That is a successful outcome.
Indicative pricing only · your quote is what counts The ranges below depend on scope, number of departments and language requirements. This is not a price list, not a quote, and not a commitment of any kind — the formal quote we issue is what governs.
Discover — AI Discovery Sprint
We come in, watch how your work actually flows, measure it, and find which tasks are worth handing to AI. You probably don't know how many hours a year your team spends on repetitive work — almost nobody does. That number is the first thing we give you.
Deliverables (nine, all of them)
- Current workflow map, with measured time per step
- AI readiness score, five dimensions
- Pain point inventory, each converted to annualised hours and cost
- Top 3 use cases, with an impact × effort × risk matrix
- A working prototype, run on your own data
- Data and privacy risk assessment
- A 90-day roadmap in three parts, each with something you can sign off
- Implementation recommendation and quote
What the three days actually look like
Not "we'll come do some research." Every hour is scheduled — who is in the room, which questions get asked.
- Day 1 09:00 kickoff (everyone, owner included) · 09:30 owner interview (one-on-one, always) · 13:00 manager interviews · 17:00 daily convergence
- Day 2 09:00 frontline interviews · 10:30 shadowing — we watch them actually do it · 13:00 quantify annual hours · 16:00 readiness scoring
- Day 3 09:00 use-case ranking · 10:30 build the prototype · 13:00 write the deliverables · 15:30 handover meeting
The owner interview is 12 fixed questions, written out word for word. One of them: "Is there something you know is inefficient, but have never had time to fix?" What comes out of that one is usually the first thing we build.
Want to see a finished one first?
We keep a fully worked Discovery example — a fictional Chiang Mai hotel, all nine deliverables filled in, not a single "TBD." You can see what you'll get before you sign anything.
When you should not buy this
- You already know exactly what you want built — go straight to implementation, don't pay for a stage you don't need
- Your organisation is mid-restructure — the processes themselves are moving, anything we measure expires next month
- No decision maker can sit down with us — we will ask you to postpone rather than proceed
Enable — role-based AI enablement
We don't teach "how to use ChatGPT". We teach your staff to put AI inside the work they already do every day. They finish with a configured Skill pack, not a stack of slides.
When you should not buy this
- You haven't done a Discovery yet — it degrades into generic AI training, and government, universities and large vendors are teaching that for free. We cannot compete with free, and we shouldn't try.
- Your team is actively resistant — training doesn't fix resistance; that's a management conversation
Implement — skills, agents, integration
Three levels: a reusable Skill pack, an agent wired into one business process, or integration with your existing systems. We pick the level from your actual readiness, not from your budget.
Starter Kit — four components, not a blank page
Most of what clients need falls into four buckets. These are already designed. Our job is wiring them to your process and your data — not inventing from scratch.
- Voice — speech to text for meetings, interviews, field notes. Runs locally; your audio never leaves your machine.
- Documents — extract, translate, summarize, compare, rewrite, search. Every claim clicks back to the exact passage it came from.
- Creative — Thai/English marketing copy. Length counted by each platform's real rules, not a naive character count.
- Assistant — meeting → minutes → tasks → human approval → release. That approval step is not optional.
Two things we do by default
- Every AI output that reaches your customers is reviewed by a person first. Full automation is something you switch on in writing — one wrong automated reply costs far more than the minutes it saves.
- We measure the baseline before we start. Without a before-number, neither of us can prove afterwards that it worked.
When you should not buy this
- Your readiness score came back too low — we will put it in writing that you shouldn't do this yet
Operate — managed AI
Going live is not the end. Models change, prices change, APIs change, processes change, staff leave. Without this stage, the value of the first implementation decays within six months.
When you should not buy this
- You have your own IT team and only took the lightest level — they can maintain it; a quarterly review is enough
| Service | Range (THB) |
|---|---|
| AI Readiness Call30–60 minutes | Free |
| Discovery Sprint3 days | 25,000 – 45,000 |
| Discovery Sprint5 days | 45,000 – 80,000 |
| AI Enablement | 20,000 – 60,000 |
| Implementation · Skill pack | 20,000 – 60,000 |
| Implementation · Agent / workflow | 60,000 – 150,000 |
| Implementation · integration | 150,000 – 400,000+ |
| Managed AIper month | 3,000 – 30,000+ |
Model usage is billed separately and shown transparently through our gateway — we do not mark up tokens inside the service fee. Travel is charged at cost. Early reference clients can negotiate a discount in exchange for a publishable case study.
Not sure it's worth it? Start with a 40-minute call. We ask which work repeats most; you decide whether to go further — and if we think you shouldn't adopt AI yet, we'll say so.
See the letter we send →