Fixed-Price AI Pilot: A Working Model on Your Data in 4 Weeks
Our AI pilot tests one task on your real data in four weeks, at a fixed price agreed in writing before we start. You get a working prototype, an accuracy score against how your team does the task today, and a written recommendation to build, change or stop.
Reply within a few hours on WhatsApp, fixed written quote within 48 hours. If AI is the wrong tool, we say so on the first call.
Most AI projects are bought before anyone checks they work
A vendor shows a demo on clean sample data. You sign for a full build. Three months later it meets your real invoices, your real customers and your real edge cases.
By then the budget is spent, and the first question is still open: does this work on our data, at our volume, well enough to pay for itself?
A pilot answers that question in four weeks, for a fixed price, before you commit to anything bigger.
Tasks that make good pilots
High volume, repetitive, and a clear right answer. These come up most often.
Document and invoice extraction
Invoices, delivery notes, KYC forms and prescriptions read, checked and pushed into your system instead of retyped.
02WhatsApp and email support assistant
Answers the questions your team answers every day, from your own documents and prices, and hands over to a person when unsure.
03Lead scoring
Ranks incoming enquiries by how likely they are to buy, so your sales team calls the right people first.
04Dropout and fee-recovery prediction
Flags students likely to drop out or miss fees early enough for the school to act.
05Search across your own documents
Staff ask a question in plain language and get the answer from your SOPs, contracts or policies, with the source attached.
Sorting and routing
Emails, tickets and forms classified and sent to the right person or queue without someone reading each one first.
The pilot is the first step of our AI and ML development and LLM and AI agent development work. If it works, the same team builds it into your website, app or WhatsApp.
When we will tell you not to run a pilot
A pilot costs money, so we turn some down on the first call. Usually for one of these reasons.
You do not have the data yet.
If there are no past examples to learn from or test against, the first job is collecting them. That is cheaper than a pilot.
The volume is too low.
A task that happens a few dozen times a month rarely pays back an AI build. Under about 100 cases a month, a better form or checklist usually wins.
A cheap tool already does it.
If a ₹2,000-a-month subscription handles the task, we will name it and tell you how to set it up.
Rules would work better.
When the right answer follows fixed rules, plain software is cheaper, faster and never guesses. We will build that instead.
What you get, and what we need from you
What you get at the end of week 4
Everything you need to decide with a number in front of you.
- A working prototype running on your real data
- An accuracy report on your own held-back examples, next to how your team does the task today
- An estimate of what a production build would cost and take, plus the monthly running cost
- A written go, change or stop recommendation
- All code, prompts and data in your accounts
If it is a stop: you stop there, with the reason in writing. There is no full-build pitch afterwards.
What we need from you
A pilot is only as good as the examples behind it.
- 50 to 500 real examples of the task: invoices, chats, records
- One person on your side who knows the task and can answer questions
- How the task is done today: time per case, and the error rate if you track it
- Two calls: a kickoff in week 1 and the results in week 4
Before you send anything: we sign an NDA.
How the pilot works, week by week
Four weeks, one task, with the result measured at the end.
- 01
Week 1: pick the task and check the data
We agree the one task, how success will be measured, and go through your examples. If the data cannot support it, we tell you here, before the build starts.
- 02
Week 2: a first working version
A prototype running on your real documents or records, built on your examples rather than sample data.
- 03
Week 3: improve it on the hard cases
We work through the cases it gets wrong, tune it, and set a confidence threshold so unclear cases go to a person.
- 04
Week 4: score it and decide
We test it on examples it has never seen, compare it with how your team does the task today, and walk you through the report on a call. Then you decide.
What the pilot costs, and what comes after
Each stage has its own fixed price in writing. You can stop after any of them.
AI pilotStart here
Fixed price, agreed in writing before we start.
One task, four weeks, your real data. The price covers the prototype, the accuracy report and the recommendation, and it only changes if the task does.
Fits: a first AI project, or getting internal sign-off with a real number.
Production build
Quoted after the pilot, as a fixed scope.
If the pilot says go, we quote the full build using what the pilot measured: scope, timeline and cost in writing. It runs inside the tools your team already uses.
Fits: a pilot that hit its target.
Running cost
Monthly, estimated in the pilot report.
Model usage and hosting, billed to accounts in your name. The report shows the expected monthly cost at your volume before you decide to build.
Fits: every AI system in production.
If the pilot shows AI is not worth it, you stop there. That is the point of running one.
Scope My AI Pilot on WhatsAppYour data stays safe
NDA before you send anything.
We sign an NDA, and a data processing agreement if you need one, before the first example reaches us.
We do not train public models on your data.
We use API tiers where the provider does not train on your inputs, or open models running on your own servers.
It can run on your servers.
If the data cannot leave your network, the pilot runs on your infrastructure with open models such as Llama or Mistral.
You own what the pilot produces.
Code, prompts and any trained model are yours, in your accounts, whether or not you go ahead.
Why run the pilot with us
A number, not a demo.
Accuracy measured on your own examples, against how your team does the task today.
We tell you when it is a no.
A stop after four weeks costs far less than finding out halfway through a full build.
We build the software around it.
If it works, the same team builds it into your website, mobile app or WhatsApp.
You own all of it.
Code, prompts, models and data, in your name from day one.
AI pilots for businesses in Delhi NCR
We are based in New Delhi and run AI pilots for businesses across Delhi, Noida, Gurugram and Ghaziabad, and remotely for clients elsewhere in India.
If you are in NCR, we will come and watch the task being done. An hour with the person who does it every day tells us more than a week of calls.
New Delhi, India · GMT+5:30
We meet clients in person
Delhi
Head office · AI and ML services
Noida
Document & support pilots
Gurugram
AI features for product teams
Ghaziabad
Operations & logistics pilots
Also running pilots for teams in
Frequently asked questions
Each pilot has a fixed price, agreed in writing before we start, based on the task and how much data is involved. It covers the four weeks, the working prototype, the accuracy report and the recommendation, and it only changes if the task changes. If the pilot works, the production build is quoted separately as a fixed scope.
Describe the task that eats the most hours
Tell us what the task is and roughly how many times a month it happens. We will reply with whether a pilot makes sense, and if it does not, what we would do instead.
- We reply on WhatsApp within a few hours.
- A 20-minute call about how your business runs.
- A fixed written quote within 48 hours. No obligation.
Phone
Reply within a few hours on WhatsApp, fixed written quote within 48 hours. NDA signed before you share anything.
Services