Find out if it works. In 2 to 4 weeks.
- Sprint
- Feature Sprint
- Price
- $10,000 to $25,000
- Time
- 2 to 4 weeks
- Fix window
- 30 days
Example builds.
- Can a model read our invoices well enough to skip manual entry?
- Can we answer customer questions from our own documents?
- Will this integration hold up with real volumes?
- The full build, once the answer is yes
What is included.
- Included: A written question and the bar for "yes", agreed before we start
- Included: A working prototype on a sample of your real data
- Included: A test set, and the results against your bar
- Included: A demo your team can try
- Included: A written result, a cost estimate to run it, and a quote for the full build
- Included: All code and accounts in your name
How the AI part works.
- What the model does
- The task you are not sure about, such as reading, classifying, matching, drafting or answering. We try more than one approach and keep the one that does best on your data.
- What it must not do
- Look good only in the demo. Every result is measured on a test set the prototype did not see while we built it.
- How we test it
- Against the bar you set at the start, for example how many documents it must read correctly. You see the results batch by batch, including the misses.
- What it costs to run
- We measure the cost per task, so the result tells you what the full system would cost to run, not only whether it works.
What this looks like
You have an idea, and one part of it nobody can promise: whether a model can read your documents well enough, whether your data is good enough, whether an integration will hold. Building the whole product to find out is expensive.
A proof of concept answers that one question first. We agree on the question and on what “yes” means, build a prototype on your real data, and test it.
How it goes
Week one: the question and the bar in writing, a sample of your data, a first prototype. First weekly demo.
The following weeks: we improve the prototype and test it batch by batch against the bar. You see every result, including the misses.
The last week: the written result, the cost to run it, and, if the answer is yes, a fixed quote for the full build.
What you get at the end
A clear answer on your own data, a prototype your team has tried, and the numbers to decide what to build next.
Questions we get on the first call.
What if the answer is no?
Then you have spent a small, fixed amount instead of a full build, and you get a written result that says why and what would need to change. That is a good result too.
Is the prototype thrown away?
Not usually. The code is yours, and when the answer is yes, the full build starts from the parts that worked.
Does it have to involve AI?
No. A proof of concept can test anything risky, such as an integration, a data migration or a performance limit. AI questions are the most common ones.
Book a call, or send the brief. A written price in three working days.
No NDA needed for the call. We sign one within 24 hours if you want it.