You have a list of AI ideas. Which one is worth building?
The AI Kickstart is a three-week, fixed-price engagement for companies of 30 to 250 people. We question every idea until one or two survive, I build working prototypes of those, and you see what it would really take to run them.
By David Canepa-Talamas, PhD. At Microsoft I lead Data & AI work with some of the largest companies in the world. The Kickstart is my independent practice.
Book a 30-minute callNo pitch. If I don't see a clear opportunity, I'll tell you.
| Idea | Number it moves | Verdict |
|---|---|---|
| Draft tender responses from our past bids | Hours per bid | Build |
| Chatbot on the website | Nobody could name one | Discard |
| Answer technical questions from our own standards | Engineer hours spent searching | Build |
| Predict machine failures | Unplanned downtime | Park: no data yet |
| AI meeting notes for everyone | No owner, no target | Discard |
What happens in the Kickstart
We question your ideas
Your team brings every AI idea it has. For each one I ask the same things: what is the business impact, which number does it move, and what does it look like when it's solved? If the answers are weak, we discard it and move on.
I build the one or two that survive
Not a slide deck. Working prototypes on your own material, good enough to put in front of your team, your sales people or a customer and get real feedback.
You find out what's missing
Building one real use case shows the foundations you lack: the data, the access, the security, the rules for using AI. Fixing them for one use case opens the door to the next.
It used to take months to learn whether an AI project was even viable. Now a prototype can answer that in days, so the meetings can be about something real.
Who this is for
A good fit if
- You lead a company of 30 to 250 people.
- Your team has AI ideas, or already uses AI tools, but nothing has reached day-to-day work.
- You want to know the return before you commit a budget.
- You would rather see something working than read a strategy.
Not a fit if
- You only want a chatbot on your website.
- You want a roadmap but aren't ready to build anything.
- You are a large enterprise with its own AI team.
Engineering-led firms are where I'm most at home, but the method is the same in any industry.
Does this sound familiar?
- "We know we should be doing something with AI, but where do we start?"
- It is the question I hear most. A good first project is small, has a number attached and matters to someone who will own it.
- "Everyone is using it, and nobody agreed how."
- People paste company information into whichever tool they like. A one-page policy and a short list of approved tools settle most of it.
- "We ran a pilot and it went nowhere."
- A demo that impresses a meeting is not the same as something people use. That is why the questions about numbers and ownership come first.
- "Our people need training, but courses don't stick."
- People learn these tools by building something they care about, with someone next to them to get them unstuck.
- "I can't tell if the return is there."
- Then that is the first thing to find out, before any money goes into building.
What you have at the end
- A ranked list of your use cases, with the reason each one made the cut or didn't.
- One or two working prototypes that you own and can show to staff or customers.
- A short foundations report: the data, access and security the prototypes needed that you don't have yet.
- A one-page AI usage policy and a list of approved tools.
- A clear next step: scale it, change it, or stop.
How it works
-
Before we start
A free 30-minute call
I learn about your business and tell you whether I see a clear opportunity. If I don't, I'll say so.
-
Week 1
The use-case workshop
Two to three hours with your leadership and the people who do the work. Every idea goes on the scorecard, and one or two come off it as the ones to build.
-
Week 2
Build
I build the prototypes on your material. You see progress as it happens and can steer.
-
Week 3
Demo and decide
A live demo for your team, the foundations report and the usage policy. Then we decide together what happens next.
Optional: a hackathon day for your team
A short session on the fundamentals, then everyone works on their own problem with the tools while I help them get unstuck. It is the best way I know to learn AI, and people leave with something they built themselves.
Why me
I do this with some of the biggest companies in the world
I'm a Strategic Partnership Technical Lead for Data & AI at Microsoft. For more than five years I've worked with very large companies across manufacturing, pharma, consumer goods and finance on exactly this: where to start with AI, which problems are worth solving, and what it takes to get an idea into production.
I've sat through their pain points and seen which projects pay back and which stall. The questions that tell them apart are the same at any size. The Kickstart is that process, sized for a company of 30 to 250 people.
An engineer first
My PhD is in mechanical engineering, in metrology, the science of measurement. I like a number attached to a claim.
I build, too
Outside my day job I build working software with AI: a coaching app that is live in the app stores, a sales-analysis system a retailer uses to make decisions, an administration prototype for a construction company, and an open-source engineering simulation tool.
Business and technology in the same conversation
I've spent years in rooms with executives turning AI into business terms. I can also sit down and build the thing.
What's included
AI Kickstart
One fixed price, agreed on our call before anything starts.
- The use-case workshop
- One or two working prototypes
- The foundations report
- A one-page AI usage policy
- A live demo for your team
- Done in three weeks
Hackathon day
An optional add-on, priced on our call.
- A fundamentals session for the whole group
- Everyone builds on their own problem
- I'm there all day to unblock people
- Remote, or on site with travel at cost
If a prototype is worth scaling, I quote that build separately, also at a fixed price. You don't need a retainer to start.
The first three clients pay half
I'm building a public track record for this work. The first three companies get the Kickstart at half the price. In return I ask to write the project up as a case study, anonymised if you prefer.
Questions people ask
Which AI tools do you use?
Whichever fits. Claude, ChatGPT, Gemini, Copilot, open-source models. I have no reason to push one vendor, and if you already pay for one we start there.
Do we need technical staff?
Not for the Kickstart. If a prototype goes on to become a real system, we'll talk honestly about who will look after it.
What if none of our ideas is worth building?
Then you have found that out in a week instead of after a six-month project, and I'll tell you so. If I can't deliver what we agreed, you don't pay for what wasn't delivered.
Is our information safe?
I work under a confidentiality agreement, use only the material you give me for the prototypes, and build in accounts you control wherever possible.
Remote or on site?
Remote by default. I'm based in Barcelona and run the Kickstart remotely for companies in the UK and the US. On-site workshops and hackathon days are possible, with travel at cost.
Who owns the prototypes?
You do.
Find out which idea is worth building
Book a free 30-minute call. I'll ask about your business and the ideas on your list, and tell you honestly whether I see an opportunity.