What is a Design Sprint, and how can it help your business?

Spoiler: it has nothing to do with pixels. And it's not another workshop where you sit back and wait for coffee.

The problem with "let's do something with AI"

If you walk into most mid-sized companies today, you'll hear some version of the same sentence: "We need to do something with AI."

What follows is usually one of two things. Either nothing, because nobody knows where to start. Or something expensive: a consulting engagement that ends with a 60-page PowerPoint deck and no working software. Or a batch of ChatGPT or Copilot licences that land in inboxes and quietly gather dust, because nobody defined which problem they were supposed to solve.

According to research cited by kipreneur.de and Computer Weekly, up to 85% of AI projects fail. Not because the technology doesn't work. Because the wrong problem gets solved, users aren't involved, and assumptions about data go unchecked until it's too late.

That's exactly the gap the AI Design Sprint is built to close.

What actually is a Design Sprint?

A Design Sprint is a structured, time-boxed process for getting from a vague idea to a validated, tested solution, fast. The original format was developed by Jake Knapp at Google Ventures and runs over five days. It's a genuinely useful tool, especially for teams building new digital products.

Here's the honest catch: most business teams don't have five days to spare. And more importantly, the classic Sprint was built for greenfield product development: starting from zero and imagining what something new could look like.

Most companies aren't working on a greenfield. They have existing processes. Procurement workflows. Quality controls. Certification cycles. Onboarding processes. The challenge isn't inventing something new. It's figuring out where AI can fit meaningfully into what's already there without breaking it.

That's a fundamentally different problem. And it needs a different tool.

The AI Design Sprint: built for processes, not pixels

The AI Design Sprint is a modified, faster, and more focused version of the classic format. It was developed specifically for teams that want to apply AI to their existing workflows without any prior technical knowledge and without an army of external developers.

It covers five phases, and the whole thing, from first idea to working prototype, takes about a week.

Here's how it works:

Opportunity Mapping (half a day) Management and senior decision-makers use AI capability cards, physical cards that translate AI concepts into plain business language, to work out which parts of the business are worth exploring. No technical knowledge required. Just an honest look at your own operations.

Framing Workshop (half a day) The specialist team responsible for the chosen area maps out its actual workflow, step by step. No glossing over the messy parts. Then AI categories get assigned: what type of AI could realistically help at which point in the process? The highest-priority workflow moves forward.

Concept Development (2 days) Now it gets concrete. The team works through the prioritised workflow in detail: what's the exact input at each step? What does the AI do? What does it output? Where does a human need to step in? This isn't a high-level overview, it's a full specification. Without it, you can't build anything real.

TechCheck (half a day) Before anyone writes a line of code, we check whether the idea actually holds up. A team member, their IT contact, and the facilitator sit down and answer one simple question: is the data we need available, and in what format? For the prototype we work with an offline copy: no API connections, no authentication layers, no GDPR architecture. That comes later. Here we just need to know if the idea is viable.

Optional: Proof of Concept (2 days) The team builds the prototype itself. Using tools like Claude Code, and with support from the facilitator, the team codes a working piece of software using their own data, for their own workflow. No external agency. No dependency on IT. The prototype belongs to the company. The knowledge stays in the company.

The moment when everything clicks

In almost every sprint I run, there's a moment, usually somewhere in the framing workshop, where the energy in the room shifts.

It happens when someone on the team realises AI doesn't need a chat interface. That it doesn't have to be a box you type questions into. That it can run quietly in the background, processing data, spotting anomalies, and making predictions, without anyone talking to it.

That moment is worth more than any presentation I could give. Because it changes the whole dynamic. AI stops being "that ChatGPT thing" and becomes a real design material you can weave into how the company actually works.

A real example: six-figure savings per product

One of the most memorable sprints I've run was with a team building safety components for industrial applications.

The challenge: a long, expensive certification cycle. Every time a product went through electrical testing, there was no way to predict which component was likely to fail. You just ran the tests and found out.

During the sprint, the team built an AI solution that analysed historical test data to predict component failures before testing even started. By predicting the most likely failure points in advance, the team could prioritise and cut out major chunks of the certification process.

The result: a shorter certification cycle and six-figure savings per product.

That wasn't a technological breakthrough. It was a process insight, made possible by taking a workflow, mapping it carefully, and asking the right question: where does AI actually add value here?

Why throwing licences at the problem doesn't work

Here's the most common, and most expensive, mistake I see in business teams before they find a better approach: buy the software first, never define the use case.

A stack of Copilot licences arrives. An email goes out. Staff are told to "use AI." And then... productivity doesn't magically go up, because nobody defined which problem the tool was supposed to solve or where it fits into which workflow.

AI tools are everywhere. But they sit unused, or get treated as a slightly smarter search engine, because the integration work got skipped. The question isn't "do we have AI?" The question is "where in our process can AI actually make a difference, and what can it replace or support?"

That's what the sprint answers.

What happens if the idea doesn't work?

This is actually one of the most valuable things a sprint can do: fail fast and cheap.

If a prototype gets built and the idea hits a wall, the data isn't good enough, the workflow assumption was wrong, the use case isn't really feasible, you find that out after a few days and for a fraction of the budget a full infrastructure investment would need.

No Kubernetes clusters. No AI infrastructure spend. No six-month build cycle that ends in an awkward conversation.

Among other things, the sprint is a cheap, fast way to check whether an AI idea actually works with AI. That's not a consolation prize. That's a core function.

And if it does work? Then you've got a working prototype, a validated concept, a concrete implementation plan, and a team that built something themselves and actually understands it. Those internal champions are harder to find and more valuable than any outside consultant.

After the sprint: what's next?

Teams that go through an AI Design Sprint don't just walk away with a prototype. They leave with a method they can apply themselves to the next use case, and the one after that.

The sprint trains teams to think in workflows, ask the right questions about AI integration, and use modern AI tools productively. That knowledge stays in the company, not with the consultant who flies home at the end of the week.

For the prototype itself, there are two paths. If it's viable, a lean team of outside AI specialists and in-house developers can turn it into a production-ready MVP. If it still needs validating, the sprint has already given you everything you need to convince management, including a demo that makes the case in ten minutes when a slide deck would take months.

Is your team ready?

If you're reading this and thinking "sounds interesting, but we're not quite there yet", let's talk. If you're genuinely not ready, I'll tell you straight. No pitch, no pressure.

My job is to help you find an AI idea worth backing, present it to decision-makers, and, if the conditions are right, put a working prototype in their hands.

But before we do that, there's a free way in.

I've put together a free 20-minute course that shows you how to find a real AI use case for your own workflows, and how to work out which steps actually benefit from a GenAI chatbot and which don't. Most teams skip this step. It's the most important one.

Get the free course

Back to overview