AI Innovation Design Lab

Starting to build more openly around AI Innovation Design Lab. Focus: practical AI products, prototypes, and workflow tools. First live tool: MessyData, for turning messy pasted text into cleaner structured data. I’ll share what gets built, what works, what breaks, and where AI actually helps.

MessyData is now live on StartupBase. I’m treating it as a small distribution experiment. Can a launch platform bring people with a real messy-data problem to the product? I’ll watch referral visits, actual use and feedback. startupbase.io/products/mes... #buildinpublic #ai #innovation

Most of my work on MessyData lately has been trying to work out how someone would actually discover it at the point they have messy text to clean up. Uneed and AppBoard were both small tests of that. The harder question now is whether the right people ever come across it when they need it.

MessyData.app has answered a useful question: can AI turn inconsistent text into a structured table? On the examples I tested, yes. A bigger question is still open: is that problem frequent enough for people to keep using a dedicated tool? That’s the evidence I need... #ai #innovation #messydata

I was asked whether I define the evidence threshold before every experiment. Ideally, yes. But the first test around a new hypothesis can be exploratory too. Sometimes that first cycle is what tells you what is actually useful to measure in the next one. #ai #innovation

AI makes it easy to keep building. I’m trying to decide what evidence would change my mind before I see the result. Lab method: question → smallest useful experiment → evidence → decision. aiinnovationlab.co.uk/blog/how-i-s... #ai #innovation #productdevelopment #buildinpublic

AI can play quite different roles in the build of a product. In the Lab I designed MessyData to use AI in the product itself. While PhysioTracker doesn’t need AI in the user experience, but I did rely on AI to help build it. Two different examples of AI-powered experiments for different reasons.

I expected AI to make coding PhysioTracker faster. What surprised me was how much it sped up scoping, testing and even the temptation to add more. That made prioritisation feel even more important. I wrote about what I learned, here: aiinnovationlab.co.uk/blog/what-bu... #ai #productdevelopment

Seeing MessyData described on AppBoard helped sharpen where it fits. Airtable and Zapier cover much broader data/workflow needs. MessyData sits in a smaller step: paste inconsistent text, turn it into a table, review, export. appboard.pro/product/mess... #AI #DataCleaning #BuildInPublic

Building MessyData and PhysioTracker has made me think about how easy it is to reach for the same app architecture each time. MessyData is stateless. PhysioTracker keeps its data in the browser. Avoiding extra infrastructure helps keep the product focused. #BuildInPublic #IndieDev #WebDevelopment

Another live experiment in the Lab is PhysioTracker: a simple way to log physio exercises and build up a record over time. I kept it deliberately local-only. No account or cloud sync; the data stays in the browser unless you export it. physiotracker.app #ai #buildinpublic #indiedev

RenewalCheck started because I was going through my own insurance renewal and thought a small tool might help me check the documents. Prototyping it made the privacy and trust questions obvious, without evidence the idea was worth pushing further. So I stopped. aiinnovationlab.co.uk/experiments

I’ve been making the Lab website a better record of the experiments themselves. I’ve now included live, paused and concluded projects, with a short note on what each one produced or why I stopped. I think the dead ends are worth keeping visible too. aiinnovationlab.co.uk/experiments

A core design choice in MessyData is that it works from pasted text. It doesn’t upload or parse PDFs directly. Text copied from a PDF may work, but multi-column layouts would probably scramble the reading order. Curious if future work would benefit from document support.

One question behind MessyData is whether a focused tool is useful when a general AI assistant can already structure messy text. I’m testing whether removing the prompt-writing and wrapping the task in review and export makes the workflow easier. messydata.app

MessyData started with a fairly ordinary frustration: information can make perfect sense to a person while still being awkward to get into a spreadsheet. That looked like a useful place to test whether AI could help. I wrote about how that became MessyData: aiinnovationlab.co.uk/blog/why-i-b...

I’ve been reshaping the Lab website so the experiments and the thinking behind them sit together. MessyData now links directly to the article on why I built it. Over time I’d like the site to become a record of what I built, why, and what happened next. aiinnovationlab.co.uk #ai #innovation

In this workshop-cost example I posted, one entry was marked “amount pending” and the formatting varied between rows. MessyData still structured the table, but surfaced those issues for review rather than treating every inferred value as certain. That's the benefit of AI to infer &/ ask questions.

Getting MessyData working answered a technical question: can AI turn awkward pasted text into a useful table? What it hasn't answered is whether enough people have that problem, or whether the workflow is useful enough to keep developing. That is what I'm curious to learn next. #ai #innovation

AI is making it easier to turn an idea into working software. For me, that puts more weight on deciding what is worth building and what the evidence supports. That is why I started the Lab. 👉 aiinnovationlab.co.uk/blog/why-i-s... #aiinnovation #buildinpublic #ai #productdevelopment

I’ve enjoyed building MessyData by keeping the core workflow simple. Paste messy input, make a server-side AI call, return structured data, review, then export. No need for a database. It’s been useful seeing how much can be built around a fairly small bit of AI functionality.

MessyData is featured on Uneed today. I submitted it as a small experiment in getting something I’ve built out to places where people might discover it. I’m curious whether directories like this generate any useful feedback. www.uneed.best/tool/messy-d...

A hard part of MessyData is defining when not to trust the AI interpretation. It works best with repeated patterns, not long prose or unrelated notes, so review, warnings and safe-use limits are part of the product. It quickly becomes a fuzzy boundary though!

Here is a concrete example of what MessyData does. This fictional input contains copied expense notes with inconsistent dates, separators and a missing amount. MessyData turns the pasted text into a table, highlights points to check, and exports it as CSV. messydata.app

A prototype doesn’t always need to prove an idea will work. Sometimes it just needs to answer the next question. - Will people understand it? - Can it actually be built? - Is the problem worth solving? Build enough to learn something, then decide what comes next on the path to success.

Building small software experiments means not every idea needs a huge commitment. Just build enough to make it real. Put it in front of people. And learn what happens next. Some ideas grow, some change direction, and some stop. All three outcomes teach you something.

One thing building MessyData has taught me: The first version of a product isn’t just there to validate the idea. It’s there to teach you how to explain it. The questions people ask often reveal more than the features they request.

The best feedback I’ve had building MessyData has been the questions I’ve received: “Does it handle PDFs?” “What’s the tech stack?” “Why not agents?” Good questions expose assumptions you didn’t realise you made. Addressing confusion can be more valuable than adding features.

One thing I enjoy about building small AI products is discovering where not to add complexity. The first version of MessyData is deliberately simple. Every feature has to earn its place. I think that’s easier to improve than starting with something clever but hard to understand.

Building MessyData has made me appreciate something: The product isn’t the AI. The product is everything around the AI. The prompts, the review step, the expectations you set, the places you deliberately don’t automate. That’s where trust comes from.