# Jessica Mumby — jessicamumby.com (full content)

> Award-winning product leader and consultant with 12+ years across Banking, Big Tech, Pharmaceuticals, and UK Government — including Google, The Guardian, Made Tech, Oliver Wyman and GE Healthcare. Builds AI-powered products from 0 to 1 with user-centered design, ships her own apps (Scroll Books, Brushlog), and is available immediately for new roles. Also known as Jess Builds Tech.

## About

I'm an award winning product leader and consultant with 12+ years across Banking, Big Tech, Pharmaceuticals, and UK Government.

I build products from 0→1, turn around legacy platforms, and lead organisational change. I care about outcomes, not outputs.

Deep expertise in AI powered products, from the UK Government's first AI fraud engine to brain scan quantification, leading multidisciplinary agile teams from backlog to boardrooms.

I'm now back in Newcastle after a five month break travelling Australia and New Zealand, and I'm available immediately for my next role.

### Work history

- 2025–26: Lead Product Manager, hedgehog lab
- 2021–25: Product & Delivery Lead (UK Gov), Freelance
- 2020–21: Senior Delivery & Product Manager, Made Tech
- 2019–20: Programme Manager, New Products, The Guardian
- 2018–19: Strategy & Operations, Google
- 2017–18: Senior Management Consultant, Oliver Wyman
- 2014–17: Global Product Manager, GE Healthcare

### Stats

- 12+ years in product
- 5+ legacy transformations
- 15+ product launches
- 10M+ users across all products

## Selected work

### Gov AI Fraud Engine (UK Gov · Cabinet Office)

Built One Login's AI fraud product, the first of any Government, which now screens millions of authentications to systematically identify and reduce fraud.

### NHS Virtual Visiting (Made Tech · NHS)

Award winning COVID-19 virtual visiting product for the NHS, 250k+ visits at 1hr+ average duration. Featured in Wired and won "Best IT Project".

### Scroll Books: Beat Doomscrolling (Scroll Books · Myself)

Shipped an App to beat doomscrolling by leveraging dopamine from scrolling into reading books. Available on Android and iOS.

### EdTech AI (hedgehog lab)

Led an EdTech AI product through product market fit and GTM, 400k+ users and growing MRR by 16%+.

### Brain scan AI/ML (GE Healthcare)

Shipped an AI/ML product for quantifying brain scans, and launched pharmaceuticals into new markets around the globe.

### RBS Digital Bank (Oliver Wyman · RBS)

Led product strategy and delivery for a new digital bank alongside 20 developers and 5 designers with a focus on UX, technical architecture, and stakeholder management.

## Apps

### Scroll Books (Live)

An infinite scroll reading experience. Flutter. Custom algorithm to chunk books into a readable scroll, very nerdy under the hood.

- Android: https://play.google.com/store/apps/details?id=com.scrollbooks.scroll_books
- Apple iOS: https://apps.apple.com/app/scroll-books/id6770981737

### Brushlog (WIP)

A miniature painting tracker for tabletop hobbyists. Flutter. AI assisted color suggestions.

## Manifesto

- AI is a collaborator, not a vending machine
- Share the embarrassing first version, don't ship it
- Product includes the work of deciding what not to build
- Fall in love with the problem, not the solution
- The interesting work happens at the edges

## Writing

### AI is a collaborator, not a vending machine (2026-08-18)

https://jessicamumby.com/blog/ai-is-a-collaborator-not-a-vending-machine

I've been teaching computers to think since ~2013 and over the last few years I've been building almost exclusively AI powered products, from the first AI powered fraud risk engine in Government to an EdTech product used by 400k+ teachers. If I had to compress what that taught me into one line, it's the first line of my manifesto: Treat AI as a collaborator, not a vending machine.

A vending machine takes your coins and dispenses a fixed product. You press a button, something comes out, and the transaction is over. That's how a lot of people are using LLMs. They type a prompt, take the answer, and move on. The problem is the answer is only useful if you already know what you're looking for. Vending machines don't have conversations, they don't push back, they're sycophantic.

The AI powered fraud risk engine I built for Government wasn't a vending machine. It wasn't built to have questions fed into it and take whatever fell out. Instead, it screened millions of authentications and flagged what looked wrong, and then humans decided what to do with the flag. The product allowed Government to act at scale whilst still paying attention, but it didn't make a judgement call.

Since the early days of computing, having a collaborator changes how we think. Remember rubber duck sessions with colleagues on hard problems? Collaborators help you spot patterns you would have missed on your own, but it's still up to you to decide if it's a real problem or not.

I've recently spent a few months in Australia and New Zealand, and in my spare time whilst travelling I built Scroll Books - a reading app for beating doomscrolling, with AI assistance for coding. The app went from idea to the App Store and Google Play because I could use AI to move faster on the parts of the code I already understood, and to argue with me on the parts I didn't so I outputted something of value faster. The decisions that matter stayed with me, of course. A vending machine can't have good product sense.

The EdTech product taught me the same lesson from the other direction. 400k+ teachers used it, and the users who got the most from it were the ones who treated it as a tool to work with, not a vending machine to get answers from. AI that hands you certainty is a shortcut to the wrong answer. AI that shows its work and lets you push back is a tool for getting less things wrong, which compounds over time to create better outcomes.

So when someone asks me how to use AI in product work, I don't tell them to prompt better. Instead, use these models as a sparring buddy. Give it enough context to go to battle with you, show it your work, and then ask it to disagree with you. The vending machine gives you what you asked for, whilst the collaborator gives you what you actually need to hone your craft.

### The first few weeks of Scroll Books (2026-06-25)

https://jessicamumby.com/blog/the-first-few-weeks-of-scroll-books

Scroll Books has been live in [the App Store](https://apps.apple.com/app/scroll-books/id6770981737) and on [Google Play](https://play.google.com/store/apps/details?id=com.scrollbooks.scroll_books) for about a month. It's a reading app built around short passages from [public domain books](https://www.gutenberg.org), meant to fit into the gaps in a day I'd normally be doomscrolling social media.

I've been building it since February, but the last few weeks were the first time it was a real product that strangers could download. [This is a write up](https://jessicamumby.com/blog/hello-world) from the first few weeks.

Here's what I actually did, and what it taught me.

#### Shipping is mostly the part after "it works"

The app "worked" months ago. Getting it into the stores was a different job entirely. Apple won't approve an app that lets people create an account but not delete it, so I had to build in app account deletion before I could ship at all. I also needed privacy policy links in Settings, correct build numbers, and submit data safety forms for both Apple and Android. None of it was interesting but all of it was required; I'd quietly assumed it would take an afternoon. It did not!

Around the same time I spent a full day paying down dependency debt I'd been ignoring: eight or nine major version bumps in one sitting, plus a major Flutter upgrade and clearing out ~100 warnings. Deferring major upgrades feels free right up until the day you do them all at once. A lesson learnt.

#### Then I looked at the data, which was humbling

Once there were enough users to bother measuring anything, I sat down and did a proper review against [Amplitude](https://amplitude.com). It's a small sample, ~100 people over ~three weeks. The data was not flattering. Day 1 retention was ~8%, day 7 was ~2%, day 30 was near zero. ~10 people were genuinely coming back per week.

Three things stood out.

The biggest problem wasn't a missing feature. It was a bug: returning users were being forced to log in again, every single time - Gah! Auth screens accounted for ~40% of all screen views. People were spending nearly half their time in the app just trying to get back into it. I didn't see myself, as I was always already logged in.

The second was a big signup cliff. ~75% of people finish onboarding but only 37% create an account.

The third was the one I didn't want to hear; Streaks have never worked. The longest streak any user has ever recorded was ~3 days, and the streaks tab was opened a handful of times in a month. I'd built streaks because every app has streaks, and I'd assumed people would find this loop helpful for reading, but the data was telling me I'd shipped a feature nobody wanted that also made the app feel like it was nagging them.

#### Acting on it

So I spent a week just doing what the data said would help.

I fixed the re-login bug and pushed it out as a priority release. I found a few other somewhat related bugs at the same time, so I fixed those as well at the same time.

I moved account creation to after the first reading session instead of before it, so people reach the actual reading before the app asks them for anything. Hopefully this will soften the signup cliff, I'll monitor the data over the next month to validate.

Oh, and I killed streaks! I replaced them with something quieter, more focused on reading. Now each day's reading just lands as a small note, a quick read or a good read or a long read, measured in minutes, with nothing to 'break'. You can set a gentle intention, or ignore it all together. It's the opposite of gamification, and it's a bet that the kind of person who'd actually stick with a reading app doesn't want to be nagged by one. I might be wrong about that, but I'll track it over time.

#### The unglamorous floor

The app had been shipping with no crash reporting, so a crash in production produced exactly zero signal. I wired up [Sentry](https://sentry.io). I added CI that runs all the tests on every pull request, which I should have had from the first week.

I replaced my hand driven release process, where I manually bumped the version, manually built, manually uploaded and hoped, with a single repeatable script that refuses to ship if the tests fail and walks me through the store checklist before it does; which has been very helpful.

#### What I think I learned

Mostly that I'd been optimistic about the wrong things. I was adding features while the thing keeping people out was a login bug, and the thing quietly pushing them away was a feature I was proud of.

The first month's numbers are small, so I'm not going to over read them, but they were specific enough to argue with, and arguing with them turned out to be more useful than any roadmap I'd written out of my own head!

Next month I get to find out whether any of it moved.

### Hello, world — and why I'm writing here (2026-06-25)

https://jessicamumby.com/blog/hello-world

#### Why a blog

I build AI-powered products, and I have opinions about how that should go. This is where I'll think out loud about product, engineering, and the occasional manifesto.

#### What to expect

Short, opinionated posts. No filler. If I learn something building in public, it ends up here.

## Contact

- LinkedIn: https://www.linkedin.com/in/jessbuildstech
- Email: dailyjessicamumby@gmail.com