)
)
TL;DR
Anyone can build an app in 2026. But "building" and "building well" are two very different things. Here's what's actually hard:
Knowing your user well enough that your first assumption isn't the one that sinks you
Designing an experience so obvious nobody has to think about it
Making architecture decisions in month one that still hold up in month six
Treating launch as the starting line, not the finish
AI tools are genuinely brilliant for prototyping and pressure-testing ideas fast. The judgement and the strategy still very much need a human touch!
More apps than ever...
Building an app used to mean a team, a budget with a lot of zeros, and the best part of a year with your fingers crossed. Not anymore.
Now a small charity with a sharp idea and a spare afternoon can have something live by the weekend. And the receipts back it up: new app releases were up 60% on last year in the first three months of 2026, and have more than doubled by April 2026 when compared to April 2025. That’s quite a lot of movement on ye olde App Store!
We think that's brilliant. Artificial Intelligence (AI) tools have knocked down the barrier that used to keep smaller, mission-led orgs out of the room entirely. More people building means more good ideas getting a proper shot, not just the ones with deep pockets. The door's wide open, and we're here for it.
But (you knew there was a but) making an app and making an app people actually use are two very different beasts. One of them just got a whole lot easier. The other one didn't budge. And that second one is where the real work has always been.
These Hips (Numbers) Don't Lie: More Apps, Same Usage
So everyone's building. Cracking. Now for the bit that doesn't make the launch announcement: all those shiny new apps haven't won a single extra user between them.
The National Bureau of Economic Research (NBER) crunched the data on more than 100,000 developers using AI tools in 2026. The tools did exactly what they promised. Code output shot up by as much as 180%. But follow that code all the way to actual humans and the gains quietly evaporate. That 180% more code turned into 50% more finished projects, then just 30% more things that genuinely shipped. Total usage across four major app stores? Didn't move an inch.
More apps. Same number of people bothering to open them.
The app stores have clocked it too. Over the past couple of years Google Play has gone from around 3.4 million apps down to roughly 1.8 million, chopping itself nearly in half. Google did that on purpose. Call it a spring clean. It raised the bar on what's even allowed in, binning the apps that were a single wallpaper, a glorified PDF, a page of text, or just plain broken. In one year it blocked 2.36 million rule-breakers and pulled the plug on 158,000-odd developer accounts.
If you're building something that actually does a job, this is brilliant news for you. Less junk means your good thing is far easier to find.
So picture two bars moving in opposite directions. The low one, just getting something built and live, has basically hit the floor. The high one, being good enough to keep, find and use, has crept up. Anyone can make an app now. Far fewer can make one that earns its place on someone's home screen.
Why Do Most AI-Built Apps Fail to Gain Traction?
We've been in the app game since 2009, so we def have tried and true thoughts ‘n feels about why this happens.
)
1. They skipped the homework.
You have to actually know who you're building for. The old adage of, “to know thy user is the beginning of wisdom” (Ok, so we might have taken slight liberties with that one. But it’s not a vibe, or a hunch).
Who are they really? How often does this problem actually bite? What are they using right now, and why is it letting them down? That's the entire point of scoping your project with a proper Discovery and Definition (D&D) session before anyone touches a keyboard. Skip it and, as our team likes to say, you're building on quicksand.
2. They treated launch like the finish line.
Big mistake. This one makes us clutch our pearls. Loads of AI-built apps are tuned entirely for the big reveal: demo day, the press release, the App Store listing. Then what? Adoption, retention, and the unglamorous graft of watching what real people do and fixing what trips them up. That's the bit that turns "fine" into "actually useful." If the post-launch plan was never in the build, you're on the back foot before you've begun.
3. The user experience was an afterthought.
If someone has to work out your app, you've already lost them. A great app just feels right the second it opens and that feeling doesn't come from a default component library.
No clunk, no head-scratching, no "where do I even tap." This is exactly where AI tools tend to wobble. They're trained on what already exists, so by default they hand you a copy of something that's already out there. Genuinely good user experience comes from understanding your specific users in their specific lives, and that isn't sitting in any training set.
4. The app was built around a tool, not a problem.
It's tempting to start with what the shiny tool can do rather than what your user actually needs. You end up with something that technically works but doesn't really fix anything that sticks. Real users could not care less how fast it was built or which model wrote it. They care that it makes their day easier.
5. The tech was built for "works now", not "works at scale".
AI can get you to a working prototype. What it's far less good at is seeing round the corner: what happens when 10,000 people pile in at once, when an edge case nobody tested rears its head, or when a security hole gets found because the foundations were never built to take the weight. Dodging those landmines takes experience, not just a clever tool. This is where a team that's done it before earns its keep.
What Does It Actually Take to Build an App People Use?
Honest answer? It starts long before anyone writes a line of code.
It starts with understanding the problem so well you could explain it to literally anyone (yes, anyone). It means thinking six months out, not six minutes, and making calls now that still hold up long after the launch buzz wears off.
A few things AI genuinely can't do for you here:
The user research that reveals your first assumptions were wrong.
The judgement call about what to cut when scope creep comes for your timeline.
The architecture decision that feels slower today and saves you a full rebuild come month six.
And going live? That's the starting point, not the finish line. The apps that actually make it are the ones where the team keeps showing up: building in feedback loops, watching what real people do, and improving long after the launch confetti's been hoovered up.
That's what it takes. No magic wand, no single clever tool. It's the difference between an app that launches and an app that lasts. (We went properly deep on the front end of all this in our guide to scoping an app without building on quicksand).
What We've Seen Building Apps for Organisations That Can't Afford to Get It Wrong
We've built apps across a load of sectors, and the stakes look different every single time. The one constant? The gap between something that just ships and something that genuinely works for real people.
)
We partnered with Surfers Against Sewage on their Safer Seas & Rivers Service (SSRS), covering 650+ bathing locations and trusted by over 208,000 users. SSRS alerts people to pollution events, helps them take action against water companies, and supports research into the health impacts of dirty water.
For the American Red Cross Blood Donor app, that meant turning declining blood donation into a movement. The app has now booked over 19 million appointments and raised more than $90 million for the Red Cross.
For EPOG's Power Path, it meant scaling personalised coaching to students in schools where the counsellor-to-student ratio is 400:1, without losing the human warmth that makes coaching work in the first place.
For Glastonbury, it meant an app for 250,000+ attendees that had to work with no signal, handle 3,000+ performances, and never once ruin the magic of the moment. 450,000 in-app hours later, it did exactly that.
For Cubbi, the UK's first parental leave discount platform, it meant rebuilding a no-code foundation into something that could scale. Something sturdy enough to survive national TV exposure on Dragon's Den, while cutting approval times by 90%.
(Have a nosy at our other case studies for more of where this has played out.)
None of that happens without obsessing over the user journey, the friction points, and the reasons people keep coming back. Different sectors, different users, same rule every time: building well is the bit that makes the difference.
So Where Do AI Tools Genuinely Help?
Right, time to give AI its flowers, because it'd be daft (and a bit dishonest) not to. We use AI tools at Cube every single day, and they're genuinely brilliant. In the right context, with the right humans steering.
For rapid prototyping and testing an early idea? AI is a proper accelerant. Getting to a proof-of-concept that's good enough to put in front of real users, fast and cheap, is exactly where it shines.
Internal, low-stakes tools that don't have to carry the weight of scale or sensitive data? Lovely. Go wild, test things, learn quickly.
The deciding factor is still the expertise around it. Anything with real users, sensitive data, serious scale, or a mission that simply can't fail deserves proper care. That's where the judgement, the architecture, and the quality of thinking behind the build become the whole ball game.
Ready to build a great app?
Already knocked together a prototype with AI tools and wondering what on earth comes next? We've written about that exact moment. Read it here.
Or shout us a holla and let's take an honest look together 💚
No jargon, no sales-y nonsense. Just a straight conversation about what it'll actually take to build something people use.
)
Frequently Asked Questions
Why do most AI-built apps fail to gain traction?
Five reasons crop up again and again:
They skipped market validation.
They treated launch as the finish line.
The user experience was an afterthought.
They were built around a tool, not a problem.
The tech was optimised for "works now", not "works at scale".
What does it take to build an app people actually use?
Start with the user's problem, not the tech. Put real people at the centre of every decision, plan for life after launch, keep improving, and make technical calls built for month six, not just demo day. Building well is what separates an app that ships from one that sticks.
Is AI good enough to build a production-ready app?
Not on its own. AI tools are brilliant for prototyping, validating ideas, and speeding up skilled developers. But for anything with real users, sensitive data, or scale (which is most app projects), human expertise still decides the outcome.
What's the difference between building an app and building a great app?
Building an app means shipping something that works. Building a great app means starting with a real problem, designing for messy real-world conditions, and planning for what happens after launch. Most builds nail the first part. Far fewer nail the second.
When should I use AI tools vs hire an app development agency?
Use AI tools to prototype fast and pressure-test ideas. Bring in an agency when your app has real users, mission-critical requirements, security considerations, or needs to scale without falling over.
Published on 16 June 2026, last updated on 16 June 2026