AI in app development isn’t a passing trend. From writing code to generating test plans, it’s clear that artificial intelligence is changing how software gets built.
At Big Fish, we use AI to enhance our work, not replace our expertise.
The most powerful results come when human strategy and AI capabilities work together.
Here’s a look at how we’re using AI in app development at Big Fish today, where it helps most, and why experienced developers, designers and strategists are still essential.
1. AI in UI/UX Design: Exploring Ideas Faster
Explaining a complex feature in plain English and asking for layout suggestions used to be something we did on a whiteboard. Now, AI can help spark ideas by describing common UI patterns – like collapsible rows, slide-ins or modals – based on a feature’s purpose.
It’s not designing the interface, but it’s useful for breaking through creative blocks and thinking through different ways to structure interactions.
As for AI actually designing apps? We’ve tested several AI design tools, and the results weren’t usable. For now – and likely well into the future – every app we design for clients is still created by a real human who understands usability, accessibility and context.
2. AI in Coding: A Developer’s Power Tool
Another way we use AI in app development is with Github Copilot. It supports development by:
- Writing small blocks of code based on the existing code
- Suggesting functions based on the project’s existing structure
- Identifying bugs and offering fixes
It’s a productivity boost that helps our developers move faster – especially when it comes to repetitive or boilerplate code. With those basics handled, our team can focus on what matters most: laying the foundation, designing architecture, writing the rest of the source code and solving complex technical challenges.
But let’s be clear – we are not using AI to write full apps. Many of the apps we build at Big Fish are complex, with integrations across third-party libraries, APIs and data sources. We still write the vast majority of our source code by hand. AI is a tool, not the builder.
And like any tool, it comes with risk. AI lacks contextual awareness across files and projects unless explicitly prompted (and even then it’s iffy), which makes it prone to introducing integration issues.
If you blindly accept its suggestions, you may wind up with bugs or flawed functionality. We’ve seen it firsthand: AI has introduced mistakes into code that had to be caught and corrected by our team.
What makes AI feel so helpful is also what makes it risky: it presents ideas clearly and confidently – even when they’re wrong. That’s why experienced developers are essential.
We know when the output makes sense, and when it’s leading you down the wrong path.
3. AI in QA: Better Test Coverage
AI is a valuable assistant when it comes to testing. We use it to:
- Generate test cases based on clearly-written user stories and criteria
- Write automated tests based on the app’s source code
- Suggest edge cases and input combinations we might have missed
- Create realistic mock data for more effective testing environments
AI helps us test smarter, but it still needs direction. It doesn’t know what to test until we’ve told it how the feature should work. That’s why we feed it both user stories and detailed acceptance criteria.
Once it has that context, it can generate accurate test cases, suggest edge scenarios, and even flag gaps in coverage.
But the quality of what it delivers depends on the clarity of what we give it.
4. AI for Strategic Planning and Architecture
When we start a new project, we can feed AI a high-level overview and ask it to:
- Suggest processes that could be automated
- Flag redundant steps in a workflow
- Identify potential use cases for 3rd party integrations
Sometimes AI suggests something we hadn’t thought of. That’s useful. But we still have to step back and ask, “Does this support the goal?”
ChatGPT, and other general purpose AI tools, will always find a way to give you an answer (even if that means making it up). It won’t ask you why you want a feature, or whether there’s a better way to solve the problem. That’s what we’re here for.
5. AI in Documentation: Fast Drafts, Human Polish
Documentation isn’t just a deliverable – it’s how teams stay aligned with each other and our clients. AI helps us create:
- Internal feature documentation
- Client-facing feature lists
- How-to guides and quick-start docs
- Changelogs based on Git commits or Jira updates
It saves time by creating structure and rough drafts. But those drafts still need to be reviewed for accuracy and edited by a real human.
Why You Still Need a Human Expert
AI is a powerful assistant, but it still needs guidance, judgment and context.
It won’t stop to ask whether a feature supports your goals, and it usually doesn’t know to push back when something feels off.
AI will give you an answer – even if it’s not the right one.
That’s why you still need a strategic partner who knows when to follow the AI’s lead – and when to reframe the question.
Here’s what experienced humans bring to the table:
- We ask “why” when something doesn’t feel aligned with your goals.
- We judge quality, not just if something works, but whether it’s right.
- We translate context from your industry, internal workflows and priorities.
- We write the majority of the code, especially the integrations and complex logic that AI simply isn’t equipped to handle, yet.
- We catch mistakes the AI introduces (and we’ve seen it happen firsthand).
- We refine prompts to get better outcomes, sometimes this requires several iterations.
- We design systems that work in the real world, not just in theory. That means thinking about edge cases, dependencies, performance and long-term maintainability.
At Big Fish, we use AI to move faster, explore broader and test smarter. But knowing whether the result is useful, feasible or accurate? That still takes human expertise.
AI Is Advancing Fast, And We’re Ready for It
AI is evolving quickly. Some tasks we once thought required human hands – like writing test cases or drafting documentation – are now fair game for AI.
But someone still has to steer the ship.
Our role is shifting. Some tasks we used to spend time on can now be delegated to AI like you’d delegate to a junior level employee. That’s exciting.
At Big Fish, we don’t resist AI. We explore and test it.
Because the tools are only as effective as the team behind them.
Behind the Scenes.
Yes, We Used AI to Help Write this Article 🙂
Fun fact: ChatGPT helped us write this article.
More specifically, a custom GPT trained on everything we do at Big Fish – our values, our process, our brand voice, and every piece of content we’ve ever published. Think of it as an AI assistant that knows us really well.
To write this article we started by giving it a topic, a title, and a list of how we use AI in custom app development. From there, we asked for an outline.
And then the real work began.
We went back and forth refining the outline and correcting mistakes.
That’s an important theme here: AI will give you answers, but not always the right ones.
That’s why we say AI often responds confidently, even when it’s wrong. And why an experienced human needs to be in the loop. To which ChatGPT usually responds:
“Good catch—you’re right.”
or
“Yes—that distinction is really important, and the way you’ve explained it is exactly how it should be framed.”
(If you’ve used AI tools before, you’ve probably seen responses like that, too.)
Once the outline was solid we asked it for a draft of the article. ChatGPT delivered and we began several rounds of revision of the draft. Still faster than writing it entirely from scratch though.
Something else we noticed while using AI to help write this article? When we asked ChatGPT to refine one section, it sometimes updated other sections too – ones we didn’t ask it to touch. So we had to say things like, “Please keep the second version of the introduction.”
After several rounds of revision, editing, and human judgment, we arrived at the version you just read.
We share all of this to say: AI for app development isn’t a one-and-done tool. You don’t just enter a prompt and walk away with polished, perfect work. It still makes mistakes. It still needs a skilled human to guide it, shape it and know what’s worth keeping.
This article originally started with ChatGPT writing:
“AI is having a moment—and for good reason.”
Our response?
AI isn’t having a moment – AI is the future.
