AI in Product Development: How Do We Build Better Products Faster?

Every week brings new tools promising to transform product development, and somewhere between the hype and the skepticism, we can all agree how AI is changing everything.

At In The Pocket (ITP), we've spent the last year integrating AI across our entire development process. We’ve fundamentally changed how we work without the chaos that usually accompanies rapid tech adoption. This article isn't about replacing human expertise with artificial intelligence. It's a testament to what makes great product teams great—strategic thinking, creative problem-solving, and relentless focus on user value.

What Is the Real Promise of AI in Product Development?

You know the saying—”you can have it fast, cheap, or good, but pick only two”? AI is actually changing that.

However, the data supports the promise but demands nuanced implementation. Stanford research from September 2024 shows 30-40% faster code generation with AI, but there's additional waste in generated code and time lost to AI errors. The net productivity gain averages 15-20%—but this research predates major advances like mainstream adoption of Cursor and Claude Code, so gains have likely increased further.

Product development is much more than writing code, so the productivity gains go beyond just coding. Meaning, these gains vary dramatically by context:

Three Ways We’re Using AI 

We've structured our AI approach around three key dimensions, each designed to amplify our existing strengths.

1. Making our core skills better

1. How Can AI Speed Up User and Desk Research?

How can AI speed up user research? AI automatically transcribes interviews and finds patterns. We're also testing AI to run interviews and create research plans. Analysing research that used to take five days now takes one or two. This doesn't just save time, it means we can talk to users more often, learn faster, and dig deeper into what they really need.

2. Can AI Improve UX/UI and Visual Design Processes?

AI doesn't magically fix bad design processes, you need good foundations first. AI is great at patterns and repetition, as well as prototyping for quick validation. 

3. What Does AI Mean for Product Managers?

Product managers usually juggle information from dozens of sources—user feedback, business goals, technical constraints, market research. AI helps us pull all this scattered information together and turn it into clear decisions through automated user stories, searchable project knowledge, and AI-powered competitive research.

4. How Does AI Transform Software Engineering Productivity?

We now use Cursor as our main coding environment, with Claude Code for the really tricky problems. Here's what our AI-powered development looks like:

Core capabilities:

• Implementation planning assistance
• Large-scale refactoring across multiple classes (eliminating code rot)
• Synced cursor rules and CLAUDE.md files for project context
• Rapid analysis and fixing of complex errors and stack traces

Integrated tools via Model Context Protocol (MCP):

• JIRA integration: Extract context from tickets, create implementation plans
• Confluence: Quick access to feature documentation and error codes
• Figma: Reference reusable UI components (full design integration still developing)
• Context7: Up-to-date documentation for common frameworks

Here's a real example: our team took a design from Figma, fed it to an AI coding platform, and within minutes had a working application that looked exactly like the design—without writing any code by hand. It still needs a developer to review and polish it, but it eliminates weeks of routine coding work.

5. Can AI Really Improve Quality Assurance and Accessibility?

Good quality control is essential for shipping fast without breaking things, but it's usually time-consuming to set up. AI changes that:

Measurable quality improvements: Our QA automation platform transforms testing efficiency. Automated test plan generation speeds up creation, while risk-based selection reduces volume and eliminates flaky failures. CI/CD cycles run faster without compromising detection, and enhanced coverage delivers rapid bug reporting across the entire quality assurance process.

 Validating accessibility automatically: New laws require websites to be accessible to people with disabilities. After every build, we automatically generate simple reports that anyone can read—no technical knowledge required. These reports explain what's wrong, how to fix it, and how serious each issue is.

2. The Roles are Changing 

1. What Does It Mean to Be an AI Orchestrator?

The biggest changes are happening to how our teams work. Software developers are becoming "AI orchestrators"—people who:

• Stay current with AI tools and know when to use each one
• Make smart decisions about which AI tools to use for specific jobs
• Set up guardrails to ensure quality and consistency
• Orchestrate AI-generated code while maintaining oversight

‍

2. How Do Design Engineers Use AI to Go from Figma to Code?

We've been experimenting with "design engineers" for two years—specialists who translate visual concepts directly into functional code and:

‍

• Make sure the final product looks exactly like the design
• Keep design systems consistent and up-to-date
• Build accessibility into everything from the start
• Eliminate the usual back-and-forth between designers and developers

With AI assistance, design engineers can take a Figma design and immediately generate corresponding frontend components, leading to innovative, functional, accessible, and pixel-perfect products.

3. What Are AI-Powered Product Agents?

Usually, project knowledge gets shared through meetings and documents that quickly become outdated. AI lets us create "product agents"—smart repositories that contain all project information and can:

• Remember project history better than any human
• Get new team members up to speed faster
• Answer questions instantly
• Challenge new decisions based on full project context

3. Reimagining our software discovery and delivery process

The ITP way didn't emerge overnight. We let natural forces shape our approach, allowing our principles to guide us through each technological shift. We're running small experiments that will reshape our delivery processes piece by piece. These experiments test new ways of working, new combinations of human expertise and AI capability, new approaches to quality and speed.

People often think using AI to build things faster means lower quality. In our experience, it's the opposite. AI generates more code, but it also enables more testing, better documentation, and more thorough reviews. 

We're not just building faster—we're building better.

What Makes Human + AI (HAI) Teams So Powerful?

Unlike companies that claim they've got AI all figured out, we're honest about what we don't know. The AI world is changing so fast that anyone making definitive predictions is probably wrong. What we do have is more valuable: a proven way of handling change, a history of adopting new technology early while maintaining quality, and the honesty to say when we're still figuring things out.

As we continue to experiment and refine our AI integration, one thing remains constant: our commitment to principles over process, value over velocity, and solutions that truly serve the people who use them.

‍