AI in Agile Product Teams

I have been interested in how artificial intelligence as an emerging technology may shape our work since the advent of ChatGPT; see my various articles on the topic. As you may imagine, when OpenAI’s Deep Research became available to me last week, I had to test-drive it.

I asked it to investigate how AI-driven approaches enable agile product teams to gain deeper customer insights and deliver more innovative solutions. The results were enlightening, and I’m excited to share both my experience with this research approach and the key insights that emerged. 

Working With Deep Research: A New Level of Analysis

My experience with Deep Research was remarkably productive. After providing a detailed research prompt exploring how AI transforms agile product development, I received a comprehensive synthesis that went far beyond what I expected from market research handled by an AI agent.

What impressed me most was how the research agent engaged with my initial request, asking clarifying questions about industry focus, development stages, timeframes, and company sizes of interest. This collaborative refinement process ensured the final report addressed my specific needs rather than delivering generic information. (Other agents are less inclined to do so; Perplexity or Grok, for example.)

Within just 11 minutes, Deep Research compiled findings from 16 sources into a cohesive narrative featuring three in-depth case studies and a thoughtful cross-case analysis. The analysis didn’t just aggregate information — it extracted meaningful patterns and presented actionable insights in a structured, easily digestible format.

(Download the complete report here: AI in Agile Product Teams: Insights from Deep Research and What It Means for Your Practice.)

Three Illuminating Case Studies

The report examined how diverse organizations leveraged AI within their agile frameworks to transform product discovery and delivery:

Lightful: Agile “AI Squad” Powers Nonprofit Communication

This London-based tech company formed a cross-functional “AI Squad” with designers, engineers, and product managers working in daily iterations. Rather than adopting AI for its own sake, they identified specific pain points for their nonprofit clients and experimented with AI solutions in short, rapid cycles.

Their most successful innovation was an “AI Feedback” tool that helps nonprofit users improve social media posts by providing suggestions with explanations. The solution educated users while augmenting (not replacing) human creativity. The team’s agile approach allowed them to quickly adapt when new AI models became available, swapping in improved technology within Sprints.

PepsiCo: AI Uncovers the “Perfect Cheetos”

PepsiCo employed generative AI and deep reinforcement learning to experiment with Cheetos’ shape and flavor. First, they built a digital simulation of the production process. Then, they trained an AI system to optimize variables like dough moisture, temperature, and machine settings — running thousands of virtual trials far faster than physical lab tests could allow.

The AI-designed “perfect Cheetos” drove a 15% increase in market penetration by aligning product attributes more closely with consumer preferences. PepsiCo combined human expertise with AI experimentation. Domain experts set clear objectives, while the AI explored the solution space extensively, identifying non-intuitive combinations that human R&D might have overlooked.

Wayfair: Generative AI Enhances Customer Visualization

Wayfair developed “Decorify,” an AI-powered interior design tool that lets shoppers upload photos of their room and describe a desired style. The generative model produces a photorealistic image of the space filled with Wayfair furniture and decor matching that style, with products linked for purchase.

Within months of launch, the tool had generated over 175,000 room designs for users. It addressed a critical customer need: helping me envision what furniture would look like in my space. Wayfair treated this as an MVP: launching early, then improving through iterative updates based on user feedback and usage data.

Six Key Patterns for Success

Across these case studies, Deep Research identified recurring patterns that contributed to successful AI integration within agile frameworks. As the report concluded:

“Common threads in our case studies include a relentless focus on customer needs, iterative development to harness AI’s fast improvements, cross-functional teamwork, and careful attention to ethics and data quality.”

The six key patterns worth highlighting are:

1. AI as an Insight Engine, Not Just an Efficiency Tool

In all three cases, AI revealed deeper customer insights that shaped product direction — from identifying content quality needs at Lightful, discovering precise product traits consumers love at PepsiCo and revealing style preferences at Wayfair. Organizations leveraged AI to uncover latent needs and patterns, not just to automate existing processes.

2. Customer-Centric, Problem-First Approach

Successful teams started with customer problems and needs, then applied AI as appropriate — not vice versa. This discipline prevented wasted effort on “cool” AI ideas that don’t move the needle. The question was always: “How can AI help solve this specific customer problem?” rather than “Where can we use AI?”

3. Agile Methods Amplify AI’s Impact (and Vice Versa)

The fast pace of AI advancement requires the adaptability that agile practices provide. Teams integrated AI work into their work cadence: using short experiments to test viability, Sprints to build AI-driven features incrementally, and frequent reviews to assess outcome quality with stakeholders. This created a powerful feedback loop where Agile’s adaptability enabled quick AI piloting, and AI-generated insights informed subsequent iterations.

4. Cross-Functional Teams and Skills Are Essential

AI projects intersect with data science, engineering, design, and domain expertise. The most successful implementations involved diverse teams with a shared language around AI. This prevented miscommunication and unrealistic expectations, allowing for smoother collaboration and more effective solutions.

5. Human Oversight, Ethics, and Data Quality

Teams created processes to verify AI outputs and mitigate errors or bias. This included adding QA steps in the definition of done, A/B testing AI decisions against human ones before full rollout, and proactively addressing ethical considerations. Transparency with users and ensuring regulatory compliance were essential.

6. Leadership Buy-In and Culture of Experimentation

Leadership support provided vision and resources, empowering teams to iterate without fear. Setting realistic expectations — not overhyping AI as magic but as a powerful tool requiring refinement — and communicating progress in terms leadership cares about (customer metrics, ROI, competitive advantage) were crucial.

Becoming Obsolete Is a Choice, Not Inevitable

What strikes me most about these findings is how they challenge the fear narrative around AI for knowledge workers. Many professionals view AI as a threat rather than a paradigm-shifting technology like the printing press, electricity, or the Internet. Yet, these case studies tell a different story.

In each example, human expertise remained essential. AI enhanced human capabilities rather than replacing them. At Lightful, the AI provided suggestions but kept humans in the creative loop. At PepsiCo, domain experts set objectives and guided the AI’s exploration. At Wayfair, the AI visualization tool helped customers make better decisions but didn’t replace the human shopping experience.

These observations suggest that becoming obsolete in the age of AI is a choice, not an inevitability. The practitioners who thrive will be those who learn to leverage AI as a collaborator — using it to uncover insights from unstructured data, simulate complex scenarios, and enhance their decision-making.

What This Means For Your Agile Practice

As agile practitioners, we’re uniquely positioned to embrace AI. The agile mindset — focused on adaptation, continuous improvement, and delivering customer value — aligns perfectly with the evolving nature of AI technology. Here are three takeaways for your own practice:

1. Start small, learn fast. Begin with specific customer pain points where AI might offer value. Run experiments in short iterations, gather feedback, and adapt quickly.
2. Build cross-functional AI literacy. Ensure your team has a shared understanding of AI capabilities and limitations. “Understanding” doesn’t mean everyone should become a data scientist, but everyone should understand enough to collaborate effectively.
3. Keep the human at the center. Design AI implementations that augment human creativity and decision-making rather than attempting to replace it. Most applications keep humans in the loop.

Many agile teams are currently missing opportunities to leverage AI for deeper insights — particularly in transforming qualitative data from user research, retrospectives, and customer feedback into actionable patterns. There’s enormous untapped potential in using AI to extract meaning from the rich but unstructured data we already collect.

Conclusion

The future belongs to agile practitioners who can pair human judgment with AI’s analytical power. We can deliver unprecedented value to our customers and organizations by embracing this partnership rather than fearing it. Deep Research is merely a glimpse into this future.

Have you experimented with AI in your agile practice? What opportunities do you see for AI to enhance rather than replace your team’s capabilities? How might we ensure that AI serves our agile values rather than undermining them? Please drop me a line or comment below.