How product managers can use AI as a research assistant for app strategy

[Editor’s note: This is the second in a series of blog posts about how builders can inject AI into the software development process].

Our agile development process typically starts with an app strategy phase and a discovery workshop.

Discovery workshops last 2-3 days and gather our team and all client stakeholders together in a series of live exercises. The goal: Align on the business opportunity and set the course for the project.

This workshop is a force multiplier that paves the way for effectively building products and, ultimately, delivering lovable experiences. Skipping it inevitably leads to budget overages, schedule delays, and negative UX consequences.

And for product managers like me who lead these sessions, it can be magical. For all the power that we can harness from our tech tools, there’s no replacement for getting a group of passionate and insightful people together to share, listen, and collectively carve out a path forward. People, not bots.

That said, there is absolutely an important role that AI can play in this process – and it happens before we enter the room. Because there’s a lot of work that we, as product and experience strategists, need to do even before we get there. Let me explain.

The ‘wall of stickies’

One visually impressive byproduct of our discovery workshops is what we call the “wall of stickies” – a mosaic of dozens of colored sticky notes arranged on a conference room wall.

While those who see a workshop-in-progress might see visual chaos, these stickies help us visualize and align dozens of key aspects of the project — including app strategy and opportunity, measurable objectives and key results, target users, competitors, technologies, and features. Some of our workshops and these sticky exercises take place in the digital realm, as we use FigJam online collaboration boards.

The outcome of this exercise — whether physical or digital — remains the same. These stickies give us the strategic path forward and launch us into the planning phase.

However, these all-important stickies don’t just get created during our workshops. Our product team writes many of them in advance as we prepare for the workshop from our own research. Then, as we go through the workshop journey, we will arrange these stickies, edit them, and create some new ones. But the up-front research is where generative AI tools can greatly help.

How AI-led research aids discovery workshops

Our pre-workshop research takes place in a few forms. We meet with our clients to talk through initial ideas about the business problem they’re trying to solve and the user experience we want to achieve. We bring knowledge based on past projects in similar industries or types of applications. And we rely on web searches to gather more information and the most recent info about an industry or company.

The temptation for builders may be to use AI as a Google search replacement – but the problem is that AI training data is often dated. Just ask AI, “How current is the publicly available data you were trained on?” and you’ll get an answer that shows how it can’t help understand the latest industry news or trends.

So, how can AI help with the discovery process, particularly before we have our human-powered workshops?

It’s early, but from our latest experiments, AI does a great job going beyond surface-level research — specifically in how it can synthesize vast information, generate new ideas, and offer inspiration for our workshops.

Here is an example.

Researching Dribbble: Example prompts

Let’s use AI to help us prepare for a hypothetical discovery workshop with a real company (that we love), Dribbble. A social media company for designers to share ideas and design concepts, Dribbble has a beloved website — but no mobile app.

We’ll use our internal WPP Creative Studio, a collection of state-of-the-art AI tools available to all agencies that are a part of WPP. For this workshop, we used OpenAI’s GPT-4. Here is a simple prompt we wrote for some initial business context before a full discovery workshop:

GPT-4 generates the following output:

Reviewing and re-prompting to go deeper

These first results are useful and certainly save us time, but there’s no insight here we couldn’t have surfaced with existing research techniques. And there are some inaccuracies — e.g. Dribbble does not have a mobile app. It’s a good reminder never to assume 100% accuracy with AI.

However, In our experience, engaging with AI becomes even more valuable the deeper you drill down in specific areas. So, let’s get more specific with Dribbble’s unique opportunities with this prompt:

And here is GPT-4’s output:

This result offers some more ideas and information worth discussing at a discovery workshop that we might not have otherwise come up with.

As mentioned before, AI can be wrong and “hallucinate.” So the final step is to follow up or drill down more by going back to primary sources and the Internet.

The verdict: AI just made our wall of stickies better

With a few good prompts, AI just gave us some great fodder we could use for our discovery workshop. For example, the AI-generated specifics on “unique selling proposition” could factor into a discussion about the competitive landscape and differentiation. And certainly, the “opportunities for delight” and “innovative opportunities” could spark discussion and alignment on the initial feature set of a minimum lovable product.

But how does our AI insight actually land on our wall of stickies?

We humans, of course, curate the sticky creation process, choosing the most important topics to put on our wall and driving our all-important workshop discussions. Our workshop leaders assimilate this AI-generated intel along with our other pre-workshop research, then manually write out the stickies. We’ve also been automating getting these into Figma (the ultimate destination for team collaboration) through this create sticky from text plug-in.

GPT-4’s responses are a big help in our workshop preparation. And in somewhat of a turnabout, our prompt-based AI research ultimately prompts us humans in the room to discuss the details of different topics and land on actions.

There are also some more intangible benefits of using AI as a research tool for our workshops. Even if a certain GPT-4 output didn’t land on a sticky, we’re more prepared by simply reading the AI analysis. Of course, we know AI’s limitations — but the information is better and more processed than what we’d find via some crude Google searches.

So, with AI, we’re a little “smarter.” And we saved time on research by avoiding a few web search rabbit holes — giving our builder brains a little more space to do things that AI can’t.

We can invest that time savings into pre-workshop feature ideation, user journey mapping, and persona sketching. We can do more innovating and problem-solving before the workshop begins. Our AI-injected process frees up our team for more of the high-value work our clients expect from us.

So, when we get in the room, we’re undeniably more prepared with AI than without it — more prepared to make that discovery workshop magic that only humans can create.

Follow along

To read our series about injecting AI into software development, join our mailing list and follow us on LinkedIn and X.

About ArcTouch

ArcTouch has been creating lovable apps for companies of all sizes since the dawn of the App Store. Contact us to learn more about our AI-infused digital product design and development services.