clever mascot

AI for Product Managers – What I’ve learned

•

Oliver Happy

So, we’ve AI on the mind, along with all the other challenges of our roles. As someone who now spends more time helping teams adapt and grow than managing my own backlog (and has gone deep into what is possible), I thought I’d share what I’ve learned about AI for Product folks so far – I hope it helps you on your own AI journey.

My AI adoption curve.

As with any new tech, I’ve gone through an adoption curve. But this time is different. Normally I’m first to adopt, play and share. This time I resisted. Why? It felt a bit like cheating to ask Perplexity or Lennybot Product questions I felt I should know the answers to. And just quietly, I also had a feeling that tech, unlike other tech before it might one day replace me.

Here’s what the stages looked like for me:

  • 90s – I couldn’t imagine AI
  • 00s – Dreamed of smarter search, platforms who remembered my preferences etc.
  • 10s – Loved recommendation and forecasting engines
  • 2023 – GenAI arrives. Stages of grief – Denial, fear (see Skynet), avoidance, depression then finally – acceptance
  • 2024 – Adopted it. Up the hype curve I went, then down into the trough of disillusionment when GenAI provided dud answers.
  • Now – Embraced it – I’ve tested and learned, now am on my way up the Slope of enlightenment, hoping to reach the Plateau of productivity.

What I’ve learned

  • Resistance is normal (yet futile). AI feels big and scary. It’s OK to explore with trepidation, that’s sensible. And yes, I am feeling the sci-fi blockbuster film references today – even Star Trek gets a mention.
  • Proper search has arrived, and so Product research got easier. AI-powered search, Leo in Brave, Perplexity, Gemini (try Advanced), Lennybot for technology, process, customer and competitor research.
  • AI is not actually AI (but no one is listening) – This current stage of ‘Narrow AI’ is just a very visible step forward in a long process of machine learning advancement. It doesn’t think for itself yet, but it is good at drawing connections. Just know in your heart that Marketing teams got a bit over excited in calling it AI, while nodding and smiling at execs who call it such.
  • It’s data time (finally). You’ve told interviewers you are good with data for years, while wishing enough data was available to you to really make this true. Well, you are finally getting that data you needed. Just remember – Garbage in, garbage out. So when execs get excited about AI, take them by the hand and have a little chat with your Data team about the data you need and how to get it.
  • Look at the pretty patterns. Pattern recognition and recommendations are now finally in reach for you, your Platform, Engineering, Sales and Operations teams. And it turns out this is most of the value in your data – previously hidden from you and the business. Explore here first, then look at GenAI options.
  • Don’t believe the hype. GenAI is already useful for search and summarisation and some content generation, but in real world application it often needs moderation/disclaimers, can hallucinate and is not quite as useful as the hype promises.
  • Product owns AI Strategy. As we learned at MTP Pendomonium 2024, we are seeing a global rise of the CPO (10x role growth in Fortune 100 alone). With C-suite membership, we’re also seeing all roads lead to Product for forming AI strategy. Product logic is helpful when applied to both internal and external use cases, so you can expect to see CPOs driving AI strategy in up to half the cases according to my event notes.
  • What, internal users? Turns out PMs have internal users too – Ops, Marketing, Sales. And AI is good enough for them already. In fact, they are already experimenting with it and could do with your help and your logical Product approach to solving their problems.

Where this is headed

The bad news is that AI is coming for your job – but only the boring bits. Your role is changing, but in good ways. It may feel like much of your daily life is being automated but if you step back a moment, I reckon you’ll find it’s the monotonous part.

  • Less Admin. This means less admin, and more time for strategy, discovery and with your people (you know, those who build and buy your product)
  • Product activities are getting automated – This is truly worth getting excited about. Your tools are getting better. Discovery docs, PRDs, tickets and bugs no longer need hours of writing. You now have templates!
  • You have new tools. A new generation of Product Management Platforms is arriving – Craft.ai, Fibery, Click-up. And existing tooling is getting an upgrade – Jira, ProductBoard, Aha! et al. This all adds up to easier planning, discovery and daily backlog and board management. Hallelueyah!
  • Other Admin is getting automated too – Marketing, Sales, Ops, Finance. They will want your help applying GenAI to their processes.
  • Take my money now – Leadership teams are willing to invest, but need educating on problems worth solving, ROI and an iterative approach to exploring the value of AI

What are people trying out?

Experimentation is reaching a frenzied level and can be grouped as follows:

Some interesting external use cases:

  1. Self-service on web or in-product (chatbots)
  2. Search and summaries (GenAI)
  3. Document summaries and suggestions (GenAI)
  4. Analytics in-product (Machine learning)
  5. Loyalty programmes (Machine learning + GenAI)

And some internal use cases:

  1. Marketing and content creation (GenAI)
  2. Search in SharePoint (Microsoft CoPilot)
  3. Coding (GitHub CoPilot)
  4. General research (GenAI)
  5. Product Discovery (Craft.ai, Productboard, Jira with AI)

Some tips for you

So what can you do with this new information? How do you forge your own path through the field of unplowed AI possibility?

  1. Become an internal problem solver. You know those internal reporting processes you love to hate? Well now perhaps is the time to suggest how they could be automated or supported by GenAI or new tooling. SharePoint automations are getting really handy. Just as you do for user problems, but for the internal processes all around you.
  2. Do try this at home – This is a great chance to experience being a user first, then build the features. Notice what you like and don’t like as a user. Start with web search via Leo in Brave, Perplexity or Gemini. Get good at asking questions and follow up questions.
  3. Get good at asking questions. Research just got easier, so do it. On tech, tooling, operating models and competitors.
  4. Seek first the problems. Put the hammer down and look for loose planks, then pick up some nails. Find the first problem worth solving.
  5. Start with what you know – Take your problem worth solving, then seek first the data you already have in order to answer them. This may mean you are gonna need a bigger boat (Data team) to fish for this particular shark and that’s OK. If a desire to benefit from AI-fuelled efficiencies results in investment in data first, all the better.
  6. Don’t get overwhelmed, take it step by step. Be careful where you look and with what you read. It is easy to be overwhelmed by big info and be ‘AI washed’ by slick Marketing pitches. Look to what Product folks are doing and reading. Start there.
  7. Thank them for calling – Be available to talk about ‘what we could do with AI’. Thanks Leadership for calling, add that ‘of course we are looking into AI’ – then be prepared to share the state of the tech, the guiderails you need implemented and a summary of some experiments you are running. Help them look beyond the hype and to understand the data-driven ethos.
  8. Don’t go too big to soon. Public failures in AI rollout stem from a big bang approach, no different to any delivery. Optimise existing features and processes using Machine Learning (pattern recognition and recommendations) first, then optimise some more using GenAI for search and summarisation. Then and only then, look at GenAI for content generation. Unless you run a Marketing or Content platform, in which case this is becoming your core feature already.
  9. Build in stages. Like anything, build and ship in the smallest valuable pieces you can. Test internally and test often to minimise the impact of AI ‘failing silently’ with hallucinations and providing dud answers to trusting users.

Conclusion:

This might be the most exciting tech advancement you see in your life. And it is because AI promises so much that the Product works is avidly exploring it both in daily tooling and for their own offerings.

But thus changes nothing about your approach, so stick to what you know and look first for problems to solve, explore options with your Tech leads, and experiment iteratively. Start by looking at the data you need to extract value via machine learning and/or GenAI and take a moment to consider governance and privacy along the way.

And remember – AI feels big and scary at first. It’s OK to hesitate, but it is important to not get left behind.

Fare thee well riding this latest hype curve!

•