AI won't destroy your PM career (or save it)
You can just get popcorn and watch a livestream of the hype cycle play out.
To set some context, I’ve been building AI-based systems for a few years, but I’ve been reading sci-fi since I could read so I know all about the impending end of the world.
Or lack thereof.
I’ve also been around the software block half my life, always evaluating technology just enough to understand potential impacts, and have seen many a hype cycle play out. I’ve been dabbling with creative outlets both in a corporate setting and as a hobby, and appreciate the creative process. I’ve seen the use and limitation of AI in such endeavours.
This post is about how AI applies to the practice of product management, which has elements of creativity as well as craft, and why the current AI buzz will neither save you nor destroy your career.
What goes up🚀 Must come down📉
Artificial Intelligence has been around for a long while, as have been machine learning and other variations of the subject. So has generative AI — remember the big explosion we’ve been seeing since late last year following the introduction of ChatGPT? It was based on GPT 3.5. So while the public chat interface was new, the underlying models weren’t.
Still, this exploded into the public consciousness, with text, image, and other generative models that are seemingly able to do magic with limitless potential. Exciting stuff.
But then, of course, your news feed is suddenly filled with prompt engineering experts (last year’s Bitcoin bros) explaining how you can do everything — from constructing the perfect PRD to user research! — with ChatGPT. Or, alternatively, those who see the rise of the machines and the end of humanity. AI will make us all redundant as a first step before the robots fire the nukes to clear up some space.
I’ve only applied a pinch of literary license, I’m not exaggerating much. Being an author, I also see similar discussions around creative writing — from “AI will steal our food” to “How can you get AI to write your books”. Same caveats apply.
Both sides are victims of the hype. For those not familiar with it, Gartner has described this behaviour over 25 years ago, and termed it the Technology Hype Cycle. It goes through predictable stages, and looks like this:
There’s a trigger, followed by mass hysteria about the potential. These inflated expectations naturally crash down, followed by a calmer realisations of the potential and limitations of the technology, and a more stable plateau of productivity.
I think we’re around the peak now (probably just past) when it comes to generative AI. The good news? It almost completely cleaned my feed from the discussions on Crypto-currency and Web3 😜
Before settling into sanity
So let’s get back in to the central core of product management, in a (vain) attempt to get over the hype and predict where AI might be of use.
First, have we forgotten why we write anything?
90% of product management is a communication of some sort (and 80% of those communications should be listening, but I digress). It’s why soft skills are so important. The purpose of communication is the conveyance of ideas between people.
Your job as a PM is to gather context from users, help the company understand that context to build the right products in the right way, and then deliver those solution back to the user and reinforce the value-learning loop. Producing an artefact around that communication isn’t the end goal in itself.
This is why PRDs, or any sort of specification documentation, fails. It’s the desire to produce a perfect document and then lob it over the wall at the developers, without taking the time to communicate. Without taking the time to do the hard thing, to answer questions you haven’t fully thought through and face feedback you don’t want to hear.
Add to that how generative AI models have to be trained on existing data. How good is the input into the model? Paradoxically, a more accurate input (accurate for your specific needs) means narrower scope, which means the generation will be narrower. So you are running the risks not only of errors in the input and generated content, but also that anything created would be more of the same. And doing more of the same isn’t the path to innovation and breaking the mould.
So while AI may help in some cases (like analysing large volumes of customer feedback or calls), it’s not a replacement to actually sitting and chatting with customers users. There is an element of empathy building that is critical, that you won’t get via the degrees of separation. It’s the same as why dragging devs on rotation to customer interviews has a profound impact, more that any amount of removed discussion.
Furthermore, in the same way that a PM doing interviews and and gaining insights into their product doesn’t magically transfer their distilled knowledge to dev teams and executives, neither would AI magically do it. Clever formatting of knowledge into a magic document won’t help if people aren’t reading it, or taking away wildly different conclusions.
The whole point of the communication is the exchange of ideas, until we reach — and verify we reached — an alignment on the goals and next steps. It’s understanding that communications are two-way, that insights are personal and everyone has to reach them for themselves. Any document or tool is there to support and facilitate this function, not replace it. So whether you use AI tools or not isn’t the point — it’s whether it’s helping you reach the minds and hearts of your audience for an exchange (both ways!) of ideas.
And note that this doesn’t mean more meetings. You can have a writing culture in an organisation that is still centred on collaborative discussions (just asynchronously). The document with comments is a medium for the open exchange of ideas, and isn’t a dictation, something that is chucked over the wall as part of a ‘gateway’. No tool obviates the need for communication, no AI replaces talking to actual humans.
How does creative writing link to this? I write fantasy stories for the joy of creation. This is very different from someone who does creative writing or painting for a living, but it does give me insights into the both the creative process and the ‘transfer’ of the story from my mind into those of the readers. It’s my observation that the ‘story’ happens as much in the reader’s head as in the author’s. Whether it’s the reader who commented about how all my characters are blond (I counted four out of several dozens), or the conflicting impressions from two readers about the same character in the same scene (one though she was a cardboard cutout and lacked agency, and the other loved her spunk and vivacity).
When you’re using storytelling for business you have a distinct advantage over a fiction author. This isn’t a write-publish-wait-for-the-reviews process, but much more interactive. You can can (and should) use the story as the springboard for deeper understanding. Obviously depending on the context, but ask questions as much as you give answers, use briefing and back-briefing to ensure alignment, be transparent about your view and curious about the others’ — all in real time.
So what’s the ‘plateau’ for AI in product management?
Back to the hype cycle. Generative AI will undoubtedly have a great impact on our world, but that has been a trajectory around machine learning and AI technology for years. And as we learn from each step on the way, the goal of understanding what ‘intelligence' really is is as elusive as ever. This is just a bump of hyperactivity on this journey.
I have no doubt that generative AI might be able make sifting and analysing large volumes of data easier (and will improve over time). You could use bots as a slightly less cryptic rubber duck when you think through a problem. As developments come, there will be tools that increase productivity by automating tedium. But that’s not the point.
One, the tedious-but-necessary part was never he crux of any creative role.
Two, any increase in productivity doesn’t obviate the need to communicate with others.
So no tool, no matter how much it promises to increase your productivity, will eliminate the need to think for ourselves and then transfer those thoughts to other humans (and validate them with received thoughts).
Spreadsheets enabled people to perform faster analysis, but didn’t obviate the need for financial literacy. Photoshop gave creatives a new medium and tools, but didn’t eliminate the need to study art and color theory. And AI won’t remove our inherent need to create nor will it replace the need to share thoughts with other humans, no matter how good it gets at enabling us to solve some problems faster.
Regardless of how good AI gets at taking orders and building ‘stuff’, it will not translate a half-thought-through idea into a successful business (in the same way that ‘no-code’ tools didn’t replace programmers or business sense). You need to understand the users’ problems space, your first try will likely fail and you may have to pivot, and a single person just can’t tackle large wicked problems. For all those you need to interact with other humans, ie communicate and share ideas and then evolve those ideas based on multiple inputs. Having a good idea — whether in starting a business or writing a novel or building a product — has very little to do with eventual success. You won’t find magic answers by asking a genie (or AI). Success in those endeavours lies with focused execution, which require collaboration with other people, the sharing of ideas and aligning creative efforts.
So, first, understand what the crux of product management is. It’s why I harp on soft skills and mental models. Second, when evaluating these emerging technologies, ask yourself whether they help with facilitating the fundamental shared understanding between humans, or whether they are distracting with empty promises to eliminate necessary interactions. You should know what to choose and what to ignore.
Hope you enjoyed this post, with my attempt at a calmer, more reasonable tone in facing the hype. AI will have its uses, but will never replace the human creativity and the need for shared experiences.
All of this doesn’t touch on integrating AI into your actual product, a subject fraught with it’s own pitfalls and opportunities. Maybe next time 😉 (Though I urge you to read Rich Mironov’s AI Washing, and Cambridge Uni’s Dept of Comp Sci Oops, We Automated Bullshit.