AI has its place in product development, but are we offloading work or offloading thinking?
AI as a tool, used correctly, is useful. If it’s helping you reorganize what you thought, helping you edit in your job or your development work, or just helping you get organized, it has a place and really has changed a lot of the way we work.
But here’s the thing I think we’re forgetting – using AI is just the offloading of work, and the trick is understanding what work we should be doing as humans.
What is the work that actually helps us make sure we’re understanding both our supply side and our demand side, and that we’re not handing off to a machine that’s aggregating words and taking things at face value?
What gets said without being said
Take customer interviews as an example.
If you’re using AI to run interviews, or you’re doing the interview yourself and using AI to capture everything and then taking it at face value, I would caution against that, because there’s a lot said without being said.
There’s a lot said in the way someone uses a word, not just the word they use. AI might get good at that one day, and it might be good at some points, but the human ear still matters.
Conducting these interviews ourselves, asking the whys and the whats in a meaningful way, getting people to explain, and feeding it back to them so they hear it from a person instead of an emotionless bot on the other end is where the impact and meaning come from.
AI doesn’t have discernment, and it doesn’t understand a lot of the nuances. It can give you facts and aggregate a range of information. For example, it could tell me that 15% of people said they wanted it to be cool, or safe, or tasty, but what do those words really mean? They could mean different things to different people, and we need to unpack them rather than take them at face value.
AI is most confident voice in the room
Another challenge we can have with AI is that it becomes overly confident too quickly.
If we know anything from product development, it’s that being humble and staying humble matters, whether that’s knowing you don’t have all the answers, conveying that to your team, or working together to reach a consensus instead of defaulting to the loudest, most confident voice in the room.
For a lot of people, that’s exactly what AI becomes.
Without giving it context, it’s still able to become confident well before it had the evidence.
When do you lose yourself?
When people say they use AI to sharpen their thinking, I often get worried, because I wonder whether it’s really their thinking or other people’s thinking. When does it lose you, and when do you lose yourself to AI?
As humans, we often want to delegate the things that are hardest. Personally, I believe we should be doing the hard things and letting other things handle the easy ones. If I lose the ability to think, to talk to customers and figure out where their words means, and to work with them to understand the trade-offs they’re making, I’ve lost something I don’t want to lose.
As product developers and leaders, we need to think hard about what we’re offloading and why.
Are we offloading something because it’s a time sucker, or because it’s hard and we don’t want to think that hard? If it’s the second, we need to take a good long look in the mirror.
A weekly report that doesn’t change much and just needs someone to check the right line is on the right line is a good thing to offload, but thinking about how people might react to something, or the different ways someone might learn something, deserves our time and our working through those nuances together as humans.
Counting all the costs
People also think AI is cheaper, and there is a dollar cost where it might be cheaper than hiring two or three people, which I totally get, because companies need to make those decisions.
But there’s also the cost of not understanding or learning something, and of what that overconfidence might cost you when you launch your product and it doesn’t hit.
Humans aren’t always great at that either, and our failure rate is astronomical, but you should know there’s a cost.
There’s also a cost to institutional learning. Unless you want to offload everything, you still need knowledge in your people, because you can’t have empty-vessel people alongside a full-vessel computer that could hallucinate and that you have to keep going back to with questions.
So are you really saving money, or is this just a cost shift? AI has its place, so offload the work, but keep the thinking.