AI (Probably) Won't Solve Your Problems

Jamming AI into your product won't make it good. AI can't substitute for legitimate product development.

I’m not a doomer, decelerationist, or a luddite…but have you noticed software products you use now have “AI” or use that little magic icon? My financial portfolio, LinkedIn, and productivity suites are all now ‘enhanced’ with AI. The problem is, as a user, I usually can’t tell the difference and I doubt the added user value justifies the costs in many cases. I suspect we’ll see many of these features quietly deprecated.

The SISP trap

If you’re a product manager, founder, or tech executive and you’ve asked yourself “How can I use AI to enhance my product?” then you should ask your doctor about SISP.

SISP is a Solution In Search of a Problem or an attractive, shiny, new gadget that you’re not **totally/kinda/even a little bit** sure how it helps but you want to use it. As technologists, devs, gadget nerds, and (when something is really trendy) executives, it’s tempting to want to use the latest and greatest tools. This is for a good reason: you don’t want to be seen as outdated in front of your boss, customers, or leadership.

Put plainly, SISPs are actually unnecessary features that don’t solve problems for users. Or even worse, SISPs are wholly unnecessary products (see this batch of YC companies for many examples).

There are real risks to the SISP trap:

• You waste development resources on features that don’t add value for users

• Adding unnecessary features confuses the user experience and can slow adoption among new users

• You confuse users about your brand, differentiation, and value proposition

• AI will cost significant money to use at scale

• Bad AI experiences feel like bad chatbots and dilute your brand equity

How to avoid the SISP trap

First, allow yourself to explore new tools without forcing a use case. Just experiment internally with your team or find a side project. Just because you found a cool hammer does not mean everything is a nail.

In product management, you always start with problems and try to be agnostic about which solution you use to solve them. If a basic solution works, don’t make it any more complicated.

For example, let’s pretend you’re the product manager for a CRM tool and you notice users don’t engage with the calendar feature you spent months advocating for. Resist the temptation to double down and create an ‘AI-powered calendar’ just because 2 of your competitors are launching them.

The approach is to ask ‘Was I wrong about adding this feature?’ and then challenge your assumptions. Set up user interviews and testing. Re-factor the design and UX. Etc.

Where should I be using AI?

If you have a specific problem that AI is especially good at solving, it might make sense. So what is AI especially good at?

• Routine, repetitive chat-based interactions like customer support requests

• Summarizing dense text, video, or audio

• Generating certain types of imagery

• Reframing language (ex: a prompt like “switch this text to third person POV”)

• B+ schoolwork

• Pattern recognition in large data sets

• Customizing generic experiences to user archetypes

This list is expanding but that doesn’t make it a silver bullet for all conceivable products and features.

None of this is to say you shouldn’t keep up with new developments. Excitement and curiosity around new tech are useful. Use it to fuel internal experimentation and exploration but just wait for legitimate problems to arise.