Unlocking spikes and latent needs - consumer AI!

My goosebump moment with AI came with Deep Research. Being able to find, comprehend, synthesize the entirety of literature on a subject within 5 mins which typically would have taken 5+ hours was magical.

But the moment you know a domain even slightly well, the cracks become obvious. It'll base strong hypothesis on sources that are fluff articles at best, most numbers are misquoted, nuances behind being an expert is MIA. It feels like an over-eager intern: energetic, occasionally brilliant, but not someone you would trust with decision making.

Yet, I find myself addicted to it. I do multiple searches daily, even for areas that I have no relation with. Essence here is that its role has shifted from a decision maker or smartest colleague to someone who can help you ask the right questions, which eventually aids in decision making. This alone compresses a multi-hour ramp-up into minutes.

Had a similar experience with Granola - captures 50-60% of the essence of meeting accurately but that's 100x better than relying on memory/ not taking any notes at all

Which makes me realize two things about consumer AI:

1) Spikes overpower imperfections: Initial versions of AI products will be highly imperfect. But they'll be so good in one thing (that actually matters) that it becomes too hard to ignore them. For Deep research, it's incredible productivity gain (5 mins vs 5-6 hours). For Ash, it's memory/ user context and high EQ. For Suno/Runway, it's suddenly having the capability of being a creator. For Granola, it's enabling the user to be truly present in a meeting. When the spike is powerful enough, trust catches up.

We saw the same movie with Google 2001 (irrelevant ads, miraculous recall), Swiggy 2015 (questionable ETAs, magical convenience), and early Uber (surge complaints, but cabs actually showed up).

2) Latent needs unlock invisible markets: The thing about truly innovative products is that they capture a latent need that users never knew existed. When friction vaporizes, usage exponentially increases, not just shifts. So it isn't anymore about market share capture but new market creation.

Deep Research eases the process of being an expert, so you end up wanting to learn more. It's not that curiosity never existed, but it's more that the curiosity mellowed down as it had to go through high friction in pre AI world.

We've seen this before - FB/IG fulfilled the need of knowing tiny details in your friends' lives (at least 10 times a day). Before YT, we didn't know so many people wanted to be creators. Swiggy/ Zomato didn't end up replacing the dineout needs but rather increased occasions of having non-home cooked meals

Breakout consumer winners will be those that increase total manifestations of a behaviour by making it absurdly easy. This is esp true in Consumer AI as a lot of use cases will be net new. Question will be what latent need is being answered and how strong that truly is.