Home Artificial Intelligence Principles for AI product design

Principles for AI product design

Principles for AI product design

Subscribe here to receive latest Angular blogs, data reports, and newsletters on to your inbox.

There’s a famous saying that customers don’t desire a drill, they desire a hole within the wall.

Given the out-of-the-box power and flexibility of LLMs, this aphorism has gained latest meaning. Might we be entering an era where most software will get much closer to producing the opening itself? In a world where AI can do a whole lot of what humans can do, how will software evolve?

To assist think through what products might appear like in our AI-powered future, it’s value recalling what’s one of the powerful and ubiquitous AI-powered products of our past: .

Back within the early 2000s, paid marketing was a wholly different beast than it’s today. Marketing teams spent a whole lot of their time determining what amount to bid. Go far enough back in Search Engine Land and also you’ll see what life was like.

Then, in late 2007, Google’s conversion optimizer launched. And in what could have been the primary job that artificial intelligence replaced, swiftly, paid marketers didn’t actually need to bid anymore. As an alternative, they refocused their efforts on two things:

First, data. The more relevant data marketers could pipe back into Google, the higher the conversion optimization worked. In the beginning, this meant sending back conversion data. Eventually, this began to incorporate product usage data as well (and anything that was seen to enhance performance).

Second, creative. The more creative iterations marketers could provide to the system, the more likely they were to seek out a latest winner and boost performance. Top performing marketing teams were creative-production machines, testing latest iterations on a weekly or day by day basis.

Yes, the conversion algorithm itself was a black box, but that didn’t stop paid marketers from reorienting themselves to do whatever they may to enhance its performance on the margins.

So, let’s say you’re constructing a product with AI at its core. What does the instance of Google’s conversion optimizer teach us about how you would possibly design it?

First, . In case you recall, Google launched conversion tracking long before launching conversion optimizer, so Google had all of the conversion data they needed to construct the initial model. Like Google, you would like some type of internal data to bootstrap the model at first. It often is the case that an out-of-the-box LLM is nice enough to get you began, but some type of proprietary data will likely be key to long-term differentiation.

Second, . Like Google enabling marketers to pipe back every form of “conversion” they wanted, it’s worthwhile to empower your users to enhance the model with external performance data as well. This has the extra advantage of making your product stickier. Not only will the model’s performance improve, it’s going to be a user’s job to enhance the model’s performance!

Third, . Like how marketers pivoted to extreme creative experimentation as a method of improving the model’s performance, take into consideration what additional variables your users can iterate through. By giving your users tools to enhance the model’s performance through experimentation, you’re creating users who will define themselves by how good they’re at extracting one of the best performance out of your system, improving stickiness even further (as we’ve learned prior to now, customer-built products create power users and evangelists.)

Given how versatile LLMs are, it’s tempting to assume a world where these models replace humans entirely…where they’re capable of simply produce a “hole within the wall.” But, just as with Google’s conversion optimizer, I feel the fact will probably be way more nuanced than that.



Please enter your comment!
Please enter your name here