Drop-In Class #17: GenAI ROI

Who is using GenAI in production and making money?

Feb 13, 2025

Welcome to my newsletter, which I call Drop-In Class because each edition is a short, fun intro class for technology concepts. Except unlike many instructors, I'm not an expert yet: I'm learning everything at the same time you are. Thanks for following along with me as I "learn in public"!

Why ROI matters right now

For a while there, ROI for GenAI use cases didn’t really matter. But now we’re at a point where we’ve seen way too many chatbot demos (my own demos included). And predictions like “could add trillions of dollars in value to the global economy.”

So the question is: When do the demos make it to the real world? Where are the trillions going to come from? Who is going to be able to monetize AI-powered products?

I’m not just talking time saved or internal productivity gains. I’m saying, who is going to be able to charge for a product because it has GenAI capabilities? And what is the pricing model going to look like?

I looked into which companies are starting to get their consumer-facing AI use cases moving, and how they might start monetizing AI-powered capabilities. Let’s put on this week’s music!

What are the current GenAI use cases in customer-facing apps?

There are really two ways you can think about ROI: either getting more efficient, or making more money.

Internal efficiency gains are tricky to measure. Worker productivity is obviously a major use case for GenAI, but there’s not a clear payoff. There are mixed reviews on whether AI copilots are worth the investment, for instance.

And companies can’t keep losing money on endless prototypes for internal employees. They need to get use cases off the ground. As one of my favorite tech reporters Tom Krazit pointed out, “Right now, cloud companies are making lots of money from selling the picks and shovels needed to make AI prototypes while customers struggle to get across the finish line.”

So I’m more interested in customer-facing apps right now. How are businesses actually incorporating AI in production? Who is making it across the finish line? Here are a few early success stories:

Text and image generation for design: Canva and Adobe are charging for premium subscriptions to get full access to AI functionality, like text-to-image generation. Wayfair launched an AI-powered home design tool called Muse. And entire startups like Arcade AI (AI-powered product design) and Prezent (AI-generated slides) are built on GenAI functionality. I think there’s a lot of potential to monetize here.
Customer support: Yes, the chatbot! Enterprises like Best Buy, ADT, Alaska Airlines, Six Flags, IHG Hotels & Resorts and ING Bank have implemented LLM-powered customer assistants. But it’s not like their apps charge extra for using the assistant (yet). They need to prove ROI in other ways, like showing that customers made more purchases due to better experiences, or that they could accomplish more with fewer customer support resources.
Personalized data insights: We know about the power of GenAI for internal BI applications, but what about making data insights more accessible for end users? I’m seeing fitness apps start doing this. Strava recently launched Athlete Intelligence, which analyzes your workout data and pops out summaries. So far I find it amusing. Sometimes it’s helpful, and I learn something about my runs. Often it’s a little too nice to me — it never tells me I had a bad run. But maybe that’s a good thing.

BTW, I was very proud of this Strava art. I was randomly running around a cul-de-sac neighborhood and this was the result. Somehow it did not go viral.

So all these AI-powered capabilities have finally made their way into production, customer-facing apps. The models are powering stuff out in the wild. But how do you charge for it?

How will AI-powered offerings be priced?

Can you really add an AI capability to your app and expect it to move the needle for customers?

Will people be willing to pay more for access to a conversation with an LLM-powered customer support agent? If someone is considering canceling their Strava subscription, will Athlete Intelligence be enough of a differentiator to convince them to stick around?

Now that companies are getting their use cases in production, we’re going to find out.

As a product marketer, I’m always curious about pricing and packaging. And GenAI is a huge pricing challenge, with so many unknowns. Box CEO Aaron Levie had a few different ideas for pricing models:

Hours of work: Just like a human charges, except it’s AI!
Per-outcome basis: Charge by thing accomplished.
Price it close to the underlying compute costs: Seems obvious, but then you’ve got a smaller margin.
Seat subscriptions: Traditional software approach — charge per user.

There’s also consumption-based or usage-based pricing, which has been the go-to model for SaaS (e.g. public cloud infrastructure). For instance, Salesforce is pricing its AI agent $2 per conversation. And OpenAI is considering usage-based pricing for ChatGPT. But it’s unpredictable for both customers and the businesses.

So no one really knows yet what the best approach is. It’s interesting to consider the Wild-West options that are still in front of us.

My hope is everyone else’s hope — that all the investment into the public cloud providers and LLM vendors building the AI “shovels” will ultimately turn into AI-powered consumer products that make gobs of money. And it’s looking like we’re finally about to start figuring out how.

If it’s any indication, I actually think I’d be kind of sad if Strava’s Athlete Intelligence went away. I’d miss it telling me that my lame runs are still an impressive effort. Everyone needs a little bit of that in their lives.

The cooldown: Extra reading

This is a 101-level drop-in, so I keep things at a high level! For deeper reading, check out these articles.

See you in the next drop-in!

Cheers,

Alex

Drop-In Class