As generative AI finds its footing in enterprise products, how will companies charge for its use?

@alex / 9:30 am PDT • May 4, 2023

Image Credits: Getty Images

Enterprise software companies aren’t wasting any time integrating generative AI into their products. Despite the relatively recent explosion in interest around large language models built by companies like OpenAI and Anthropic, tech companies big and small are charging into baking the technology into their products and services.

You’ve seen the headlines by now: Slack is working on AI tooling; Anthropic is building a version of its own LLM for use inside the corporate chat service; Box is working with generative AI tooling, and so is Ada. And Microsoft’s Bing just ripped the waitlist off its chat product. Corporate excitement for this new system of user-software interaction and user-directed creativity is widespread, and more examples are cropping up.

On Wednesday, HubSpot CEO Yamini Rangan told investors during the company’s earnings call that generative AI is going to “rapidly” change the business landscape.

Rangan detailed why HubSpot would prove a useful hub for generative AI to accelerate work, arguing that it has “unique data and broad distribution” while residing in the “center of [its] customers’ workflows.”

Nothing that HubSpot said about generative AI and how it expects customers to use the stuff was surprising. What did catch our eye, however, was the company’s discussion of how it plans to monetize the new tools.

During the call, an analyst asked when the investing public will be able to “see the benefits of AI in your financial results,” per a transcript of the conversation.

Here’s the key portion of Rangan’s response:

Some [generative AI] features are just going to become table stakes. They will just be part of our core products and part of how we drive customer engagement and adoption within the core product. Now, if there are more specialized use cases, then maybe it goes into higher value additions like co and enterprise, and will increase our ASP longer term, and that’s how we’re thinking about monetization.

That’s a very interesting answer providing much food for thought:

HubSpot expects that generative AI will find sufficiently wide implementation inside enterprise software that some of its capabilities will become “table stakes.” Put another way, demand for LLMs is going to be massive and widespread enough to be anti-endemic. This perspective explains why venture capitalists are paying through the nose for a handful of shares in OpenAI.
Offering specialized generative AI tooling could become the way to monetize the new technology cleanly. This could lead to better net-dollar retention, for example.

There are a number of ways or plans to monetize generative AI tooling in the market today. HubSpot is seemingly adopting one approach. GitHub is taking another path by charging for its “Copilot” service (powered by OpenAI) directly.

GitHub charges individuals $10 per user per month and companies $19 per user per month. That’s simple enough, but it is quite costlier than the middle tier of GitHub’s core service (around $4 per month per user).

We aren’t saying that’s a bad thing. If GitHub’s Copilot saves the average developer a single hour’s work in a month, it more than pays for itself. But this is a more direct pricing mechanism than what we’re expecting from HubSpot.

These diverging approaches provide a useful way to think about how generative AI tools might find a path to revenue.

Here’s a loose outline of what we can see today:

Ad-supported: Bing and other search engines are going to use advertisements to monetize their generative AI-powered services. This will presumably work in the search market, as the ease of search engine optimization on a per-search basis will likely outstrip the compute costs required to deliver results.
Value-add via integration: HubSpot expects generative AI to be used throughout its product. Its CEO’s note about “table stakes” tooling implies that AI will become a component of existing software products. If it works as intended, this will make existing software better (more impactful) and therefore worth more.
Paid add-ons: The GitHub method. Products and services in which AI-powered features are separate from the core offering may be priced with an additional charge. It will be interesting to see where AI tools manage to separate from the core offering of an existing product or service in revenue terms and where they do not.

Note that the above methods are not necessarily exclusive: HubSpot could employ both the second and third approaches, for example.

There are many more monetization techniques that could find mainstream adoption. Artists could purchase credits to a particular generative AI model for adapting or extending their artwork under a type of pay-for-what-you-use model. And the companies building generative AI models will find their own ways to extract value, just as the underlying cloud platforms that they themselves leverage are happy to take their own cut of total generative AI GDP.

The lesson from HubSpot’s earnings is that while there are broad corporate aspirations to employ generative AI, the money side of the equation is less concrete. Startups harboring hopes of toppling incumbents appear to have a pretty wide field ahead of them when it comes to monetizing their own generative AI-powered products. May the best service and the method of selling with the least friction win.

As generative AI finds its footing in enterprise products, how will companies charge for its use?

Latest Stories

Google lays off workers, Tesla cans its Supercharger team and UnitedHealthcare reveals security lapses

Human composting and timber marketplaces: talking “industrial” VC with investor Dayna Grayson

Women in AI: Tara Chklovski is teaching the next generation of AI innovators

Farewell, dunks? Threads launches quote controls for all users

Deal Dive: Amae Health is building an in-person approach to mental healthcare in an increasingly digital space

Why RAG won’t solve generative AI’s hallucination problem