Lots has been happening in the world of AI chat bots, not all of which has been particularly savory. It looks like we might be on the very precipice of AI-powered chat bots diving head-on into the trap of optimizing for user engagement — which is rarely ever a good idea.
So let’s take a look: AI chatbots apparently getting very good at convincing people they should change their mind (even though the study is widely accepted to be unethical, so: take the results with a big grain of salt). Meta reportedly has been playing it loose when it comes to adding the necessary guard rails to their chat bots if that might restrict engagement even a little. And, as so often, reportedly shows a blatant disregard for privacy when it comes to their ability to sell more ad space. Over on Hard Fork, Casey Newton and Kevin Roose just talked about this, too, and it’s very much worth a listen.
A while back, I considered gen AI & chat bot business models:
Generative AI’s potential business models
- Subscription models for generative AI tools like Midjourney and ChatGPT are pretty straight forward. They are, however, estimated not to be profitable. There might be some areas where professional tools might be possible to run at a profit, say a copy-writing tool for agencies or the like: Enterprise pricing might work, and be not too problematic. It’s a niche. Possibly a large niche, but still.
- Integration of AI tools into existing tool suits like Microsoft Office or Google Docs might pay some licensing fees, but it’s entirely unclear if this might be a sustainable business. I doubt it, the big companies will just absorb the tools at some points (or in the case of Microsoft, they co-own OpenAI anyway.)
- Take social/companion chats, one of the hot areas of applying generative AI. Besides potentially being a bit sleazy, these services would most likely make money from subscriptions, which requires them to keep users engaged. Engagement-based business models online were the root of many of the issues that got us into trouble with social media. So did running targeted ads against the contents discussed on the platform. So we might be committing the same mistakes we did with social media platforms all over again. (Let’s not.)
Number 1 and 2 have come to fruition: There’s subscription-based gen AI services (even though they tend to run at a loss, but we’ll get to that), and of course all the usual office products and apps have AI sprinkled on them generously.
For number 3 — companion chat bots — the jury is still out. The news I shared at the top indicate to me that the space is heating up, and not in a good way.
Most generative AI services lose money because they are simply very expensive to run, both in terms of upfront investments and in the operating costs: You need very expensive model training and server infrastructure to get going, and every single user request is expensive to run, too because of compute time, hardware depreciation, energy and water costs, etc. According to their CEO Sam Altman, OpenAI loses money even on their $200-per-month subscription.
As a rule of thumb, if you lose more money the more your users want to use your product, you’re in trouble.
There are some obvious pathways out of this situation:
- Users use the tools less. The path there is to make the product worse so users grow frustrated. This is unlikely to be the strategic decision that OpenAI and others will make.
- Running AI services will get cheaper. This is likely. We’ve seen big steps in that direction with China’s DeepSeek, which appears to run much more cheaply. Alas, usually cheap and available tools tend to also be used more the cheaper and more available they become, so this might not solve the issues, but who knows: We’ll see when we get there. This does not apply today.
- Introduce new revenue sources, and/or lower the costs and quality of existing services. This is where I personally see things heading in the short term: As more AI companies need to demonstrate profitability, AI-powered chat bots — the social, companionship or romantic type of chat bot — are likely to become a focal point to drive revenue. They’re comparatively simple and cheap to run, and it’s an obvious pathway to create more ad inventory to sell.
You’ll have noticed that I quite unsubtly pointed towards scenario 3: I’m fairly confident we’ll see a whole flood of chat bot offerings that will be optimized for engagement, i.e. trying hard just to keep users in the chat, and then to run ads against this, aggressively. (Cough cough, enshittification.)
And it’s frankly the worst.
Meta has long since announced replacing human users’ posts with bot-generated posts more and more, and is leaning heavily into chat bots. As the reports above show pretty clearly, Zuckerberg seems to push hard in favor of user engagement. At Meta, the old adage about the fish smelling from the head truly rings true. I see little reason to believe that the other big companies will choose differently (with the possible exception of Apple, which tries to position itself very in another spot altogether).
So, we might be entering the era of Engagement AI. Take care out there.
