AI is all the rage these days and for good reason. Investment on AI is at an all time high. Like the dot.com boom of the 90s, it’s hard not to turn a corner and see yet another companyname.ai launching with “the latest must-have feature!”
But sometimes, as you look under the covers a bit more, you see that a large number of these companies are just wrappers around cloud AI providers like ChatGPT, Anthropic, and more with a tiny amount of “unique” features. Some of these are a collection of pre-canned prompts that are fed into the cloud AI providers to generate the desired output.
The fundamental problem with these types of businesses is that they are parasitic. They only exist based on the cloud AI provider. The pre-canned prompts aren’t particularly defensible and are relatively easy to replicate. If the cloud AI provider finds enough demand, they could easily offer those canned prompts themselves and wipe out the entire reason this other company exists.
This isn’t a new business approach with AI. There are many examples of companies that just wrap the APIs of another company and provide a thin layer of insights on top of them. Think of companies working to analyze your cloud spend as an example.
But with AI, because generating prompts and getting output is so easy, the effort to produce them is very low. A business user could put this together in a few hours or a weekend & think, “Wow, I’ve just built something amazing!” And because every investor has FOMO, they may even get some funding.
But there are two very important issues that make this business approach particularly vulnerable today.
- Costs today are subsidized
- Your product value is limited
Most cloud AI providers today are losing money… well, hemorrhaging money is probably more accurate. Low cost subscriptions provide an all you can eat approach for users (including parasitic AI companies) that don’t reflect their true costs. Investors have put in billions if not trillions and one would have to be insane to believe they don’t want to see a return on that investment. So, those subscriptions will either raise significantly in costs or users will be forced onto an API model where they pay directly for usage or customers will be rate limited.
If you’ve built your business on the assumption that AI will always be cheap, you’ll be in for a rude awakening when the prices climb. As your costs climb, you have to charge your customers more. While they might have been willing to pay $20/month for the feature you had, is it realistic to think they’d be willing to pay $1000/month?
This leads us into the second issue. Prices are based on the value received. Most people using cloud AI today are experimenting. They aren’t yet seeing ROI even at the insanely low subscription prices, let alone if they had to pay what it actually costs to deliver the inferences. If it was so easy for you to build your companyname.ai business, how much value do you think it is actually providing to your customers? While they might be willing to deal with the errors from hallucinations or having inconsistent results from each agent run when it was $20/month, will that still hold when prices reflect your actual costs?
Building true value for your customers isn’t something that comes easy, despite what we are told from all the AI hype. You need to understand the challenges your customers face and come up with a valuable solution to them. Price it where it is better than the approach they do today & you can win them over. The more defensible your solution is… i.e., not easy to replicate with different prompts with cloud AI & you are set up for longer term success. This is why data matters. If you own the data, it is much harder to replicate your business. If you build your own models using your data, rather than being diluted by a commercial LLM attempting to be everything to everybody, you can make it cost effective too.
The companies that will win are those that can use AI effectively, both in terms of the value delivered and the costs of delivering it. If you’re a parasite… your outlook doesn’t look good if you don’t make some changes.
Need help ensuring your company is on an effective path to using AI? I’m available for fractional consultations.