Anthropic’s top executives are warning companies against cutting back on their use of AI as costs increase.
“As we try to spend a little bit of time talking to our users, our top priority is really that the last thing they want is for them to stop using AI. That’s kind of the wrong thing to do,” Angela Jiang, head of product at Claude Platform, recently said on Sequoia Capital’s “Training Data” podcast. “And some of our customers actually do that.”
Katelyn Lesse, head of platform engineering at Anthropic, said focusing on cost is part of the “normal, natural cycle for businesses” when considering the best way to deploy AI.
“What becomes dangerous is when you say, ‘Here’s the hat, but I’m trapped inside the hat,'” said Lessy, who was interviewed with Jiang.
Zhang said Anthropic has often noticed a “spike” in AI spending at companies where employees are procuring Anthropic’s AI models themselves through “some kind of shadow IT.” He said companies can find ways to use AI more efficiently, rather than using less.
“What we try to encourage our customers is that we don’t want to stop innovating,” she says. “If we can get returns on top of that and be able to ship faster and operate more efficiently, those are benefits.”
Lesse said the aim is to “encourage innovation” while understanding different ways to achieve desired results.
“One is like running Opus all night and doing something crazy,” she said. “And the other thing might be to be a little bit smarter about the strategies you put together to produce the same outcomes at a lower cost. And I think that’s the next layer of thinking that everyone is going to start doing.”
AI companies face an increasingly skeptical corporate America that AI-related fees will rise without what some executives have said is a decent ROI to justify the spending. In response, AI companies like Anthropic have emphasized the cost efficiency of their models and services, which allow AI to be better tailored to the needs of specific companies.
Cost concerns could weigh on the overall AI market as companies like Anthropic approach long-awaited IPOs.
A new kind of router.
Companies like Vercel are seizing on this cost-conscious moment by providing customers with a way to route their AI usage to the best model suited to the task. Analysts say that as long as the cost of AI tokens remains high, demand for routing requests will remain high.
Jiang said a router “inside Claude space” makes sense for Anthropic.
“I think what we feel really strongly about on the model routing side is that we’re kind of designing a platform for Claude. We want to make sure that Claude is good at solving all of these things,” she said.
In the meantime, companies will continue to compete for position rather than price.
OpenAI CEO Sam Altman has made that strategy clear since the company released a series of new advanced models under the GPT-5.6 banner to compete with Anthropic’s Fable 5.
“GPT-5.6 sol is often half the price and about twice as token efficient as Fable to accomplish the same task,” Altman wrote in Tuesday’s X. “We are happy to be able to offer it at a quarter of the price.”