Dan Shipper says he doesn’t write many of his routine emails.
Instead, the CEO of Every — an AI-focused media, software, and consulting company — said he uses OpenAI’s Codex to read his inbox, check his calendar, propose meeting times, and draft responses.
The tool is not allowed to send emails without his approval, but Shipper said much of the scheduling correspondence around his Business Insider interview was generated by AI.
“All the words are pretty much Codex,” Shipper said. He’s considering changing the “From:” field in his email to indicate when messages are drafted with his AI agent’s help.
His New York City-based business is built around AI. Shipper cofounded the company in 2020 with backing from Starting Line and Reid Hoffman. It writes its own technology newsletter, sells AI consulting and training services, and builds apps like Sparkle, a computer-file automation tool.
The firm also gets early access to models and tools from major AI companies, including OpenAI and Anthropic, giving Shipper a close-up view of where the tech is heading.
That perspective has made him optimistic about how people will work in the future. And he thinks his business is structured like a company of the future.
That optimism is expensive.
Every’s AI expenses
Extremely smart take on the tokenmaxxing panic and why it won’t last: https://t.co/Su1yOhAuL7
— Dan Shipper 📧 (@danshipper) May 29, 2026
Shipper told Business Insider he spent about $13,000 on his personal Codex overages last month, one of the highest AI bills he could recall.
“I got some side eye from our COO, Brandon Gell, on that one,” he said.
Asked how much larger that bill was compared with the same month last year, Shipper responded: “Way more. Way, way, way, way, way, way, way, more.”
At Every, AI access is just part of the cost of employing people. All 27 full-time employees get the entry-level $20-a-month subscriptions, while technical workers get $200-a-month plans, and the company pays overages. Token budgets are akin to other employee costs, like health insurance or company-issued laptops, Shipper said.
Even with the high costs, nobody at Every has been told they’re spending too much on tokens, he added.
“We’ve started to figure out how we will think about overages,” he said. “As long as you don’t spend so much money that we go bankrupt, we’ll be fine.”
AI changed how the company works
Every’s spending shows how deeply embedded AI has become in its workflow. Shipper said the company once experimented with giving every employee their own AI agent, but moved away from that model because the agents required too much upkeep.
Instead, Every now uses a smaller number of agents that serve specific teams or the whole company. One of them, called Claudie, helps the consulting team draft initial slide decks, put together sales proposals, and track client to-dos by reading transcripts and updating the company’s task manager.
Shipper said he does not see those tools as replacing Every’s workers outright. Instead, he said that AI could push more people into manager-like roles earlier in their careers.
“Very few people actually get the opportunity to be managers, because managing humans is very expensive and risky,” he said. “I think many, many more people are capable of that than we think.”
Still, Shipper said AI has limits. As a writer, he said, the technology is useful for research, drafting, and editing, but it is not especially good at knowing what is interesting.
That is where he thinks humans still matter most: not in doing every task by hand, but in knowing what is worth doing.
Predictions for AI’s future
Right now, Anthropic is all the rage.
The AI startup just launched its well-received update to its Opus model, leapfrogged OpenAI as the nation’s most valuable AI company, and is racing toward its IPO.
Shipper is betting that OpenAI’s Codex is better positioned than many people realize. He said Anthropic has “a ton of momentum,” but predicted the narrative around OpenAI could shift over the next few months.
“What they’re doing with Codex is incredibly impressive,” he said.
Also, for anyone looking for help with their prompting skills, Shipper said he’s still nice to the robots just in case they — you know — accidentally start to dominate us humans.
“I’m usually very nice, because you never know when they’re going to take over the world,” he said with a laugh.
Read the original article on Business Insider
The post An AI CEO explains how much he spent on Codex last month — and why he’s still ‘very nice’ when prompting the tech appeared first on Business Insider.
