I thought the AI boom was producing a new level of bad behavior, what with MongoDB CEO CJ Desai peacing out, not to mention Google’s earlier controversial “acquihire” of Windsurf’s executive team. But I was wrong: Money has a long history of hollowing out our ethical norms.
The problem in these and other AI moves isn’t really about the money, but rather about the trust, or lack thereof, left in the wake of the moves. My former MongoDB colleague Meghan Gill brought this into focus after Desai’s departure. In a post following her sharply critical tweet about the exit, she put the business case plainly: “A leader’s word is a business asset.” Otherwise stated, investing in employees and planning leadership transitions help build a company that people can trust.
That sounds almost embarrassingly obvious. Yet some responses to Gill’s thoughts treat integrity as an unreasonable expectation in a capitalist economy. Take the money, their logic says, because everybody else would.
Perhaps. But even if you don’t think ethics matter in some absolute sense, Gill’s point doesn’t require you to have Leave It to Beaver (look it up) sensibilities. After all, enterprise software companies succeed largely by fostering trust, which manifests in long time horizons for supported products and more. If AI companies want to cash in on the enterprise, they’re going to need to start acting in ways that earn trust, rather than burn it.
Leaving well is part of the job
MongoDB’s September 28 filing records Desai giving notice on September 24, with his resignation effective four days later, the day Meta announced him as its chief enterprise platform officer. The filing doesn’t tell us everything discussed privately, and neither announcement establishes that a bigger paycheck drove the decision. What it does establish is a remarkably short interval between the formal notice and the departure of the person running the company. A rank-and-file employee would normally give at least two weeks’ notice.
Compare that with Eric Schmidt’s move from Novell to Google in 2001. His Google employment agreement explicitly recognized that he needed to complete his duties as Novell’s CEO. He would initially serve as Google’s chairman and take on the full-time CEO position once his Novell employment ended. Novell subsequently named Jack Messman CEO, retaining Schmidt as chairman. Google’s offer accommodated Schmidt’s responsibilities to the company he was leaving. The handoff was part of the arrangement. It’s what you do when you don’t want to create $6 billion swings in a company’s stock price, as happened in the wake of Desai’s abrupt resignation, though MongoDB’s stock price has recovered somewhat since then.
I’m not saying an executive should be forced to stick with a company any more than an employee should. No matter how loyal you may be to your employer, the reality is that companies routinely cut employees and aren’t in a position to demand lifetime loyalty. Loyalty is a two-way street. The nature of leadership should, however, include preparing an organization to function without you. That’s particularly important when you’ve asked others to organize their careers and businesses around your plans.
It’s not about do-goodism. It’s about long-term thinking about the consequences of our actions.
Company trust is part of the product
Windsurf offers a particularly revealing example. In July 2025, Google agreed to a licensing arrangement reportedly worth $2.4 billion while hiring Windsurf’s founders and selected researchers. Google took no equity stake or controlling interest. Most employees and the operating business remained at Windsurf, as Reuters documented.
For an enterprise customer, that distinction matters. Google got the people and access to the technology it wanted without taking responsibility for the product customers had bought. Keeping that product useful, supported, and moving forward remained somebody else’s job. That somebody was interim CEO Jeff Wang and the team that stayed.
In his public account of the weekend, Wang described employees asking for cash to be distributed while he needed money to pay bills and keep the product working for customers. Cognition subsequently agreed to acquire the remaining business and team, giving the product a new corporate home. Good. But the need to sort all this out after the founders’ departure is precisely the problem. The message customers could reasonably take from the Google deal was that a lucrative opportunity could secure the founders’ next chapter without settling the future of the product they’d sold.
Enterprise customers have to care about that future. They’re buying into a road map, a support organization, and a company’s ability to keep delivering. A product can perform beautifully today and still become a questionable investment if customers lose confidence in the organization behind it. Founders don’t have to stay forever. But they do need to build a business whose commitment to customers survives their departure. Otherwise, customers are being asked to make a long-term bet on a product whose own leadership may be making a much shorter one.
Trust has to outlast the deal
AI’s grand promises make this tension harder to ignore. OpenAI’s 2015 founding announcement described a mission to benefit humanity, “unconstrained by a need to generate financial return.” Its structure has changed substantially since then, although its nonprofit foundation continues to control its commercial business.
Raising capital and earning revenue can support that mission. Compute isn’t free. Indeed, the financial commitments are becoming enormous. In a recent Brookings conference paper, Columbia Business School’s Stijn Van Nieuwerburgh develops a scenario, rather than a forecast, involving $10.3 trillion of US investment in AI infrastructure during 2025 to 2032. Delivering anything approaching that scale requires lenders, infrastructure providers, and technology companies to plan years ahead. Their commitments have to survive changes in leadership and shifts in which part of AI looks most attractive.
Enterprise customers are making their own long-term investments. A developer adopting an AI tool has to learn how it behaves. A company deploying it more broadly has to evaluate its output, connect systems, establish security controls, and decide who handles failures. Those investments don’t disappear when the vendor’s leadership gets an attractive offer.
Of course, customers shouldn’t need a particular CEO or founder to stay forever. Building an organization that can survive departures is part of the product’s value. That requires successors who understand the business, teams with enough context and authority to maintain it, and a transition that doesn’t leave customers guessing whether yesterday’s road map still applies.
That work rarely generates the excitement of a billion-dollar talent deal. It’s nevertheless what makes the software worth depending on. If AI vendors want customers to entrust them with more consequential work, continuity has to count as an achievement alongside recruiting another famous researcher or announcing another financing round. After all, the people who remain have to answer the support tickets, maintain the integrations, and deliver the promises that brought customers in. Their work should figure in how we judge an AI company’s success, including the success of its most celebrated deals.
If customers depending on your software is the business model, making your company dependable is part of your job.