Nobody posts the day 400 update

Has anyone dropped one of those "we replaced Salesforce" posts into your team chat yet? They tend to come with a dollar figure, a weekend, a screenshot of something that looks a lot like a CRM, and a line about a cancelled contract. Honestly, I like them, because somebody built something that worked and wanted to share it. What I keep noticing is that they almost always get written in the first few weeks, when the new tool is still fast, the person who built it is still around, and nobody has run into the strange edge case yet. That's probably why nobody seems to post the day 400 update.

The best known of these this year is Curative. In Salesforce Ben's roundup of the viral CRM swaps, Fred Turner is reported to have cancelled his $600,000 Salesforce contract after vibe coding a CRM in two months, and Business Insider Japan adds that Curative plans to cut its overall SaaS spending by about 80% this year and put that money into AI. That's the version that gets passed around, but there's more. In the Business Insider Japan piece, Turner says the company's Anthropic spending has grown sixfold a month over the past six to seven months, from tens of thousands of dollars to several million dollars a month. And when he was speaking to Salesforce Ben, he called maintenance "definitely one of the most challenging pieces."

I'm not saying Curative got it wrong, because I don't know, and neither does anyone scrolling past it. What interests me is that the second chapter was right there, told by Turner himself, and the version that went around stopped at the big number. A lot of the genre has the same shape. Salesforce Ben lists Auger's founder describing a productive weekend in which he built what the team needed ("Took a night and a morning"), and Atonom cutting a $40,000 Salesforce bill to $1,200, a story published by Lovable, the company that helped build the replacement. Vendors post their wins too, and as a marketing person I'm hardly one to talk, but it's worth knowing who's telling the story.

Klarna is probably the clearest example. Its CEO, Sebastian Siemiatkowski, said the company had "shut down Salesforce," and later, as Salesforce Ben reports, he wrote that he was "tremendously embarrassed" and felt his comments had been widely misinterpreted. His own clarification was "We did not replace SaaS with an LLM." What Klarna actually did was pull together company knowledge spread across too many tools. That's less a scandal than what happens when a sentence written for a feed gets read as a plan. A launch post describes a moment, while an operating decision has to hold up for years, and buyers keep forwarding the first kind as if it were the second.

The survey data has the same trap in it. Rajesh Beri's breakdown of McKinsey's 2026 State of AI survey points out that 32% of organizations decided against buying at least one software product or feature because they could build it internally with agentic coding tools, which is a real shift. As Beri notes, nobody was asked whether the replacement shipped, passed an audit or is still running a year later. In his words, "The 32% is a decision not to buy, not a system in production." Meanwhile, the share attributing any EBIT impact to AI stayed flat at 37%. Then there's the boring number that never trends, because estimates of maintenance as a share of software lifecycle cost, Beri writes, cluster between 60% and 90%, which makes the weekend of building the cheap part.

Three questions before you forward it to your CFO

If one of these lands in your team chat with "should we do this?" underneath, I'd want three things answered first. The first is who owns it in year two, and I mean a specific person, not "the team." Beri's advice is to ask for "a named owner and a named backup, by person, for years two through five," and he adds that if the answer is "the platform team," the proposal has no owner, which is funny mostly because it's such a familiar answer.

The second is what it costs to run for five years. Beri says, "Treat a build proposal like a vendor proposal," which means making it produce a five-year cost rather than a sprint estimate, and that cost has to include the model bill. Curative's story shows the AI line can grow quickly too.

The third is what happens when the builder leaves. Salesforce Ben asks exactly this, alongside who handles security incidents, integrations and regulatory changes. A weekend tool built by one person lives in that person's head, so either it gets written down or you plan on rebuilding it.

None of these are reasons not to build, since any vendor would face the same questions in a sales cycle and an internal build deserves the same treatment. It's also why we care at Workbricks about an app having an owner and access rules from the day it's made, not bolted on later.

Some of these swaps will work, and I mean that. Some companies really were paying for a giant product and only using a few screens of it, and a focused tool their own team owns will serve them better. I'd love to read about those. So, builders, please post again on day 400 and tell us what broke, what the bill looks like, who's on call, and whether you'd do it again. That's the post I'd actually forward to a CFO, and it would help the rest of us decide far more than the launch-week version.