Lexful’s SaaSpocalypse Insurance
The AI era system of record startup for MSPs believes targeted context will ensure its long-term relevance to the industry.
I contributed to the SaaSpocalypse earlier this year.
As I mentioned some months back, I use a vibe-coded app written in Claude Code to run Channelholic’s editorial calendar. It costs me nothing to operate and serves my needs far better than the jerry-rigged Airtable system I’d been using previously.
So I canceled my Airtable subscription.
Which means I’m one small part of a series of events resulting in Airtable’s shocking decision last week to sell most of the company to Bending Spoons, the tech industry scrapyard purveyor currently wringing a few last remaining drops of value out of such once promising ventures as Vimeo, Evernote, and AOL.
What’s to stop companies more familiar to MSPs from the same fate? I suggested strength in risk mitigation and subject matter focus in a March post that would reference relationships too if I had it to write over again. Lexful, the AI-era system of record startup you last read about in February, is betting on the focus part of that equation for reasons best illustrated by something CEO Pinar Ormeci (pictured) experienced recently when she asked Anthropic’s cutting-edge Claude Fable 5 model to produce a dashboard for her showing daily active users, subscriber totals, and other critical metrics.
The result she got back was impressive, so she showed it to her head of operations, who soon asked her a question she couldn’t answer: where did that tenant count come from? Because it takes an API call into Lexful’s application to get that, and Fable doesn’t have API access.
Which didn’t slow the model down for a moment. It just made something up.
The problem, according to Ormeci, is that Claude is both eager to please and knowledgeable about everything except the all-important specifics that define you and your business. “The context window is simply too big for it not to hallucinate, or for you to trust that it will not hallucinate,” she says.
Narrower context alone won’t fix the problem, though. AI needs more than just MSP-specific data to give help desk technicians accurate answers. It needs clean, consistent, structured MSP-specific data, and if you don’t supply it, as we’ve just seen, it’ll fabricate something that looks right but isn’t.
Which poses an obvious problem for any MSP using LLMs to automate their help desk. “They’re not going to tell you we couldn’t find it or your data isn’t structured enough,” Ormeci says. “They’ll just give you very confident answers.” And it’ll be up to your technicians to determine if that confidence is warranted.
“Before they were wasting time searching for stuff,” Ormeci says. “Now they’re going to have to keep validating everything the AI is giving them, because you can’t trust anything that it’s saying.”
Which is why Ormeci isn’t worried about becoming the next Airtable. “You have to have a trusted structured knowledge foundation for AI to properly work,” she says. “That’s the fundamental problem that we are solving.”
Venture investors aren’t worried about Lexful becoming a SaaSpocalypse victim either, apparently. Two in particular, Top Down Ventures and York IE, recently committed $7 million of seed funding to the company, whose solution became generally available last week following a phased, multi-month test rollout.
“Our adoption numbers are not public, but I can tell you this: we have hundreds of MSPs in our ecosystem, either as customers or in our pipeline,” says Ormeci. How many more she’ll get in the months ahead depends on how trustworthy Lexful’s structured MSP data turns out to be.