A Marc Benioff-backed startup thinks AI can solve the AI deployment problem

Getting an AI agent to actually work inside a big company is apparently a lot harder than getting one to sound smart in a demo. That gap has spawned an entire cottage industry of "forward-deployed engineers," specialists who parachute into enterprises to wire AI tools into the tangle of Salesforce instances, ServiceNow tickets, and Workday records that run the actual business. Efrat Rapoport, a former Salesforce executive, thinks that's a broken model. "AI, paradoxically, increases the demand for professional services," she says. Her answer is a company called June, which came out of stealth Monday with $20 million in pre-seed funding led by Marc Benioff's Time Ventures, plus checks from Michael Dell, Aaron Levie, and George Kurtz.

Rapoport isn't a first-timer here. She and her three cofounders — Ohad Hen, Barak Goldstein, and Idan Tsitiat — previously built Bonobo AI, a voice-to-text company from the pre-transformer era that Salesforce acquired in 2019. Years spent inside Salesforce's own AI push apparently taught them exactly where enterprise AI deployments go to die: not in the model, but in the plumbing underneath it. Duplicate database fields. Years of technical debt. Ten different teams using the same field to mean ten different things. Building a slick agent template, Rapoport argues, is the trivial part. Making that agent survive contact with a company's actual data is the whole problem.

June's pitch is that it can automate the diagnostic work an FDE would normally bill hours for. The platform scans a client's existing systems, maps out where the bottlenecks and redundancies live, then hands back something like a project plan: remove these duplicate fields, connect this data source, click build on this task. It's less a magic AI brain and more a systematic checklist generator that happens to also do some of the building itself, notifying relevant teams through whatever Slack or Teams setup they already use.

The real-world test case is CMG, a major U.S. mortgage lender. Chief strategy officer Paul Akinmade had already moved engineering work to Claude Code, but ran into a wall trying to hook it up to Salesforce — a wall made worse by the fact that he'd publicly promised, at Salesforce's own conference, to have 100 agents live within a year. Weeks of meetings with architects and forward-deployed engineers produced little progress, he says, until June gave his team a concrete map of where and how to deploy safely. Tellingly, Akinmade says he warned Rapoport upfront that he didn't want another black-box tool requiring a specialist priesthood to operate. That's arguably the more interesting tension here: Rapoport describes June as complementary to FDEs, but her own customer is buying it explicitly to avoid needing them at all.

Whatever the pitch deck says — and there wasn't one, notably, for this raise — June is betting that the next phase of enterprise AI isn't about smarter models but about somebody, or something, finally cleaning up the data wreckage those models have to run on top of.