June AI emerged from stealth this month with a $20 million pre-seed led by Marc Benioff’s Time Ventures, joined by Michael Dell, Aaron Levie, Diane Greene, and CrowdStrike’s George Kurtz. The pitch, from CEO Efrat Rapoport and three co-founders whose last company Bonobo AI was acquired by Salesforce, is that enterprise AI has become so hard to deploy it needs its own AI to deploy it.

“The industry’s answer to AI implementation is, ‘let’s hire more and more and more people,’” Rapoport told TechCrunch. June’s platform does process mining across Snowflake, Databricks, Salesforce, ServiceNow, and Workday, then generates implementation roadmaps and builds agents automatically. CMG chief strategy officer Paul Akinmade described his team burning weeks trying to wire Claude Code into Salesforce before June unblocked them.

The numbers behind that stuck feeling are grim. ISG’s 2025 enterprise study, reported by Calcalist, found only 31% of AI use cases ever reach full production, at an average spend of $1.3 million each. Forbes, citing Boston Consulting Group, put measurable ROI at 5% of generative AI projects against $30–40 billion poured in, with 70% of failures traceable to organizational culture rather than the technology itself.

So the June thesis is coherent: if the bottleneck is humans and forward-deployed engineers, automate them.

But it’s worth noticing what that thesis quietly assumes. It assumes the customer already runs Snowflake and Workday and has a Salesforce org complicated enough to justify a $1.3M implementation. That’s a tiny slice of the economy, and it’s the slice that generated the 5% ROI figure in the first place.

Small and mid-sized businesses live in the inverse world. The BCG-cited winners weren’t running process-mining tools across five platforms; per Forbes, they automated narrow, repetitive, well-documented chores like invoices and scheduling. That’s the design brief behind no-code SMB tools like LemonLime, which skip the deployment layer entirely rather than papering over it.

June is being built to hide enterprise complexity. The more useful bet for everyone else is software built to never accumulate it. Two roadmaps, one industry, and the gap between them is where the next decade of AI spending will actually get decided.

Sources