AI Agents Just Moved Into the Workflow — Three Signals in One Day

On June 23, three things landed at once: Menlo Ventures’ $3B fund, Anthropic’s Claude Tag living inside Slack, and MoEngage’s bet on millions of marketing agents. Enterprise AI agents are leaving the demo stage and entering real workflows. What founders should watch isn’t the flashy launch — it’s the governance gap.

What Happened

Three moves stacked up in a single day. Menlo Ventures raised $3 billion to mark its 50th anniversary — its largest fund ever, and the driver is clear. After putting $500 million into Anthropic in 2024, its stake grew across rounds to roughly $14 billion in value. Anthropic’s valuation has crossed $900 billion, and managing partner Shawn Carolan calls the bet a “bet-the-firm moment.” The same day, Anthropic shipped Claude Tag. It replaces the old Slack app with a persistent AI teammate you summon via @Claude: it reads threads, breaks tasks into stages, learns from other channels it’s granted access to, and accumulates context over time. It even has an ambient mode that jumps in on its own. MoEngage went further, acquiring Aampe — which assigns a dedicated agent to each individual customer — and layering Merlin AI custom agents onto a platform used by more than 1,350 brands. The bet: the future of marketing is millions of agents. Put together, the picture sharpens. Agents are no longer demos; they’re starting to live inside the tools people already work in.

What This Means for Founders

The numbers tell the trend. Gartner projects 40% of enterprise apps will embed task-specific agents by 2026, up from under 5% in 2025. Yet only 17% of organizations have actually deployed agents, while more than 60% plan to within two years — meaning the buying happens now. At the same time, Gartner warns over 40% of agentic projects risk cancellation by 2027, driven by governance gaps and unclear ROI, and only 21% of organizations have a mature governance model for autonomous agents. For founders, that gap is the opening. While OpenAI and the FAANG players race to embed agents into their own suites, and YC’s latest batches fill with agent startups, the seats next to specialized internal stacks stay empty. There’s a reason the most-hyped general agent rarely wins the enterprise: buyers fear what they can’t audit. If an agent can’t prove what it learned and how far its permissions reach, procurement stalls. Chase the gap, not the announcement.

What You Can Do Now

First, draw a hard line in your product between “execute” and “propose.” The more authority an agent holds, the heavier the governance burden — and that friction becomes an adoption blocker. Second, translate global launches into specific verticals and stacks. If Claude Tag lives in Slack, who builds the equivalent that lives inside the niche workflow tool your customers actually use? Third, design audit and logging from day one — show what the agent knows and where it learned it. That’s the core of the cancellation risk Gartner flagged. Fourth, attack narrow workflows where ROI is immediate, like marketing or support — the exact wedge MoEngage targeted with per-customer agents. A narrow agent that owns one task end to end sells faster today than a flashy general-purpose one.