The Next Systems of Action: Agentic Workflows and the End of Toil with John Foreman, CPO, Clio

John Foreman delivered what was probably the most operator-honest talk of the day: a walk-through of why so many teams end up shipping “AI features” that look impressive in a demo but never change the actual workload. His argument was simple: if you want real leverage, stop thinking about interfaces and start thinking about toil.

He began with a split that feels obvious once you hear it: in any job, there’s the judgment work, and then there’s everything around it: the follow-ups, the formatting, the searching, the checking. That’s the toil. GenAI is good at almost all of it, yet most companies still wrap their AI efforts around a chat box. Foreman’s view: the chat paradigm has flattened everyone’s imagination. Most users don’t actually want to “talk” to software. They want the work to disappear. At MailChimp, SMB customers literally told him, “I don’t want to log into your product. Just do it for me.”

‍

Vertical SaaS, he argued, is uniquely suited for this shift. These products already own the nouns, the records, the documents, the timelines, the messages. With AI, they can own the verbs: parsing, updating, scheduling, reconciling, drafting, routing. Foundation models alone can’t do this; without proprietary context they’re “eloquent idiots.” That’s why Clio made its big legal content acquisition so every AI-generated claim or citation could anchor to sourceable authority.

‍

The real hurdle wasn’t technical. It was organizational.

‍

Clio’s first shot at AI relied on a specialist pod: ML engineers over here, product teams over there. Unsurprisingly, it stalled. The AI experts knew RAG and vector stores; the product teams understood lawyers, workflows, and the dozens of micro-tasks that actually constitute a legal day. They weren’t solving the same problems. So Foreman cut the structure in half. No AI PM track. No ML-only engineering team. “We will not have two job descriptions.” Instead, every product team inherited AI expertise, and every team, rather than a single pod, became accountable for identifying and eliminating toil.

‍

To make the concept stick, Foreman used two “bumper stickers” that Clio teams now rely on:

‍

Jesus Take the Wheel: AI handles the task end-to-end including deadline extraction, intake, timekeeping, and invoicing.

‍

God Is My Co-Pilot: AI drafts or analyzes; the lawyer applies judgment.

‍

Most legal work sits somewhere on this spectrum. High-stakes reasoning stays with the attorney; nearly everything else is fair game for automation.

‍

Clio’s day-in-the-life demo drove the point home. It walked through inbox triage, deadline extraction (with linked citations), matter updates, client intake, consult prep, document review, motion drafting, template generation, structured review of huge file sets, billing assembly, collections outreach, and end-of-day briefing. Only a few moments involved anything resembling a chat interface, and even those are being standardized into structured workflows, because, as Foreman joked, “lawyers write prompts like Google searches.”

‍

His closing argument was aimed squarely at product teams. Agentic AI isn’t an add-on. It changes the job of product. You don’t roadmap features anymore; you roadmap the elimination of toil. The Vertical SaaS companies that put AI into every team, tie it to proprietary context, and measure themselves by how much work they lift, not how clever their interface looks, are the ones that will define the next generation of Systems of Action.