agentic-ai

The last 100 yards of the marathon

Willem LarsenWillem LarsenJuly 16, 20263 min read

The queue of finished work lands on a human being who now has to fix the decisions the AI made. We have been solving that problem for decades. We just called it something else.

ShareLinkedIn

Talking recently with a potential collaborator, he framed the thing I care about most in the most succinct way I have heard.

"What you are doing," he said, "is all about getting these agentic systems to where they can finish the last 100 yards of the marathon. That last stretch where, for the industries we work in, safety and quality are the only thing that will get you over the line, no matter how fast your agents have worked, no matter how much labor they have done."

He was right.

There are already a thousand tools and apps for pointing agents at whatever work you like. But how often does that queue of finished work get handed to a human being who is now saddled with fixing the decisions the AI made?

A towering, unstable pile of documents, tickets, and code printouts heaped on a single office desk, spilling over the edges, one small human figure dwarfed beside it: the work AI hands back, waiting to be fixed

My wife, who heroically tackles the laundry in our household, calls the massive pile that comes from five bedrooms and three bathrooms "Mount Washington."

I see that same image for teams of knowledge workers who have been told by leadership to start leveraging AI. Instead of real empowerment, they have traded old problems for new ones. Mount Washington now sits at every desk, the agents adding to it like the enchanted brooms in the Sorcerer's Apprentice.

The solution, the one we have built at ClickChain, is the same thing we founders have done as technical coaches across decades-long careers. We treat agentic work as a flow system.

For us, "governance" is not a new concept that only applies to agents. We have been doing governance by another name all along, with technical teams and whole organizations. Chartering, working agreements, retrospectives, Kanban boards, pairing, ensembling: these systems of deliberative friction and flow management are just how you do the work in complex adaptive systems. Why invent a new mental model for yet another complex system? Stick with what works.

A lone marathon runner in the final stretch, a well-lit lane marked out and held open for them all the way to the finish line, the crowd and noise falling away behind: the last 100 yards, made safe to cross

Until colleagues and competitors catch up, just as we have always guaranteed safety and quality in our human systems and technical teams, we will be the only AI platform that guarantees safety and quality for agentic work.

  • agentic-ai
  • ai-governance
  • flow

Keep reading