The Army Behind You Is Only Worth What You Can Keep.

In 2012 I was two months into a consulting engagement at Travelers in Hartford, leading a team in India on a competitive rating project: when a website showing auto insurance rates from several companies called in, our system answered with Travelers' rates. A competing consulting firm sat in that same room, and its lead also had a team in India behind them.

The day before I started, my wife Nicole and I found out she was pregnant with our son Aidan. My in-laws had been living with us since a hurricane the previous August. A lot of people depended on me that year, at home and at work.

Then came the sprint retrospective. We had completely failed to meet the business need, and we were not even close to what was actually needed. It was not only the team's failure; it was mine. I had not fully understood what the business asked for, and the team had delivered neither the business ask nor what I had asked of them.

I walked out to the parking garage, sat in my car, and took out a yellow legal pad. What went on the page was raw: the fear of failing at a time when so many depended on me, and what I would later learn to call imposter syndrome, unchecked. I felt like a failure. I started writing, not because I knew what to write, but because I did not know what else to do.

The team in India was at different points in their careers, and even at the same level, each member had different strengths and weaknesses. My job was to aim each of them at the right work, use the strengths, and cover the weaknesses. None of that works if the person in the room has misread what the business needs.

Fourteen years later, at a dinner in Manhattan, Bob Matsuoka said that more and more, each person will represent a shadow army of agents behind them, and he wanted to know how his would work with everyone else's. I pointed out that consulting has run that way for decades. A moment later someone at the table put the real question plainly: are you getting any value out of it? Hartford is my answer to that question.

The shadow army is pitched as headcount without the management. Each engineer becomes a team of ten, and the agents carry none of what made my team in India hard to lead: no careers at different stages, no uneven strengths, no trust to build. They are interchangeable capacity, so the leader's job shrinks to dispatching work and approving what comes back, and leverage is measured in agents per engineer. Almost every technical leader has seen the slide with ten agents behind each engineer, and none of those slides shows who decides which of the ten to trust.

Agents are uneven too, and they lack something that team had. They are built for different purposes, and they are not equal however good the inputs are. They do what is asked, with no personal experience or judgment of their own and none of the differing viewpoints that sharpen a deliverable; their only context is what they were trained on and your prior interactions with them. Each new task starts with a tuning period before the output is consistent. They have to earn trust the same way people do: through consistent output quality over time.

Treat the army as uniform capacity and three things follow. First, the army multiplies the input: if the person in the room misreads the business ask, as I did in Hartford, every member executes the wrong thing in parallel, and a fleet of agents does it in an hour, fluently, with output that looks finished. Second, uniform trust forces uniform review, so the person in the room either checks everything closely and becomes the bottleneck or checks everything lightly and ships the weak work with the strong. Third, with no record of which member produced consistent work on which kind of task, trust cannot grow, and review time never moves from what is proven to what is new.

The cost shows up where it did for me: late, in the retrospective or in front of the business, after the army has produced a great deal of output nobody should keep.

The leaders who get real leverage from an army do not have better agents. They check the work against the business sooner.

After Hartford, I changed the loop. We ran more frequent demos and talked with the business far more often, because a two-week sprint is too long to wait to find out whether you understood the ask. Those demos became one part of what I named The Cadence: the structured, recurring moments, from standups to planning to demos, where a team synchronizes, surfaces what is blocking it, and closes open decisions before ambiguity piles up. The demo with the business is the moment that catches a misread ask, and it only works if it comes around often enough. With agents, the rhythm has to run faster still, because the army produces in hours what my team in India produced in days.

So size the army to your feedback loop, not your budget. Check the business ask at the speed the army produces. Extend trust to each member, person or agent, for each kind of task, only as consistent output earns it. The question that matters is how often their output meets the business before it ships, whatever the size of the army.

List the agents and AI workflows your team runs today. Next to each, write two things: the last time its output was put in front of someone on the business side before it shipped, and how much of its work you still check line by line. If the first answer is "end of sprint" or "never," the army is producing faster than you can tell whether it is worth keeping. If the second answer is the same for every row, you are trusting them all the same, which means you have not measured which of them has earned it.

LeadershipOS™ has the full Cadence layer: how to build a rhythm that closes decisions predictably, so the work meets the business before an army has run two weeks in the wrong direction: http://TheLeadershipOSBook.com