The Skip Slot: The Near-Break-Even Agent, Not the Biggest Loser

The SeatCompress action plan reserves one slot for an agent you should not deploy. The obvious way to fill that slot is with the agent that loses the most money, the one the engine can point at and say "this would lose $2 million at your scale." We tried that. It was useless. The slot now goes to the agent closest to breaking even, and the difference between those two choices is the difference between an alarming dashboard and a useful one.

The problem: the biggest loser is not a decision

Every action plan the engine builds has a skip slot: one agent it recommends you pass on, surfaced next to the agents it recommends you deploy or renegotiate. The question is which agent earns it.

The first version picked the biggest dollar loser. If one agent would burn $600,000 in year one against your stack and another would burn $2 million, the $2 million agent won the slot. It looked decisive. A big red number, a clear "no."

It was also uninformative. An agent that loses $2 million at your scale is an obvious skip. Nobody needs a dashboard to tell them not to spend $2 million to save a fraction of it. The recommendation carried no information, because the decision was never in doubt. Worse, it was alarmist: a $2 million figure on a screen reads as a threat, when the honest read is "this was never a candidate." The slot was spending its one recommendation on the least decision-relevant agent in the set.

The methodology: rank by closeness to break-even

The decision-relevant agent is the one near the line. So the scoring flipped to reward closeness to break-even instead of size of loss.

The score for a skip candidate is:

50,000 / max(net annual loss, 5,000)

Smaller losses produce bigger scores, so the agent with the net closest to zero ranks first. An agent losing $14,000 outranks one losing $290,000, which outranks one losing $2 million. The slot goes to the top score, which is the agent sitting closest to profitable.

The floor of 5,000 on the divisor does one specific job. Without it, an agent that loses $50 would score 50,000 / 50 = 1,000, an astronomical number that would seize the slot for a rounding error. With the floor, any agent within $5,000 of break-even divides by 5,000 and scores 50,000 / 5,000 = 10. So everything inside the break-even band ties at a ceiling of 10 rather than spiking toward infinity for whichever agent happens to be nearest zero. Trivial losses do not dominate. They cluster at the top, which is the honest picture, because an agent losing $200 and an agent losing $4,000 are both just "basically break-even, watch it."

Why the near-break-even agent is the one that matters

The skip slot has a specific purpose, and it is not "here is a bad idea." It is "watch this one as you scale."

An agent sitting a few thousand or a few tens of thousands of dollars underwater is not a permanent no. It is a not-yet. The economics that make it unprofitable at today's seat count often flip as the targeted tool's footprint grows, because compression scales with seats while the agent's cost does not scale as fast. That is the agent a CFO should actually track: the one that becomes a deploy the moment the stack crosses a threshold. So the card does not just name it. It shows the unlock threshold, the deployment size where the agent stops losing money, so you know exactly what to watch for.

The $2 million loser has no such story. It is not going to flip profitable with modest growth. It is structurally wrong for the stack. Surfacing it teaches the CFO nothing they did not already know. Surfacing the near-miss, with its unlock threshold, tells them something actionable: this one is close, here is the number that turns it green, check back when you hit it.

A worked example: the 12,000-employee stack

Reuse the stack from our year-one realization write-up, a 12,000-employee company evaluating several agents at once. Two of them come back unprofitable in year one.

AiSDR on Outreach. 80 Outreach seats at $100 a month, AiSDR compressing 30 percent. Gross annual compression is 80 times 0.30 times $100 times 12 = $28,800. At the 0.4 year-one realization factor that is $11,520 realized. AiSDR is a flat-fee agent at $900 a month plus a $15,000 setup floor, so year-one cost is $10,800 plus $15,000 = $25,800. Net: $11,520 realized minus $25,800 cost is a loss of $14,280.

Moveworks on ServiceNow and Freshservice. The same write-up works this one out to a year-one loss of $326,136.

Now add a hypothetical structural mismatch, an agent that would lose $2 million against this stack.

Score each one:

AiSDR wins the skip slot with a score of about 3.5, more than twenty times the Moveworks score and roughly a hundred forty times the $2 million agent's. The old logic would have shown the $2 million agent. The new logic shows AiSDR, because AiSDR is the one worth watching.

And AiSDR is genuinely worth watching. Its steady-state economics are already positive; it is only the first-year ramp and setup that put it underwater. So the card computes the unlock threshold. Year-one net turns positive once realized savings cover the $25,800 year-one bill. Each Outreach seat delivers 0.30 times $100 times 12 times 0.4 = $144 of realized year-one compression, so break-even lands at $25,800 / $144 = about 179 seats. The company is at 80. The card says, in effect: skip AiSDR now, revisit when your Outreach footprint clears roughly 179 seats. That is a decision a CFO can file and act on later. "This loses $2 million" is not.

The zero-overlap filter

One more piece keeps the slot honest. Before an agent can even be scored for the skip slot, it has to actually compress something in your stack. The engine applies a hard filter: an agent whose gross annual savings against your tools is zero never enters the ranking.

This matters because a catalog has agents that target tools you do not own. An agent built to compress a helpdesk product you do not run has zero overlap with your stack. It is not a skip recommendation. It is catalog noise. Ranking it would be like advising you to skip a product you were never going to buy for a job you do not have. The filter drops those entirely, so the skip slot only ever holds an agent that genuinely touches your stack and genuinely does not pay off yet. The action-plan engine treats the skip slot as advice, and advice about tools you do not own is not advice.

What the CFO does with the skip slot

Read it as a watch list, not a warning. The agent in the skip slot is usually close, not catastrophic. It is there because it might become a deploy, not because it is a disaster you narrowly avoided.

Note the unlock threshold. The card tells you the deployment size where the agent flips profitable. Write it down. When the targeted tool's seat count approaches it, re-run the numbers, because the skip may have become a deploy.

Do not go hunting for the biggest loser. If you find yourself wanting the dashboard to show you the most alarming number, resist it. The biggest loser is the least useful output the engine can produce. The near-miss is where the decision lives. This is the same reason we discount rather than inflate the numbers we put in front of you: the honest number is the useful one, not the dramatic one.

Try the free calculator, 15 seconds, no signup. If an agent lands in your skip slot, look at its unlock threshold before you dismiss it. It might be one growth quarter away from a yes.