Adding Levers as You Learn

An executive who wins a hard decision on the third attempt usually tells it as persistence. They made the case, heard an objection, came back with an answer to it, heard a different objection, and came back again. Looking back, the first two attempts seem like the price of getting in the door, and the third seems like the case that won.

The Lam System for Political Interaction writes the same history as one cycle with three levers, each declared only after the one before it had been played and read. Two of the readings failed and the third validated. Every lever needed something the actor did not hold, so each one opened a sub-cycle before it could be played. This is the adaptive stance carried through to the end, and the record it leaves shows what the persistence story leaves out: what each attempt taught, what each one spent, who learned about the plan along the way, and how little the finished record can say about which lever decided the matter.

The firm and the decision

Illustrative. The people and the company are invented. Maren Holt runs platform engineering at a software firm that sells accounting software to small businesses. Partner developers build integrations on the firm's public API, and each product team maintains the endpoints for its own part of the product. Maren wants ownership of the public API moved to her team, with its maintenance budget. That is the interest, and the cycle opens with it under her agenda.

Detection names the target and the choice in front of him. The decision belongs to the chief product officer, Teodor Vasile. The preferred position is that Teodor moves the API to platform engineering. There are two adverse alternatives: Teodor leaves each product team owning its own endpoints, which is what happens if he does nothing, or he hands the API to the solutions engineering team, which works with partners every week and reports to the chief revenue officer. The objective is that Teodor moves the API and the move brings its ownership and budget to Maren. That completes the movement the system calls Declare.

Maren does not know which of Teodor's concerns will decide the matter, so she declares one lever and holds the rest of her thinking back until she has seen what it does. Stances Toward Uncertainty sets out what that choice buys and what it pays.

The first lever

The first lever is to show Teodor what split ownership costs: most of the past year's customer-facing incidents began at endpoints that no single team owned from end to end, and those incidents cost engineering time. Its projection is that Teodor will come to see the split as costing more than the disruption of moving it.

The lever needs two things. One is the incident record, which Maren's team keeps because it runs the on-call rotation. The other is the engineering hours those incidents consumed, which sit in time-tracking data owned by the vice president of engineering, Rafael Quintero. The gate tests every resource a lever declares, and it fails on the one she lacks.

She escalates. Teodor's cycle waits at its gate, and a sub-cycle opens under the first lever, aimed at the hours figures and nothing else. Its target is Rafael, who can release the figures or keep them inside engineering. Maren's lever on him is to show that the figures will argue for less interrupt work landing on his product teams, which he has wanted for a year, and it spends her standing with him. He reads the request as serving his teams, releases the figures and gains a case for lighter on-call loads made by someone outside his own organization. The sub-cycle closes and the figures flow back.

Teodor's cycle resumes at its gate, which tests again and passes. Maren executes the lever, and the incident record and the hours figures are spent. Then she reads him. Teodor accepts the incidents and the hours and says the product teams can fix their endpoints themselves under a common standard. Moving the API in the middle of the year would put the roadmap dates he has committed to at risk, and he will not trade those dates for a reliability gain the teams could get another way. He still prefers the first alternative. Maren writes the reading down as she observed it, and writes it as a failure.

The second lever

The failed reading leaves the cycle open, and the target's selection cannot be written on the strength of a lever that did not move him. Maren has to choose an exit, and Perception Failure describes the three. Her diagnosis is that the position is still within reach and the lever was aimed at the wrong concern: Teodor weighs the roadmap above the incidents. So she declares a second lever, fresh, with a number of its own. The first stays on the record with its failed reading beside it.

The second lever is to show Teodor a transition plan that moves the API over two quarters without touching any product team's roadmap dates, staffed in a way the vice president of engineering has agreed to in writing. Its projection is that Teodor will see the move as costing the roadmap nothing. It needs Rafael's written agreement to the staffing plan, which Maren does not hold, and the gate fails.

A second sub-cycle opens, with the same person as its target. The system keeps it as a separate cycle, under the second lever, with its own detection and its own levers, and nothing in the record merges it with the first. Rafael now knows what the hours figures were for, and the standing Maren had with him went on the first request, so this time she pays with something else. Her lever is to take the API's on-call rotation off his product teams from the first day of the transition, and it spends her team's spare on-call capacity. Rafael reads the offer as cutting his teams' interrupt load at once, agrees in writing, and the agreement flows back.

The parent resumes at its gate, passes, and Maren executes; the agreement is spent. Teodor accepts that the roadmap is safe. Then he says that the partners who build on the API complain that nobody answers for it, that solutions engineering talks to them every week, and that platform engineering never has. If anyone should own the API, he now thinks, it is the team that already knows the partners. His preference has moved from the first alternative to the second. The reading fails again.

The third lever

Between the two readings Teodor's comparison has shifted. The product teams keeping their endpoints no longer leads, and the rival that detection listed at the top of the cycle now does. His remaining concern is the partners. Detection had named solutions engineering as an alternative, so the reading found something already on the list. Had it not been listed, the same reading would have looked like an objection from nowhere. Detection takes up why that list carries so much risk.

She declares the third lever: to show Teodor that the partners themselves want the API owned by the engineers who run it, with the partner program staying their point of contact. Its projection is that Teodor will see the move to platform as what the partners are asking for. It needs a written statement from the partner council, which only Philippa Tremayne, who runs the partner program, can convene and send. Maren does not hold it, and the gate fails.

The third sub-cycle has Philippa as its target. Maren's lever is to commit her team to brief the partner council on every API change once a quarter, and it spends her standing with Philippa. Philippa reads the commitment as giving her partners the engineering access they have asked for while keeping the relationship in her program. She convenes the council, the statement is written and it flows back.

The parent resumes, the gate passes, and Maren executes the third lever; the statement is spent. This time Teodor reads the choice as she projected. With every lever read, he selects: the public API moves to platform engineering. The outcome resolves. His obtained utility is fewer incidents at the API, roadmap dates left intact and a change the partners asked for. The cycle closes, and ownership of the API and its budget flow back to Maren.

What the shape held fixed

Three levers ran three passes through the middle of one cycle, each in the same internal order, and the cycle converged on one selection and one close. The cycles waiting on a sub-cycle rose to one and fell back to none three times. Maren never had two failed gates waiting at once, so she never had to choose which to escalate first.

One-at-a-time declaration also changes the exits at a failed gate. Each lever was the only one in the cycle not yet read, and the system will not let an actor moot the last live lever on its own, because nothing would be left to carry the cycle to a selection. At each failed gate her moves were to escalate, to declare another lever before setting this one aside, or to end the cycle by mooting or abandoning it. Dropping the lever and carrying on without a replacement was not among them. An actor who declares every lever at the outset can set one aside at its gate while the others continue. An adaptive actor setting a lever aside is either replacing it or stopping, and the record has to say which.

Counted, Maren started with four things the maneuver touched: the incident record, her standing with Rafael, her team's spare on-call capacity and her standing with Philippa. Four things flowed back: the hours figures, Rafael's written agreement, the partner council's statement and, at the close, the ownership of the API. Seven were spent. She ends holding one thing from the maneuver, the API, which is four plus four less seven. Every resource a sub-cycle fetched was spent by the lever that needed it, and the resources spent on the two failed levers stay spent.

What each lesson cost

The adaptive stance learns from failed readings, and the record shows where the price of each lesson was paid. Both lessons were paid for in front of Teodor. Each failed reading was a lever he had already heard, and by the third meeting he had heard two arguments for the move that, by his own reckoning, had not answered his concern. The record counts only what Maren held, so Teodor's patience appears nowhere in it, though each meeting drew on it.

Some of the price was paid in front of other people. Rafael was asked twice. The second time he knew the plan, knew that the first request had not been enough to carry Teodor, and could set his price accordingly. Escalating widens the circle of people who know what a maneuver is attempting, as Escalation describes, and an adaptive maneuver widens it one lever at a time, often by going back to the same people, each of whom knows more at the second request than at the first.

In return the actor gets a reading of the target taken after each change. Maren's three readings, in order, say that Teodor weighed the roadmap above the incidents, then the partners above the roadmap, and that once the partners' concern was answered nothing else stood in the way. She could not have written that list at the start. Teodor might not have written it either, because his concerns came out in the order her levers met them.

What the record can and cannot settle

Afterward, Maren will be tempted to tell the story as persistence: the partner statement won it. The record supports less than that. It holds three executions, two failed readings and one validated one, and the selection that followed. Teodor selected once, on the comparison as all three levers had left it. The incident case and the transition plan each failed to bring him around, and each left its mark on the comparison: after the first he had accepted the incident costs, after the second he had accepted that the roadmap was safe. The third lever landed on a comparison the first two had already moved. Whether the partner statement alone would have carried him, had she played it first, is a question about Teodor, and the record holds no answer to it.

It cannot say either whether the roadmap concern existed before the first lever or was raised by it. Detection named Teodor's options, and no entry before the first reading says how much any concern weighed with him. A concern that appears in a reading may have been there all along, or the lever may have brought it out.

What the record does settle, it settles exactly. It shows which lever each sub-cycle was opened for, what each lever spent, that Teodor's selection waited for the third reading, and what Maren held at every point. Of the nine questions below, five the record settles and four it cannot.

What the maneuver teaches

An adaptive record carries more failed readings than a careful or batched one, for reasons the stances essay gives, so their number says little about how well Maren reads people. The question worth asking of each failed reading is whether the diagnosis written after it held up. Maren concluded twice that the position was reachable and the lever had missed, and the third lever tested both conclusions at once. Had it failed in the same way, the record would have pointed back at those two diagnoses and the readings under them.

The readings also outlast the maneuver. The next time Maren needs something Teodor controls, she knows that he puts committed dates above operational gains, and that he listens to partners. She has it as three readings with their projections beside them, in the order Teodor gave them, which is more than her memory of three meetings would keep.

Rafael and Philippa keep something too. Each knows part of what Maren attempted and roughly what it cost her, and Rafael knows that his first answer was not enough to carry Teodor. None of that appears in her record, though it will bear on the next request she makes of either of them.