Automattic’s Reciprocal Severance Agreements and the Architecture of Approval

When the approver and the beneficiary are the same two people, the approval step stops doing its job — and that is a structural property, not a judgment about anyone’s conduct.

Key Facts — Automattic Severance Story

$8.15M

Combined severance figure per HR documentation reviewed by TechCrunch

2

Executives who each signed the other’s severance agreement during the leave window

60 days

Notice window in the “for cause” procedure under the agreements

30 days

Cure period required before a termination qualifies as “for cause”

What Happened

TechCrunch, which reviewed the underlying HR documentation, reported on September 16 that during a brief period in early September 2026 when Automattic’s chief executive was on paid leave, the company’s then-interim chief executive and its chief legal officer signed reciprocal severance agreements — meaning each signed the other’s. The terms, as reported by TechCrunch, include twelve months of base salary as a lump sum, accelerated equity vesting, the ability to exercise vested stock options, and one year of health coverage. TechCrunch puts the combined figure at $8.15 million that Automattic would now owe both executives following their departures. The circumstances of the leave and the subsequent return are disputed by the parties involved; this piece takes no position on that dispute and goes no further than that single sentence.

The agreements define “cause” narrowly, requiring a 60-day notice window, a 30-day cure period, and board majority approval before a departure can be classified as for-cause. One clause specifies that the interim chief executive’s removal from that interim role would not trigger severance, provided he remained in his substantive role as chief financial officer. Both executives subsequently departed when the chief executive returned. Automattic has been asked for comment — a pending request rather than a declination — and attempts to reach the two executives were unsuccessful. No inference is drawn from either fact.

Nothing in this piece alleges wrongdoing, bad faith, self-dealing, breach of duty, dishonesty, or illegality by any person. This is not legal advice, not investment advice, and not an allegation of wrongdoing by any person. The analysis below concerns the architecture of the instruments and the structural properties of approval controls generally.

The key insight: An approval carries information only when the approver and the beneficiary are different people. When A signs B’s package and B signs A’s, the set of approvers and the set of beneficiaries are identical — and the approval step, by design, no longer performs its function of introducing an independent perspective outside the transaction.

The Structural Read

Every governance control is, at its core, an information mechanism. A signature on a compensation document is not a formality — it is a claim that someone who does not receive the money examined the arrangement and judged it acceptable. That is the entire payload of the approval step. Strip out the independent party, and the signature still exists on paper, but the information it was designed to convey does not.

The reciprocal structure here moves the control that is designed to sit outside a transaction to a position inside it. This is a point about what a signature conveys and about the architecture of controls generally — not about the intentions, conduct, or judgment of any individual involved.

Structural Principle

“A control that has been moved inside the transaction it was meant to govern is no longer a control. It is a record of the transaction. The two are not the same thing, and conflating them is how organizations discover — after the fact — that their approval architecture was doing less work than the paperwork suggested.”

The procedural language in these documents — notice periods, cure windows, board majority requirements — is where the economics actually live. Each condition added to the definition of “cause” removes some departures from the for-cause category and places them in the category that triggers payment. The two categories are exhaustive: tightening one necessarily loosens the other. A tighter definition of cause is therefore a broader severance entitlement, expressed in procedural language rather than monetary language. That is precisely why the procedural clauses are the ones most often skimmed, and why skimming them produces an undercount of what a document actually promises.

One further structural observation about the text: Davies’s agreement specifies that losing the interim chief executive role would not trigger severance provided he remained chief financial officer. A clause addressing what happens when an interim arrangement ends is a clause written with the end of that arrangement in contemplation. That is simply what drafting is — an instrument anticipates the contingencies it provides for, or it is a poor instrument. The observation stops there. No state of mind, expectation, plan, or intention is ascribed to anyone.

Finally, there is a timing property worth naming on its own. Governance controls — notice periods, committee review, independent approval — all presuppose enough elapsed time for someone outside a transaction to examine it. A compressed window is simultaneously the circumstance in which such controls matter most and the circumstance in which they are hardest to operate. That is an awkward structural property of control design, and it applies as a general proposition. Nothing here says that any control was bypassed, evaded, ignored, or breached in this instance, and nothing here says any board failed at anything.

Three Implications

IMPLICATION 1 — CONTROL ARCHITECTURE IS THE PRODUCT

Governance documents are not paperwork that accompanies a decision — they are the mechanism by which the decision gets validated. Any organization that treats approval signatures as administrative formalities rather than as information-carrying instruments is running controls that look complete on paper but are structurally hollow. The Automattic situation is a clean case study in why the independence of the approver is the entire point of the approval step, not a procedural nicety.

IMPLICATION 2 — PROCEDURAL CLAUSES ARE ECONOMIC CLAUSES IN DISGUISE

Boards and compensation committees that focus review time on headline numbers — base salary multiples, equity acceleration percentages — while skimming procedural definitions are reading the cheaper part of the document. The notice periods, cure windows, and cause definitions determine how many real-world departures fall into the severance-triggering category. In any contested or fast-moving leadership transition, those procedural clauses do more financial work than the dollar figures they surround.

IMPLICATION 3 — TRANSITION WINDOWS ARE HIGH-RISK CONTROL ENVIRONMENTS BY DESIGN

Leadership transitions — planned or otherwise — are precisely the moments when governance controls face their hardest operating conditions: compressed timelines, reduced quorum, unclear authority chains, and high stakes. Organizations that design their control architecture for normal operating conditions and assume it will hold during transitions are making a structural bet that transitions are normal. They are not.

Business Engineer Framework

The Permission Layer: How Control Architecture Shapes What Organizations Can Do

The Automattic severance story is, structurally, a Permission Layer story — a case study in how the design of approval and authorization mechanisms determines what decisions can actually be made, by whom, and under what conditions. The Business Engineer Permission Layer framework maps how control architecture either enables or constrains organizational action. Understanding where approval authority sits, and whether the independence built into that authority is real or nominal, is foundational to reading any governance structure clearly.

Explore the Framework →

The Bottom Line

The Automattic reciprocal severance story is not primarily a story about money — the $8.15 million figure reported by TechCrunch is the outcome, not the mechanism. The mechanism is simpler and more general: when the set of people who approve a document and the set of people who benefit from it are identical, the approval step has been converted into a record-keeping step, and the independence that gives an approval its meaning has been removed. That structural property holds regardless of the amounts involved, regardless of the specific organization, and regardless of anyone’s intentions. Whether an approver sits genuinely outside a transaction is therefore the property that determines whether such a control is doing any work at all.


Source: TechCrunch — “Automattic’s interim CEO and legal chief signed reciprocal severance deals during Mullenweg’s brief ouster” (September 16, 2026). Structural analysis is original to Business Engineer / FourWeekMBA. This piece is not legal advice, not investment advice, and not an allegation of wrongdoing by any person.

91,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.

This is not legal advice, not investment advice, and not an allegation of wrongdoing by any person. Nothing above alleges that any individual acted in bad faith, breached any duty, or did anything improper or unlawful, and nothing above should be read as doing so. The facts described rest on HR documentation reviewed and reported by TechCrunch; the combined figure of $8.15 million is TechCrunch’s, and nothing above states that any amount has been paid, is legally owed, or will be paid. Automattic has been asked for comment — a pending request rather than a declination — and attempts to reach Mark Davies and Andy Missan were unsuccessful; no inference is drawn from either fact above. The circumstances of the leave of absence and the return are disputed by the parties, and nothing above takes any position on that dispute or repeats any accusation made by anyone. The observations that reciprocal approval conveys no independent judgement, that narrowing a definition of cause widens a severance entitlement, and that governance controls presuppose elapsed time are general properties of controls and of drafting. They are not characterisations of any individual’s conduct, state of mind or intentions, and nothing above says that any control was bypassed or that any board failed at anything. Nothing is predicted.

Scroll to Top

Discover more from FourWeekMBA

Subscribe now to keep reading and get access to the full archive.

Continue reading

FourWeekMBA