The two people who could never be gone at the same time

Two founders, married, running the agency together. From outside, that reads as redundancy — two decision-makers, two sets of judgement, twice the coverage.

The business had two founders. It did not have two independent operating systems.

Responsibilities were divided; decision authority was not. One founder held client relationships, commercial decisions and business development. The other held creative delivery, production and the team. The boundary held right up until a decision crossed it, which happened most days.

A client requests a scope change. The account lead goes to Founder A. Founder A needs a delivery assessment, so Founder B is brought in. A staffing question affects a deadline: both consulted. A pricing exception affects margin: both approve. Someone is unsure whether a thing is even worth escalating, so they ask one founder, who checks with the other.

The business was not waiting on a founder. It was waiting on two people to reach agreement — and they were the same two people at dinner that evening.

Over a four-week sample, 34% of recurring operational decisions were escalated to at least one founder, and 11% required both. Combined, the founders were spending an estimated thirty hours a week on approvals and escalations that did not need founder-level involvement.

Of 23 critical recurring workflows, 9 had documented procedures. The remaining 14 lived in individual knowledge — client preferences, project history, vendor relationships, pricing exceptions, delivery nuance, held in heads, inboxes and private conversations. Seven client relationships carried material context known primarily to one founder.

There was no escalation matrix. The team knew who to ask. They did not know when they were expected to decide instead, what authority they held, which founder owned what, or what to do when the relevant founder was unreachable. Uncertainty became escalation by default.

Because both founders were highly capable, the system appeared to work. It appeared to work right up until either of them was unavailable.

Decision rights were mapped and assigned by level — team, functional lead, Founder A, Founder B, both — with a single governing rule: the lowest competent level makes the decision. Founder involvement became an exception rather than the operating model.

Then the founders were separated as an approval layer. Each took explicit ownership within their domain. Where a decision genuinely required both, that threshold was written down. Everything else got one owner. The question the team asked changed from "which founder should I ask" to "who owns this."

Escalation architecture followed, classifying every recurring escalation: solvable by the team, requiring functional expertise, materially affecting client, margin, legal exposure or strategy, or genuinely requiring both founders. The path stopped terminating automatically at the top.

The 14 undocumented workflows were prioritised by frequency, financial impact, client impact and key-person risk, then written — onboarding, kickoff, scope changes, pricing exceptions, resource allocation, escalation, complaints, vendor management, invoicing, closeout, staff absence, delivery handoffs. Every critical responsibility received a primary owner, a backup owner and an escalation threshold. The point was never to remove the founders from important relationships, only to ensure no important relationship existed exclusively inside one person's head.