How to Hand Over Decisions Before You Step Away

Start from evidence, not the organization chart. The useful raw material is the trail of things that already wait for you. Look for approval requests sitting in private messages, choices you reversed after work had begun, recurring exceptions forwarded to your inbox, and customer commitments accepted without a delivery review the startup decision-rights matrix. Those are your candidates for the map.

The discipline is naming each item as a decision, not a topic. Pricing is too broad to test. Approve a non-standard annual discount is a decision, and a stand-in can be told whether they are allowed to say yes. One focused hour with your last two weeks of messages open is usually enough to build the inventory: list every question where the answer was "I'll check" or where someone visibly waited.

Sort each decision into exactly one bucket.

Sort by risk, not by importance. A useful set of criteria is reversibility, blast radius (how many customers, records or currency units can be affected), sensitivity (finance, security, employment, legal obligations), observability (will a poor outcome be detected quickly) and precedent (does the choice create a rule others will copy) the risk criteria from the same matrix. Move reversible, observable, low-precedent decisions close to the work. Keep high-consequence, slowly observable ones with you.

A worked example shows how much this frees. Suppose a customer-facing team logs around 60 decisions in a two-week absence. Under a founder bottleneck, perhaps 45 of them end up with you. Apply the bucket test honestly and the split usually looks like this: about 8 to 10 are genuinely yours (contract terms, a hire, a refund above policy, a security incident), 30 or more fit the delegated bucket with a written rule, and 15 or so belong on the waits list. Your job before departure is not to shrink the third column. It is to stop the first column swallowing the second.

A handover note that says "use your judgment" fails on day two. The matrix that works specifies, for every delegated decision, one default owner named as a role rather than a committee, the conditions under which they may act, who they must consult for evidence (advice, not a second veto), the escalation trigger, a response-time expectation and where the decision is recorded the seven fields described. Notice what the single named owner buys you: a committee cannot be held to a response time.

Ambiguous words are where handovers die. Significant, large, urgent and strategic cannot be applied consistently by a stand-in under pressure. Replace them with numbers you can defend even if imperfect, then pilot those numbers against your last ten real cases and adjust the operational-language guidance.

Here is a row to copy and adapt. It uses a hypothetical company, so substitute your own bands.

Decision: approve a customer credit after a service failure
Default owner: customer success lead
May decide when: incident is verified; amount within policy band
Must consult: finance, for ledger treatment (advice only)
Escalate when: amount above band; fraud suspected; contract terms conflict
Response time expected: same business day
Evidence recorded: CRM decision record linked to incident

Two details make this survive your absence. First, the consult is scoped to evidence, so a slow reply from finance cannot freeze the decision forever. Second, the response time creates a visible signal: if nothing is decided by the clock, the fallback owner takes it. Legal operations practice is blunt about why the fallback comes first in the design, not last: if you cannot name who acts when the first owner goes silent, the rule is not finished the escalation workflow analysis.

Separate roles within a row, even when one person holds several of them. A salesperson may prepare the commercial evidence while a sales lead decides; a product manager may prepare customer evidence while an engineering owner decides whether a reliability threshold permits a release the preparer, decider, executor and verifier split. This matters most where reversibility changes, because that is where a stand-in's confidence is lowest.

Delegation without an escalation route feels reckless to a founder, and that feeling is what makes people check their phone from the departure lounge. Give the stand-ins a legitimate way to reach you, then constrain it. A tiered structure works best: the default owner first, a cross-functional lead second, you only for the trigger conditions you wrote down tiered resolution levels. Everything below your tier should be resolvable without you.

Define triggers mechanically, not emotionally. A spend or commitment above a figure, a legal or safety flag, a security or privacy incident, a decision that contradicts a rule you already wrote, or a deadline that has passed its clock. The Association of Corporate Counsel, cited in that legal operations analysis, recommends a written delegation-of-authority policy that assigns approval rights to roles rather than to named individuals, so the arrangement survives staff turnover without a rebuild the ACC recommendation. A named fallback for each critical role, with contact details checked before you leave, closes the obvious gap.

The failure modes to design against are documented. Thresholds written into a policy document but wired into nothing that actually routes work mean nobody escalates until someone complains. A dead-end escalation names an owner who has moved teams. And approver fatigue sets in when thresholds sit too low, flooding senior people with matters that should have stayed at the first level, which trains the team to ignore the queue the three failure modes. Each one traces back to a missing component in your row.

Then pick one channel. Not two, not a mix of email, chat and a phone number. Announce it in writing along with the daily window in which you will read it. A twenty-minute window at a fixed hour is usually enough for a two-week absence and protects the point of the trip. Mental Health America's practical guidance points the same way: clearly communicate your availability, designate windows for questions and concerns, and delegate where you can MHA's burnout guidance.

Watch for the signs in yourself before you go. Mercury's work on founder burnout names three early alarms: emotional flattening, where wins and setbacks land with the same dull weight; cognitive clutter, where you reread messages and still miss details; and a growing pull to recheck decisions you already approved Mercury on the three alarms. If you catch yourself reversing a delegated decision informally, stop. Reversals with no stated reason teach the team that the authority is unreliable, and you will return to a queue that never shrank the reversal guidance.

Most of what feels urgent survives two weeks untouched. The waits list exists to give the stand-ins permission to do nothing on those items without guilt, and to give you permission not to hear about them. Keep it short and concrete: three to seven named items, each with one line on what happens on your return.

Good candidates for the waits list:

• Vendor renegotiations whose renewal dates fall beyond your absence.
• Non-binding exploratory partnership conversations.
• Internal process redesigns and tooling evaluations.
• Hiring steps already scheduled, where the panel can proceed to a shortlist.

Items that must not wait belong in the founder-only bucket, announced as such. A contract signature tied to your name, a regulatory filing, a serious safety or security event, a termination decision and anything creating legal exposure should have an acting authority where your governing documents allow it, and a clear note that the decision returns to you if it can be deferred.

The communication carries as much weight as the document. A short team note naming the acting decider, the escalation channel and window, the parked items and your return date removes most of the ambient anxiety that drives unnecessary messages. Holding on to a sense of agency is central to preventing burnout; Psychiatry.org's clinical guidance puts agency at the center of prevention, framed as self-efficacy and the sense that you can act on your circumstances the prevention guide. For a founder, agency over how the company runs during the absence is what turns a break into recovery rather than supervision with a different view.

You cannot hand over authority you have never checked anyone can use. Do a rehearsal. Pick two recent decisions that would now be delegated, walk the named owner through how the rule applies, and let them recommend the call. Then, before you go, let them make the next similar call independently and review the outcome afterward rather than re-approving it transferring authority through observed practice.

A short review script keeps the check fast.

Treat the first exception under the new rules as data. Review whether the boundary was too narrow, too broad or missing information, then adjust the row the three-pass installation method. The first pass documents who decides today, including the founder interventions that happen outside the official process. A false current-state map produces a ceremonial future state. The second designs the boundary from recent examples. The third runs it for a fixed period and reviews the decisions, reversals and exceptions.

Escalation volume is the metric to watch once you are back. Count the items that reached you, the time from request to decision, the re-escalation rate and the exception rate. Mature legal operations treat overdue volume, time to acknowledgment, time to decision, re-escalation rate and exception rate as the diagnostics for whether an escalation workflow works at all the metrics list. Come back to forty escalations and your thresholds were too low. Come back to silence on a problem that later surfaced and they were too high.

If you need a single test before walking out the door, use this. A decision can be delegated if a competent person with your written rule, and no access to you, would reach a defensible answer within one business day. Everything that passes goes to the delegated bucket with a named owner, a boundary, a trigger and a clock. Everything that fails goes on the waits list or the founder-only list, and you accept that it waits.

That rule has a useful side effect. The decisions that fail it are usually the ones where your judgment genuinely matters, and there are far fewer of them than a full inbox suggests. Writing them down makes the real size of your role visible, which is often the opposite of what a founder fears. Treat the map as a living document too. It should be reviewed when roles, service models or material commitments change, because the useful outcome is fewer ambiguous waiting decisions with appropriate accountability, not the largest possible matrix the review guidance.

If building the supporting mechanics (a decision log, an escalation form that routes to the right owner, a stand-in briefing page reused for every absence) is where your team keeps stalling, that is exactly the kind of internal tool a specialist can build with you and hand over ready to run: Quanta.

You will know the handover is ready when you can picture a specific person making a specific call on a specific day without you, and the answer feels acceptable rather than alarming. If it still feels alarming, one row is missing a boundary. Fix that row, not your willingness to leave.