2026-05-15
May 15, 2026
Most recurring meetings exist because the team never built the artifact that would have made the meeting unnecessary. Friday is the right day to audit them.
Some part of you already knows it. The hour-long staff meeting that exists because the team can't read each other's writing. The 30-minute design sync that exists because the doc never circulated. The weekly project review that exists because nobody trusts the status field. Every recurring meeting on your calendar is a confession of an artifact your org chose not to build — and most senior tech leaders are not auditing that calendar the way they audit their AWS bill.
The cost is bigger than it looks. The average engineering org in 2026 has more meetings on the calendar than in 2020. Despite five years of remote-work normalization, the industry didn't learn to write more; it learned to bolt sync-ups, check-ins, and "alignment" ceremonies onto the front of every Slack thread. The bill, in any honest accounting, is the largest line item in the run-rate of a senior IC: half their working hours, billed at staff+ comp, spent in rooms whose only output is a Slack DM and a shared sense of fatigue.
The fix is not "fewer meetings." It is to separate the three jobs meetings do — broadcast information, make decisions, build relationships — and run each one through the cheapest medium that works. Information goes async. Decisions go DACI. Relationships stay sync but get smaller. This is the operating cadence the writing-first orgs (Stripe, GitLab, Automattic, Amazon, the better parts of Microsoft) have been running quietly for years. The rest of the industry is finally noticing.
The category mistake at the heart of broken meeting culture is treating "meeting" as a single thing. Three different jobs share the calendar slot, and the failure mode is mixing them in the same room.
Job 1: Broadcast information. "Here's where the migration stands. Here's the program against milestones. Here's a customer question that touches four teams." This is the biggest category of meeting and also the worst use of synchronous time. A written weekly does the same work at 5–10x leverage. One leader writes for 60 minutes; ten readers consume for 6 minutes each on their own schedule; the org has a searchable artifact six months later. The standing status meeting, run for the same audience, costs ten times the hours and leaves nothing behind. If a meeting's primary job is to broadcast information, kill it and write the memo. The only counter-argument anyone raises — "people won't read it" — is a culture problem the leader is not solving by holding more meetings; they are subsidizing it.
Job 2: Make a decision. This is where the synchronous room earns its cost — and where most leaders waste it. The decision meeting should be short (30 minutes is usually too long), structured (the artifact arrives 48 hours ahead, not in the room), and roled (someone is the Driver, someone is the Approver, contributors are named, the rest are informed). This is exactly what DACI — Intuit, 1980s; popularized by Atlassian — was built to enforce. The mistake most senior leaders make is running decision meetings as information meetings: opening with status, drifting through context, surfacing the actual call in the last five minutes, then deferring it because nobody pre-committed in writing. A DACI-structured decision has the call pre-articulated in the artifact; the room is there to surface the last 10% of risk and capture the Approver's signature. Forty-five minutes becomes fifteen.
Job 3: Build relationships. Trust, mentorship, the social fabric that makes the first two jobs feasible at all. Camille Fournier's audit, Which Meetings Should You Kill?, found — counter-intuitively — that the bloat in most senior calendars isn't standups or status reviews. It is 1:1s. Line managers meet with directs, with their own manager, with peers, with the PM, with the designer, with recruiting, with skip-levels; staff+ engineers meet with every member of their team, with peer staff engineers, with neighboring orgs. Each one was a good idea individually. In aggregate they are the line item to audit. Some 1:1s are real relationship work; some are async updates wearing 1:1 clothes; some have outlived the project that spawned them and should be canceled.
Put the three jobs together and the operating cadence becomes legible:
The 2026 wrinkle is that AI agents are about to make the bad version of all three categories worse. Auto-generated meeting recap emails subsidize the "I'll attend and catch up later" anti-pattern. AI-drafted status memos hide the rot in the underlying work. AI summaries of 1:1s make the meeting feel productive without producing the trust it was meant to produce. The leaders who scale through this transition use AI to strengthen the artifact, not to apologize for the meeting. An AI-assisted DACI memo is cheaper to write, easier to circulate, and reaches more stakeholders than the human-only version did three years ago. An AI-summarized status meeting is, mostly, an expensive way to launder bad writing.
There is a Goodhart warning at the bottom of all this. The metric to chase is not "meetings reduced." It is "decisions made per week, per team." Some orgs cut meetings 40% and decision throughput goes up; some cut 40% and decisions get bottlenecked because nobody owned the cadence shift. The decision is the unit of leadership output. Meetings are a means. The discipline is to audit the cadence against the decision throughput, the same way you audit your DORA pair against shipped value. If you cannot articulate what decisions your weekly meeting block is producing, the block is the wrong size.
Try this week. Pick your three most recurring meetings. For each, classify the primary job: information, decision, or relationship. For information meetings, draft the written-weekly that would replace them and ship one. For decision meetings, retrofit DACI: who's the Approver, what's the artifact, what's the 48-hour pre-read? For relationship meetings, ask if they would survive being downgraded to monthly. You'll usually find at least one block per week you can reclaim cleanly.
What it is. A four-role framework for any decision involving more than one person. The Driver leads the process, gathers input, and prepares the artifact. The Approver — exactly one named individual — makes the final call. Contributors are domain experts who provide input but don't own the decision. Informed are the people who need to know once the decision is made. DACI's lineage runs through Intuit in the 1980s as a variant of RACI; Atlassian formalized it for product and engineering orgs and made it the canonical Team Playbook decision tool.
When to use it. Any decision that has stalled in a meeting, been re-litigated in three Slack threads, or surfaced the wrong question — "should we?" — when the right question was "who decides?". Especially useful when a peer org is treating the question as theirs, when a directionally important call is bouncing between five informed-and-opinionated people, or when an executive is reluctant to be the named Approver.
How to run it:
When NOT to use it. Reversible, low-blast-radius calls a single person can make and walk back if wrong — running DACI on every choice is its own anti-pattern. Also: do not run DACI on Type 1 (one-way-door) decisions where the Approver genuinely needs days of synchronous deliberation with peers — those are board-style decisions, not DACI decisions.
Worked example. A platform team argued over which CI runner to standardize on for six months across three Slack channels and four meetings. The director ran a DACI: Driver was the platform lead, Approver was the director themselves (not "the leadership team"), Contributors were three principal engineers, the Informed list was every team that would migrate. One memo, 48-hour pre-read, fifteen-minute meeting to resolve the last objection, decision in writing. Total elapsed time: nine days. Previously-elapsed time: twenty-six weeks.
Stand-Up Meetings Are Dead (and What to Do Instead) — Honeycomb's engineering team on replacing the daily standup with async written updates plus a focused weekly sync. The post is a couple of years old but the playbook is exactly the cadence 2026 async-first orgs are converging on.
Companies Using RFCs or Design Docs — and Examples of These — Gergely Orosz's running map of which engineering orgs have institutionalized the RFC as the decision artifact, with template examples from Uber, Stripe, GitLab, and others. The right starting point if you're trying to install the practice on a team that doesn't have it.
A thorough team guide to RFCs — LeadDev's practitioner guide. Pair with the Pragmatic Engineer map for one-day implementation: read this for how, read Pragmatic for who.
"Meetings are by definition a concession to deficient organization. For one either meets or one works. One cannot do both at the same time."
— Peter Drucker, The Effective Executive (1966)
Don't miss what's next. Subscribe to Critical Path: