If your agency still runs its week on a stack of recurring Zoom calls, you’re likely burning a third of your productive hours before a single deliverable moves forward. At Choco Media, we’ve spent the last couple of years gradually replacing most of our async agency team meetings with structured async alternatives — and using AI to handle the parts that used to require everyone in a room at the same time. This post is about what that shift looks like in practice, what we kept live, and the AI-generated summary format that holds the whole system together.
This is written for small agency teams — five to twenty people, often remote or hybrid — who feel the weight of too many meetings but aren’t sure how to pull back without things slipping through. You’ll leave with a concrete meeting cadence, a set of async templates, and a summary workflow you can implement with tools you probably already pay for.
The context: we’re a small AI-first marketing agency in Rovaniemi. Our team works across multiple time zones and client accounts simultaneously. Coordination is genuinely necessary — we’re not anti-meeting for sport. But we found that most of our meetings were solving problems that better documentation and lightweight async tools could handle in a fraction of the time.
Why Most Agency Meeting Cultures Are Inefficient (and How They Got That Way)
Agency work has a natural pull toward synchronous communication. Clients have questions. Work is iterative. People like feeling in the loop. And early-stage agencies — the kind where everyone wears multiple hats — genuinely need real-time coordination. The problem is that the habits formed in a five-person startup rarely get redesigned when the team doubles.
What we typically see is a meeting stack that grew organically: a Monday kickoff, a mid-week check-in, project-specific calls that were supposed to be temporary, and a Friday review that nobody really wants but feels too risky to cancel. By the time you add client calls into the mix, most account managers are spending 15–20 hours per week in scheduled video calls.
- The kickoff is largely a status update that could be a written report
- The check-in exists because no one trusts async handoffs yet
- The Friday review rehashes information that was already in Slack
- The “project-specific” calls keep recurring because no one owns the decision
The root cause is almost never laziness — it’s a documentation gap. When you don’t have a reliable system for capturing decisions, surfacing blockers, and moving work forward without a meeting, meetings become the fallback. The fix isn’t discipline, it’s infrastructure.
The Async-First Principle: What It Actually Means
Async-first doesn’t mean async-only. It means that synchronous time — the kind where everyone has to be present at once — is reserved for things that genuinely benefit from it: creative direction decisions, complex client strategy conversations, team tension that needs to be resolved in real time. Everything else defaults to async.
In practice, async-first requires two things that most agencies skip:
- A shared single source of truth — one place where work status, decisions, and open questions live, that everyone checks before asking a question in Slack or scheduling a call
- A writing culture — the habit of documenting decisions and context in real time, not retroactively, so that people joining a thread at noon can catch up without needing a thirty-minute briefing
The good news is that AI makes both of these dramatically easier to sustain. The meeting summaries, the status updates, the “here’s where we landed” write-ups — these used to require discipline and time. Now they require a prompt and thirty seconds.
What “Async-First” Doesn’t Mean
It doesn’t mean you never have meetings. It doesn’t mean you’re always waiting for a reply. And it doesn’t mean you treat every question as requiring a written essay. Some things — especially anything involving emotional nuance, creative tension, or a client relationship under pressure — are genuinely better live. The skill is learning to distinguish between the two.
Our Meeting Cadence: What Stayed Live and What We Cut
After iterating for about eighteen months, this is the cadence we settled on. It’s designed for a core team of six to twelve, with a mix of client-facing and delivery roles.
Weekly: One Live Sync (45 Minutes Maximum)
Monday mornings, we run a single weekly sync. The agenda is fixed and takes less than ten minutes to prepare: active client accounts, top three priorities per account, blockers, and any decisions that need group input. The key constraint is that this call is for decision-making and direction-setting only — no status updates, because those live in our async standup thread.
- Duration: 45 minutes hard limit
- Facilitator rotates weekly
- AI summary posted to the team channel within 15 minutes of close
- Action items extracted automatically with owner and deadline
Daily: Async Standup in Writing
Each working day, everyone posts a short structured update to a dedicated Slack channel by 9:30 AM Helsinki time. The format is three fields: what I’m working on today, what I finished yesterday, anything blocking me. This takes two to four minutes per person and replaces what used to be a 30-minute daily standup.
We use a simple Slack workflow to prompt the update each morning. No AI here — the format is human-written intentionally, because the texture of how someone describes a blocker tells you more than a generated summary would.
Biweekly: Deep Work Review (60 Minutes, Live)
Every two weeks, we run a slightly longer live session focused on work quality, not status. We look at recent deliverables together, discuss what we’d do differently, and talk about any systemic issues in client accounts. This is the session we protect most carefully — it’s where standards are maintained and shared craft knowledge is built.
Monthly: Strategy and Operations (90 Minutes, Live)
Once a month, we step back from client work and look at the agency itself. Pipeline, capacity, pricing, what’s working and what isn’t. This one stays live because it often involves sensitive discussions and requires collective judgment calls that don’t translate well to async threads.
“The goal of the cadence is to protect the live sessions for things that actually need live sessions. Once you have that clarity, it becomes much easier to push back on calls that don’t need to happen.”
The AI Summary Workflow: How We Use It and What We Prompt
This is the part most teams underestimate. A good AI meeting summary isn’t a transcript — it’s an edited, structured document that surfaces what matters and discards what doesn’t. Getting there requires a specific workflow, not just dropping a transcript into ChatGPT.
Step 1: Record and Transcribe
We record all live meetings via Fathom (free tier covers most small teams) or Otter.ai. Both produce timestamped transcripts with speaker labels. The transcript is the raw material for everything that follows. We don’t rely on memory or notes — the transcript is the source of truth.
Step 2: Run the Summary Prompt
Immediately after the call, the facilitator pastes the transcript into a Claude prompt. The prompt we use is structured around four outputs:
- Decisions made — explicit choices that were agreed on, with enough context to be understood out of conversation
- Action items — each one with an owner name and a deadline, formatted as a checklist
- Open questions — things that were raised but not resolved, that need a follow-up in the async thread or the next meeting
- Context for absent teammates — a two-paragraph plain-English summary of what was discussed and why, written for someone who wasn’t in the room
The full prompt takes about 45 seconds to run and produces a structured document that’s posted to Slack before most people have closed their laptops from the meeting.
Step 3: Human Review (Two Minutes)
The facilitator reads the summary before posting. AI summaries occasionally misattribute a decision or miss a nuance in a complex discussion. The review catches these before they propagate. We don’t edit heavily — just flag anything that’s materially wrong and add a note if needed.
Step 4: Post and Archive
The summary goes to the Slack channel immediately. It also gets copied into Notion under a running Meeting Notes page for the relevant client or team thread. The Notion version becomes the searchable archive — six months from now, when someone wants to know why we made a particular decision, the answer is a search query away.
Tools We Use in This Stack
You don’t need an expensive purpose-built platform for this. The tools we use are things most agencies are already paying for:
- Slack — async standup channel, meeting summary distribution, Slack workflows for standup prompts
- Notion — single source of truth for client work, meeting archives, open questions
- Fathom or Otter.ai — transcription. Fathom is free and good enough; Otter.ai has better speaker diarisation for larger teams
- Claude or ChatGPT — the summarisation step. Claude handles long transcripts well and produces more editable output than most alternatives
- Google Meet or Zoom — the live calls
The total additional cost of this workflow on top of what most agencies already pay is approximately zero. The investment is in setup, habit formation, and the ten-day adjustment period while the team shifts from talking to writing.
If you’re looking to go further with automation across your operations, our AI automation service covers where this kind of meeting workflow sits inside a broader operational stack — from client reporting to content approval flows.
How to Handle the Transition: The First 30 Days
The hardest part of going async-first isn’t the tools — it’s the social adjustment. People who are used to being heard in a room can feel invisible in an async-first culture if the transition is handled poorly. Here’s what we’d do if starting from scratch.
Week 1: Audit Your Current Meeting Stack
List every recurring meeting. For each one, answer three questions: What decision or outcome does this meeting produce? Could it be replaced with a written update? If not, could it be shorter? You’ll usually find that about half your meetings fail the first question entirely — they don’t produce a decision, they’re coordination theatre.
Week 2: Introduce the Async Standup
Start with the daily standup — it’s the lowest-stakes change and the one with the fastest visible benefit. Set up the Slack channel, create the workflow, and run both formats in parallel for two weeks so the team can see that the async version captures the same information. Then cut the live one.
Week 3: Add AI Summaries to Your Remaining Meetings
Before eliminating any more live meetings, add AI summaries to the ones you keep. This builds the muscle of structured documentation before you depend on it. It also immediately makes your live meetings more valuable — knowing a summary will be posted makes people more decisive in the room.
Week 4: Cut or Consolidate
With the standup async and summaries running, look at the rest of your meeting stack again. Most teams can eliminate one to two recurring meetings entirely in this phase. Consolidate where you can — two mid-week check-ins become one, a Friday review rolls into the Monday sync.
Common Failure Modes to Avoid
We’ve tried versions of this that didn’t work before landing on the current setup. The failure modes are predictable:
- No accountability for async updates — if the standup posts are optional in practice, they stop happening. Someone needs to own the cadence and nudge people who miss it
- AI summaries that nobody reads — this happens when the output is too long or too vague. Keep it tight: decisions, actions, open questions. Cut anything that isn’t one of those three
- Async as avoidance — some teams use “async-first” as cover for not giving feedback or not making decisions. The biweekly deep work review and monthly strategy call exist precisely to prevent this
- Tool fragmentation — if some people use Notion, some use Google Docs, and some reply in Slack threads, there’s no single source of truth. Pick one and enforce it
For more on the operational frameworks behind a well-run small agency, our post on how a small AI-first agency runs in 2026 covers the broader picture — from how we structure delivery to how we filter clients before they become problems.
What Good Looks Like: Signs the System Is Working
After a few months of running this way, you’ll notice a clear shift in how the team communicates. The signs the system is working:
- People check the Notion page before asking a question in Slack
- Decisions are traceable — anyone can find out why something was decided, and when
- The Monday sync ends in under 40 minutes because the prep was done async
- New team members onboard faster because the written record gives them context that would otherwise require hours of conversation
- Clients notice that the agency is more coordinated and responsive, even though internal meeting time has dropped
The last point matters. Async-first isn’t just a productivity change — it produces better client work because people aren’t context-switching every 45 minutes. Protected deep work time raises the quality of what gets delivered.
If you’d like help thinking through what this looks like for your specific team setup, or if you’re curious about how we’ve built these processes into our client work, reach out — it’s one of the areas we’re happy to talk through in detail.