Every Meeting Should Start Where the Last One Finished
Ten minutes into the weekly team meeting, you are still talking about last week.
Was the proposal sent? Which version of the document is current? Did anyone follow up on the question raised near the end of the call?
You have the agenda, the minutes and the AI summary. The missing piece is follow-through: nothing carries the decision into the work and back into the next meeting.
Someone thinks an action was completed. Someone else has a version of the document that may or may not be current. You end up making the same decision again because going round once more feels quicker than going back to look.
The record exists. The tracking usually doesn't
Most businesses already have the paperwork: agendas, minutes and, increasingly, an AI notetaker producing a tidy summary within minutes.
The summary captures what was said. Follow-through still depends on the actions reaching the right person, staying visible and returning to the next agenda if they remain open.
I use a meeting recorder and would not go back. It does the capture well. The rest belongs to the process around it, which is what I rebuilt in my own business over about three months.
The hour after every call, and the one home for what came out of it
My version of the same problem showed up after client calls. I have had a meeting recorder since I started, so a summary landed in my inbox within minutes. Then it sat there, because a summary tells you what was said, not what matters: which actions were genuinely mine, which were the client's, and which were really a question in disguise. A recorder cannot make that judgement because it does not know what the project is for.
So every call ended the same way: read the summary, work out what was real, move it somewhere it would be seen. About an hour, by hand, competing with the next call. The casualties were always the small conditional ones, the "I'll check whether that integrates and come back to you" said in passing at minute forty. Big commitments you remember without trying; it's the offhand ones that vanish.
The fix was deciding where an action lives. Before that, one could be in my notes, in an email, in my head, or on a list I kept meaning to tidy. Four places means no places, because none of them is the one anyone checks. So I picked ClickUp, put everything on it, and stopped treating meeting notes as a to-do list. The rule is what matters: automate a process you have not defined and the mess just arrives faster.
Ten o'clock, every night, and still not enough
Every evening at ten, a scheduled job reads the day's meetings, pulls out the meeting action items, and puts them straight into ClickUp. The recorder had not changed; what changed was where its output went, instead of sitting in my inbox waiting for me to deal with it. A schedule beats good intentions because it fires on the busy weeks too, when intentions don't. That fixed capture. Nothing agreed in a meeting has gone missing since.
But capture turned out to be half the job. ClickUp was accurate now, and it said things like "write the roadmap" or "update the delivery plan": true, and still a whole afternoon's work sitting there undone. The delay had just moved from "I forgot" to "it's on the list." The other gap: everything I had built looked backwards, so I walked into meetings on preparation written weeks earlier, fresh then, stale now. Which left a better question than the one I started with: what would it take for the write-up to arrive already done, and the preparation to be written on the morning it was needed?
The missing layer was judgement
That was when I built an agent and called her Mela. An agent is a written set of instructions and permissions: what she can read, what she can produce and what she must never do alone. She reads what I would have read before the meeting. When my process changes, I edit the instructions in the same way you would update an SOP for a person.
This stage needed an agent rather than another scheduled job because it needed judgement. What matters today depends on what was agreed last time, what has moved since and what the meeting is actually for. That is the same decision I had been making by hand in my inbox at the very start.
So Mela prepares each meeting the morning it happens, then drafts the thing the meeting was for: the roadmap, the updated scope, the decision record or the recap. I read everything before it goes out, so the client gets a proper follow-up without having to chase their own project to find out what happened next.
That review is the rule that makes Mela usable. When she gets something wrong, it is caught before it reaches anyone else because there is a person in the way by design.
The hour I used to spend catching up after a meeting now goes into the work that grows the business.
The order matters more than the tools
If your meetings run the way mine did, the order is the same whatever you build it with.
Do it yourself and time it. You cannot fix what you have not measured, and the number is usually worse than the estimate.
Decide where the output lives. One home, chosen once, before any tool is involved.
Schedule it. Most of the value is in it happening reliably, not in it being clever.
Automate last, and keep a person in the way of anything that leaves the building.
Most people try step four first, which is why so much automation gets built, used twice and quietly abandoned. It was asked to hold a process nobody had agreed on in the first place.
You can put the first three steps in place yourself. When the missing piece is judgement and ownership, that is what an online business manager is there to provide: someone who understands what the work is for and makes sure it moves forward.
What changes when the follow-up holds
Good follow-up saves the hour after a call and changes the next meeting too. Decisions remain visible, outstanding actions are clear, and the conversation can move forward without covering the same ground again.
You feel the difference when you walk into the next meeting knowing exactly where things stand. Structure creates freedom, and this is what it looks like on an ordinary Tuesday morning.
FAQ
Do you need to build an AI agent to fix this?
No. Most of the value sits in getting the basics right: timing the problem, choosing the one home an action lives in, and scheduling the capture. The agent only earned its place once that foundation was solid.
What if your meetings aren't recorded yet?
The order still works. Read back your own notes and time how long the follow-up actually takes, decide where an action lives, then put a recurring slot in the diary to move it there. The recorder speeds up capture; it is not what makes the system hold.
Are the same loose ends returning to every meeting?
Meetings are where loose ends become visible. The underlying gap sits between a decision being made and the system meant to carry it forward.
In a Clarity Session, we look at where that gap is showing up across your operations and decide what to fix first. You leave with a written roadmap in priority order. The session is £75, credited in full toward any work that follows.
