How can I use AI to help my team prep for meetings with solid data?

Why meetings run on stale or missing data

The Collective 54 essay on EOS in the AI era says scorecards were built for a world where data was scarce, delayed and manually assembled. Weekly metrics were a breakthrough in 2007. It says data is now abundant and real-time, leading signals can be surfaced continuously, and when leadership teams wait for weekly or monthly scorecards they are already behind. It also warns that when the wrong metrics are in front of a team, meetings become busy rather than informative, and leaders debate symptoms instead of causes.

The operations essay in the newer book describes what that looks like day to day in a boutique. Decisions are made but not enforced. Decisions resurface repeatedly because no one held on to them. Meetings produce agreement but not momentum. The founder becomes the escalation point for everything, because the founder is the one who remembers.

As an inference, most meeting preparation fails for one of two reasons: nobody assembled the facts, or nobody remembered what was already decided. AI can now do both.

What AI should own

The operations essay assigns AI the work that overwhelms people: decision memory, meaning what was decided, why, and what was deprioritized; commitment tracking, meaning who committed to what and whether it happened; drift detection, noticing when priorities, scope or timelines quietly erode; cross-functional visibility across sales, delivery, finance and people; and pattern recognition that surfaces recurring execution failures. It keeps judgment, interpretation, escalation, people leadership and final accountability with humans.

The finance essay describes the same split for numbers: an AI capability inside the firm that ingests sales, delivery, payroll and cash data, reconciles it, reports continuously and flags variances, with people supplying judgment and benchmarks. As an inference, that is the data layer every meeting should draw on, rather than someone building slides the night before.

Build a prep pack, not a slide deck

As an inference from both essays, a useful pack for any internal meeting has four parts, kept short enough to read in five minutes.

The numbers that matter for the decisions on the agenda, with their trend, pulled straight from source systems. The scorecard answer on this site argues for economic measures such as revenue per head, EBITDA per head, pricing realization and margin by client and project, rather than activity counts.

The record: what was decided at the last meeting, what each person committed to, and whether it happened, taken from the decision log rather than from memory.

The open issues: anything the system flagged since the last meeting, such as a project running over budget, a slipped commitment or a client gone quiet, each with the evidence behind it.

The decisions needed: one or two questions the meeting must answer, with the options and their trade-offs set out.

Send it ahead. The client experience chapter of the 2020 book asks whether a firm sends prereading to clients in enough time; the same courtesy applies inside the firm.

Prepare client and sales meetings the same way

The account management essay describes AI surfacing account history, analyzing client communications and extracting decision maker preferences, and the AI for account managers answer on this site turns that into a briefing before every client meeting. The account executive essay says recorded buyer conversations can be analyzed to show what a buyer has actually said, whether a trigger exists and what evidence is still missing, and the sales playbook answer covers using that to prepare each call. The 2020 book asks whether you research meeting attendees before each meeting. As an inference, the same four-part structure works for all three: facts, record, open issues and the decision you want.

Capture decisions as they are made

The operations essay says the system should capture decisions as they are made, retain the rationale, and distinguish commitments from discussion. As an inference, the prep for the next meeting starts at the end of this one: record the decisions and commitments before people leave, and let the system check progress in between. The EOS essay says meetings in an AI-native firm still exist but become fewer, shorter and more deliberate, with reporting shifting from backward-looking summaries to real-time visibility.

Get the data right first

The EOS essay warns that AI-led governance used too early fails, and that AI cannot compensate for unclear priorities, weak leadership or unresolved accountability. As an inference, three steps come before any automation. Agree the handful of numbers each meeting needs and who owns each one. Connect the systems that hold them, such as accounting, time records, the CRM and project budgets, so the figures assemble themselves. And run the pack manually for a few cycles until the team trusts the numbers. A prep pack built on data nobody trusts produces the debate it was meant to prevent.

Keep client and financial data inside tools the firm has approved; the client data protection answer on this site covers that.

Keep a person accountable for the pack

As an inference, AI drafts can be confidently wrong, so name one person for each recurring meeting who reads the pack before it goes out and corrects anything that looks off. Over time the corrections show where the data is weak.

Start with one meeting

As an inference, do not try to change every meeting at once. Pick the leadership meeting, because it is where decisions most often resurface and where the founder is most often the memory. Build the pack for it for a month, keep a running log of decisions and commitments, and check at the end whether fewer topics came back and whether the meeting got shorter. Then extend the same approach to client reviews and sales calls.

What we do not prescribe

Collective 54 names no meeting software or AI tool and publishes no agenda template or required cadence. The published positions are real-time data making scheduled scorecards late, meetings that become busy without the right metrics, economic measures over activity measures, decisions that resurface and agreement without momentum, AI owning decision memory, commitment tracking, drift detection and visibility while people own judgment and accountability, continuous finance data, client and sales preparation from recorded conversations and account history, prereading and researching attendees, fewer and shorter meetings, and the warning against automating governance too early.

When this answer flips

If your time, billing and pipeline data are not yet kept consistently, as an inference, fix that first; AI will assemble bad data faster.

If the meeting has no decision to make, the best preparation may be canceling it and sending the pack instead.

And if your team is small and meets daily, a short shared list of decisions and commitments may be all the system you need.

The short answer

Let AI assemble the facts and the record before every meeting, so people arrive informed and use the time to decide. The EOS essay says data is now real-time, so waiting for a meeting to see the numbers means being behind, and the operations essay says AI should hold the memory of decisions and commitments that otherwise keep resurfacing. Build a short pack with the numbers that matter, what was decided and promised last time, the flagged issues and the decisions needed, and send it ahead. Use the same structure for client and sales meetings. Agree the metrics and connect the sources first, and keep a person responsible for checking every pack.