The L10® is the weekly heartbeat. The Quarterly and Annual sessions are the heart surgery. Two days off-site. Whole leadership team. Real money on the table. These are the meetings where Rocks get set, V/TO™ gets refreshed, the 1-Year Plan gets rebuilt, and the company commits.
Bringing Claude or ChatGPT into these sessions is higher stakes than bringing them into the L10®. Higher reward and higher risk. The default mistake is to under-use AI here. The second mistake is to over-use it.
Here is how to do it well.
Before the session: the prep pack a model can build
The single highest-yield AI work for an EOS® Quarterly or Annual is the pre-read. Most leadership teams arrive at the session under-prepared because the prep work takes hours nobody has.
A reasoning engine can build the entire prep pack in a few hours of focused work.
The Scorecard trend deck. Eight weeks of every Scorecard row, with deltas, trailing averages, and call-outs for any row that moved more than 20% from baseline. The agent reads the live Scorecard. Nobody types this up.
The Rocks retrospective. Each Rock from the closing quarter, with a Done/Not Done verdict, evidence trail, and a one-paragraph commentary on what worked and what did not. Drawn from the L10® notes and the milestone history.
The Issues List rollup. Every Issue from the prior quarter, clustered by root cause, with resolution status. Recurring themes flagged. The team sees patterns they would have missed reading the meetings individually.
The IDS index. Every major decision the team made in the quarter, with date, attribution, and outcome. Useful for Annual, indispensable for catching reversals.
The Vision vs Reality check. The 1-Year Plan from the current year, with current-state evidence for each goal. Green, yellow, red, with the agent's reasoning. The leadership team comes in already knowing which goals are at risk.
The 3-Year Picture™ resonance test. A short Claude or ChatGPT prompt: "Read our 3-Year Picture™ and our last quarter of company activity. Where are we drifting from the picture. Where are we surprising it positively." Useful for Annual.
The Integrator assembles the prep pack. Each leadership team member reads it before walking into the session. The session itself opens with everyone on the same page.
That alone can be the difference between a Quarterly that finds its rhythm by hour three and one that finds it by mid-day two.
During the session: the in-room rule
In-room AI use during a Quarterly or Annual is more restricted than during an L10®, not less.
Why. Because the Quarterly is where the team makes the largest commitments of the year, and those commitments need to come from the people in the room. The team's conviction in the Rocks they set is what carries those Rocks through the messy middle of the quarter. AI in the room dilutes conviction.
The rule we use: no model gets typed into during the Vision section or the Rock-setting section. During those sections the team writes by hand, on whiteboards, in their own voices. The model can be referenced for facts ("what was our Q3 churn") but never for choices.
After the Rocks are set, AI can format them, expand the Done states, and propose milestone breakdowns. The team reviews and edits. The commitment happened on paper, the AI helps articulate.
This is a subtle line. It is also the line that protects the Quarterly from becoming a meeting where the model proposes and the team accepts.
After the session: the propagation pack
Most Quarterlies leak in the 72 hours after the session ends. The team is exhausted. The Rocks get written but the milestones do not. The 1-Year Plan updates get half-typed. The decisions made on day two get half-remembered by Friday.
An agent built for post-Quarterly propagation removes this leak.
The agent reads the session whiteboards, the meeting notes, and any new V/TO™ artifacts. It produces:
- A clean version of each new Rock with a draft Done state and milestone breakdown for the owner to confirm.
- An updated V/TO™ document with the changes flagged for the Visionary to approve.
- A draft of the post-Quarterly email to the broader team announcing the new Rocks.
- A Scorecard delta list (any rows added, removed, or redefined).
- A list of new SOP work that came out of the session for the Process Component.
- A list of new agents proposed or new accountability lines drawn on the Accountability Chart.
By Friday after the Quarterly, the team has everything cleaned up and queued for execution. The new quarter starts in the new quarter, not in the middle of week 2.
Annual sessions: the cumulative read
The Annual session has one thing the Quarterly does not. A full year of context to reason over.
This is where a long-context model like Claude Opus or a similar tier of ChatGPT actually shines. Hand it the year's worth of Scorecards, Issues Lists, Rock retrospectives, and L10® notes, and ask it to surface the patterns the team would never see by reading any individual artifact.
"What were the three issues that came up in IDS most often this year." "Which Scorecard rows were consistently the leading indicator for the rest." "Which Rocks failed and what is the shared cause." "What did we say we would build in the 3-Year Picture™ and what evidence is there that we are building it."
This is not the model deciding anything. This is the model doing pattern matching across a year of data that no human could hold in their head at once. The leadership team uses the patterns as inputs for the next year's planning. The decisions still belong to the room.
For some teams this is the single most valuable AI use case of the year. It is also the one that requires the least technical buildout, because the model is just being given a folder of documents and a sharp question.
What this does not change
The Quarterly and Annual sessions are still two days off-site. Still the leadership team only. Still phones-down for the meaningful sections. Still facilitated by an EOS® Implementer® if you have one, or by the Integrator if not.
AI does not change the cadence, the format, or the rituals. AI changes the prep and the propagation. The room itself stays human.
FAQ
Should the Implementer use AI to prep the session? Yes. Most EOS® Implementers® are already doing this informally. Doing it deliberately produces better sessions.
Can the team review the AI prep pack in real time during the session? No, run it in the pre-read. The session is for discussion, not reading.
What about Annual sessions with AI-generated Rock proposals? Carefully. The model can propose Rock candidates from the V/TO™ 1-Year Plan and the prior year's data. The team chooses. Do not let the model's proposed Rocks become the default. Treat them as one input among many.
Does this require a custom-built agent or can it be done in the chat interface? For the first Quarterly, the chat interface works. Paste in the data, ask the questions, copy out the answers. By the second or third Quarterly, building a dedicated session-prep agent that runs against your live data sources saves real time.
EOS®, Entrepreneurial Operating System®, V/TO™, Vision/Traction Organizer™, Level 10 Meeting®, L10®, Rocks™, Scorecard, Issues List, IDS, 3-Year Picture™, 1-Year Plan, Quarterly, Annual, and EOS® Implementer® are concepts and trademarks of EOS Worldwide, LLC. This article is an independent practitioner perspective and is not affiliated with or endorsed by EOS Worldwide.