Stream a simulated run, inspect the notifications it would send on Slack and email, and see exactly where it sits in the 7-stage PM OS flow. No password required.
The short version
The Board Narrative Drafter agent turns the most expensive document in product management, the quarterly board update, into a ninety-minute edit instead of two to four full days. It ingests the quarter's data from billing, CRM, product analytics, the migration tracker, and the eval system, runs it against the CFO-agreed comp set (transition peers, not pure-SaaS), and composes the four-section narrative the board has agreed to. The CPO does the editorial layer: which moments matter, which framing works, what to cut. Start by writing the two YAML files this agent needs, board-data-sources.yaml and comp-set.yaml. That alone saves a day next quarter.
The agent that turns the most expensive document in product management into a 90-minute edit
Most CPOs spend two to four full days per quarter writing the board update. That's eight to sixteen days a year. Sixteen days writing slides.
Worse, the board update is mostly the same shape every quarter: leading indicators, transition narrative, what changed, what's next. The data sources are the same. The structure is the same. The audience is the same. Only the specific numbers and the specific events change.
This is exactly the work that should be agent work, with the CPO doing the editorial layer on top.
What the drafter does
Four jobs.
- Ingest the quarter's data from product analytics, billing, customer health, and the migration tracker.
- Run the data against the agreed comp set (the one negotiated with the CFO; pure-SaaS comps are wrong for transitions).
- Compose the narrative in the format the board has agreed to: trough position, leading indicators, what changed, what's next. Plus named call-outs for customers, salespeople, and product wins.
- Hand the CPO a draft to edit, with all the source data linked so the CPO can verify any number in one click.
The CPO turns four days of compilation into ninety minutes of editing.
The seven components
1. Data registry. A YAML file (board-data-sources.yaml) listing every data source the board update needs: revenue (billing), NRR (CRM + billing), gross margin (FP&A + billing), outcome volume (product analytics), migration percentage (the migration tracker agent), customer references (CRM), agent quality scores (the eval system), competitive moves (the competitive intel agent).
2. Quarterly snapshot collector. Pulls each data source as of quarter-end. Caches results. 60 lines of Python.
3. Comp set ingester. A separate file (comp-set.yaml) with the agreed set of public companies in transition (Sierra, Intercom Fin, HubSpot AI, Atlassian, etc.) plus the metrics tracked for each. Updated when public earnings ship. The drafter compares the company's metrics to the comp set's, not to pure-SaaS averages.
4. Narrative composer. The longest LLM call in the agent. Takes the quarterly snapshot, the comp set, the agreed format, and the previous board update (for continuity). Composes the narrative section by section.
5. Named call-outs collector. Pulls customer wins from CRM, sales call-outs from the sales leaderboard, product wins from release notes. Filters to the most board-worthy. Drafts a "named acknowledgments" section.
6. The leading indicators dashboard. A separate generator that produces the chart pack: trough curve, migration percentage, outcome volume trend, gross margin per outcome, agent quality, dispute rate. Charts are embedded in the slide deck.
7. The CPO edit interface. The draft lands in a shared doc with: the narrative, every number linked to its source, the chart pack as embedded images, and the named call-outs section. The CPO edits, the data is already correct, the rhetoric is the CPO's contribution.
The narrative composition prompt
This is where the agent earns its keep. The prompt takes the quarterly data and produces a board-ready narrative.
You are drafting a board update for ${company_name}, a ${stage} company in the middle of a SaaS-to-Service-as-Software pricing transition.
The agreed format has four sections:
1. Trough position (where we are on the gross margin recovery curve)
2. Leading indicators (the seven-metric table)
3. What changed this quarter (named events: customer wins, product launches, key hires, competitive moves)
4. What's next (the next quarter's commitments)
Comp set for this update: ${comp_set}.
Pure-SaaS averages should NOT be cited; the comp set is the right frame.
Quarterly snapshot:
${quarterly_data_json}
Previous quarter's narrative (for continuity):
${previous_narrative}
Compose the four sections in the company's voice (calm, specific, owns the trough, leads with leading indicators, references named customers and team members where appropriate). 800-1200 words total.
Where any number could be questioned by a board member, include the source link in markdown.
Return as markdown.
The output is a draft that needs editing, not a finished product. But it removes the blank-page problem and keeps the narrative consistent across quarters.
What this changes for the CPO
Without the agent, the CPO writes the board update from a blank page every quarter. Half the time is spent reconstructing data (where's the gross margin number, what's the migration percentage, what was the dispute rate). The other half is the actual narrative.
With the agent, the data is right. The narrative is structured. The named call-outs are identified. The CPO is doing the editorial work that actually requires their judgment: which moments matter, which framing works, what to cut.
Two days of compilation becomes ninety minutes of editing. The compounding effect over a year is enormous.
What to try this week
If you're not yet running this agent, do the simplest version: write the comp-set.yaml and the board-data-sources.yaml files. By hand. Just the index. List every data source the board update pulls from. List every company in your agreed comp set with the metrics you track for each.
The two files are 80% of what you need to compose the agent. The rest is plumbing. But you can't write the YAML files until you've thought through what your board update actually needs, which is itself useful.
If you're already writing your board update from a blank page each quarter, this exercise alone will save you a day in the next cycle.
The full agent blueprint, including the YAML schemas, the LLM prompt template, and the chart-pack generator, is at /artifacts/agent-board-narrative-drafter. The companion Board Narrative Slide Outline (the agreed format) is at /toolkit/board-narrative-slide-outline.
Related
- The PM AI Agent Fleet, the 45-agent operating system this agent slots into.
- The Investor and Board Narrative for AI-Era SaaS, the format this agent drafts toward.
- Board Narrative Slide Outline, the agreed format.
Download the artifact
Ready to use. Copy into your project or share with your team.
Also on Medium
Full archive →AI Agents and the Future of Work: A Pixar-Inspired Journey
What product managers can learn about AI agents from how Pixar runs a film team.
Many AI Agents Are Actually Workflows or Automations in Disguise
How to tell agents from workflows from cron jobs, and why it matters for what you ship.
Frequently asked
What does the board narrative drafter agent do?+
Ingests the quarter's product, financial, and customer data, runs the trough narrative through the CFO-agreed comp set (transition peers, not pure-SaaS), composes the four-section quarterly board update, and hands the CPO a 1,200-word draft to edit instead of a blank page.
How long does a quarterly board update normally take a CPO?+
Two to four full days per quarter. Sixteen days a year. The drafter agent collapses that to ninety minutes of editing once the data sources, comp set, and format are wired up. The agent does the compilation; the CPO does the rhetoric.
What data sources does the agent ingest?+
Revenue and gross margin from billing, NRR from CRM + billing, outcome volume from product analytics, migration percentage from the migration tracker agent, agent quality scores from the eval system, customer references from CRM, competitive moves from the competitive intel agent.
What does the CPO actually edit when the draft lands?+
The narrative voice, the named customer call-outs, the framing of what changed, and the three commitments for the next quarter. The data is already correct; the CPO's job is the editorial layer that makes the trough sound boring instead of catastrophic.