Executive & Administration

agenda → pre-read → live notes → decision log → action items → follow-through tracking

• Draft
  Agenda & pre-read assembly — AI drafts the agenda from open threads and compiles the pre-read from source docs and metrics · verifiable against sources; choosing what makes the agenda is a power decision the CoS owns
• Automate
  Live capture — AI transcribes and produces structured notes tagged with decisions, owners, and open questions · fully verifiable against the recording, pure repetition, errors caught immediately by attendees
• Draft
  Decision log maintenance — AI extracts decisions into a canonical log with context and dissent noted · what counts as "decided" is contested more often than you'd think; a human confirms before it becomes the record
• Avoid
  The decision itself — weighing the tradeoff and committing the org · this is the accountability axis in its purest form — an executive who lets a model tip a close call has no answer to "why did you decide this"
• Automate
  Action-item chasing — agent tracks owners, sends nudges, and rolls status into next week's agenda · high volume, rule-based, and the worst failure is a redundant reminder
• Assist
  Follow-through synthesis — AI flags decisions that quietly stalled or got relitigated · pattern detection helps, but calling out a stalled peer is political work

inbound stream → triaged queue → drafted replies → scheduled commitments → daily brief

• Automate
  Triage & prioritization — AI sorts the inbox by required action and flags the five things that actually need the exec today · misfiled email is instantly recoverable, volume is enormous, and the exec's glance is the spot-check
• Automate
  Routine replies — scheduling confirmations, intros with approved templates, acknowledgments · low stakes, template-verifiable, high repetition; anything off-template escalates
• Draft
  Substantive replies — AI drafts responses in the exec's voice for the exec to edit and send · the words commit the executive personally; a sent email is irreversible and everyone assumes the exec wrote it
• Avoid
  Sensitive communications — replies touching personnel, deals, disputes, or press · maximal stakes and relationship nuance, and an AI-drafted misstep here is an unforced error with a signature on it
• Draft
  Calendar defense — agent applies stated priorities to accept/decline/negotiate meeting requests and protect focus blocks · rules cover 80% but declining the wrong person is a relationship cost the EA prices better than any model
• Automate
  Daily brief — AI compiles tomorrow's schedule with per-meeting context, open items, and travel logistics · pure retrieval and assembly, verified in thirty seconds each evening

message intent → draft → stakeholder review → published comms → Q&A follow-up → sentiment read

• Draft
  Message drafting — AI turns the exec's voice-memo intent into a structured draft in their established voice · verifiable against the exec's intent and prior writing, but the org reads these words as the leader's own — ownership is the point
• Automate
  Channel adaptation — one message becomes the Slack post, all-hands slide, email, and manager talking points · mechanical transformation of an approved source, fully checkable
• Avoid
  Sensitive announcements — reorgs, departures, layoffs, missed targets · irreversible trust stakes and legal exposure; AI's fluency actively hurts here because polished-generic reads as evasive exactly when authenticity is the job
• Draft
  All-hands production — AI drafts slides, speaker notes, and anticipated Q&A from the quarter's material · content is verifiable against real results; the exec rehearses and owns the delivery
• Automate
  Q&A and sentiment synthesis — AI clusters employee questions and reaction themes for leadership · summarization with citations, high volume, easily audited

metrics pull → business narrative → board deck → pre-read memo → Q&A prep → minutes → follow-ups

• Automate
  Metrics assembly — agent pulls the KPI package from source systems into the reporting template with variance flags · numbers reconcile against systems of record, same structure every quarter
• Draft
  Narrative drafting — AI drafts the "what happened and why" sections from metrics and leadership notes · claims are verifiable, but the framing of a miss is a judgment call with securities-law and trust consequences the CEO owns
• Automate
  Deck & pre-read production — AI turns the approved narrative into formatted deck and memo · formatting an already-approved source, fully verifiable
• Assist
  Q&A preparation — AI generates hostile-director questions from the deck and drafts response outlines · genuinely useful sparring, but knowing which question the board will actually ask is relationship knowledge
• Avoid
  The board meeting & disclosure decisions — what to say, what to concede, what to commit to · fiduciary accountability is personal and non-delegable; there is no rubric axis this doesn't max out
• Draft
  Minutes & follow-ups — AI drafts minutes from the recording for counsel's review and tracks commitments made · minutes are a legal record; counsel edits precisely because verifiable-against-transcript is not the same as legally well-drafted

market & internal research → situation assessment → strategy options memo → plan of record → OKRs → quarterly review docs

• Automate
  Research & synthesis — AI compiles competitor moves, market data, and internal performance into a cited situation assessment · every claim carries a source link, volume of reading is the bottleneck, errors surface on citation-check
• Assist
  Options development — AI generates and stress-tests strategic options, arguing each side and surfacing considerations the team missed · superb red-team, but its options regress to the plausible-generic; distinctive strategy comes from conviction the model doesn't have
• Avoid
  Strategy choice & plan of record — betting the company's resources on a direction · lowest verifiability of any artifact in this playbook, highest stakes, and accountability for the bet is the executive's entire job
• Draft
  OKR cascade — AI translates the plan into draft team-level objectives and flags cross-team conflicts and orphaned goals · consistency-checking is mechanical, but targets are commitments each leader must own
• Draft
  Quarterly review production — AI assembles progress-vs-plan docs with variance analysis and drafts the "what changed" discussion · verifiable against tracked metrics; deciding what the variance means is the review

vague executive ask → scoping memo → workplan → stakeholder input → findings deck → recommendation memo → handoff

• Draft
  Scoping — AI helps convert "figure out why churn is up" into a sharp scoping memo with hypotheses, needed data, and interview list · the memo is checkable against the ask, but reading what the exec actually wants — versus said — is the CoS's core skill
• Automate
  Data gathering & desk research — agent pulls internal reports, prior post-mortems, and external benchmarks into an organized evidence base · retrieval with citations, high volume, verifiable
• Assist
  Stakeholder interviews — the conversations where people say what they won't write down · AI drafts interview guides and synthesizes notes afterward, but candor requires a trusted human in the room
• Draft
  Analysis & synthesis — AI structures evidence into findings, tests hypotheses against the data, and surfaces contradictions · each analytical step is verifiable, but the CoS must be able to defend every finding to the exec without the model in the room
• Avoid
  Recommendation — what the company should do about it · the recommendation spends the exec's trust in the CoS; borrowed conviction is detectable and career-limiting
• Automate
  Handoff & follow-through — AI drafts the transition doc, owner checklists, and 30-day check-in schedule · mechanical packaging of decided content