Strategy

Environmental scan → strategy hypotheses memo → BU strategy decks → financial plan/targets → board strategy deck → OKR/initiative portfolio

• Automate
  Environmental & internal scan — deep-research agent compiles market shifts, competitor moves, and internal performance trends into a briefing pack · high-volume synthesis of verifiable public and internal data, errors caught downstream
• Assist
  Framing the strategic questions — AI stress-tests the question list ("what would make this the wrong question?"), but humans set the agenda · low volume, judgment-heavy, and a badly framed question poisons everything downstream
• Draft
  BU strategy deck drafting — AI drafts situation-complication-resolution narratives and slide skeletons from the scan plus leadership interviews · verifiable against source material and heavily edited anyway, but the BU head owns the argument
• Assist
  Financial plan & target-setting — AI builds scenario sensitivities on the model; humans set the actual targets · targets are commitments with career stakes, not calculations
• Draft
  Cross-BU challenge & synthesis — AI red-teams each BU deck for contradictions, hockey-stick assumptions, and resource conflicts across the portfolio · cheap systematic critique at volume, but the CSO decides which challenges matter
• Avoid
  The strategic choices & capital allocation — where to play, what to fund, what to kill · irreversible, accountability cannot be delegated, and conviction is the product
• Automate
  Cascade & communication — AI adapts the approved strategy into town-hall scripts, BU one-pagers, and FAQ docs · pure reformatting of a settled source of truth, easily spot-checked

Research question brief → source map → analysis workbook (sizing, structure, trends) → synthesis memo → readout deck

• Assist
  Question framing & scoping — AI proposes decompositions and MECE issue trees; the strategist picks the cut that matches the decision at hand · the frame determines the answer, and the decision context lives in executives' heads, not documents
• Automate
  Source gathering & landscape scan — deep-research agents pull filings, transcripts, trade press, and expert-network summaries into a cited source map · exactly what these agents are built for — high volume, verifiable citations, low stakes per item
• Draft
  Market sizing & structure analysis — AI drafts top-down/bottom-up sizing with explicit assumptions and Porter-style structure notes · arithmetic is verifiable but assumption choices swing the answer 5x, so a human must own each one
• Draft
  Synthesis into a point of view — AI produces a first-pass "so what" memo; the strategist rewrites it into an actual position · AI synthesis regresses to consensus; the value of the memo is precisely where it departs from consensus
• Automate
  Challenging the synthesis — a separate AI session argues the opposite conclusion from the same source map · adversarial critique is cheap, repeatable, and any bad objection is harmlessly discarded
• Draft
  Readout & Q&A prep — AI generates the likely-hardest executive questions and drafts backup slides · question prediction is pattern-matching on a known corpus, but answers carry the presenter's credibility

Competitor profiles → threat hypotheses → war-game scenario book → move/countermove log → response playbook

• Automate
  Competitor intelligence assembly — agents maintain living profiles from filings, earnings calls, job postings, patents, and pricing pages · high-volume monitoring of public, verifiable signals with low per-item stakes
• Draft
  Threat hypothesis generation — AI proposes attack vectors a competitor could plausibly take, including uncomfortable ones · breadth of ideas is the point and bad ideas cost nothing, but humans rank plausibility
• Draft
  Scenario design — AI drafts the war-game scenario book with injects and role briefs; facilitators tune for the room · verifiable against the intel base, but scenario realism depends on organizational context AI lacks
• Automate
  Playing the red team — AI plays the competitor live during the game, arguing its moves in character with the competitor's actual constraints and incentives · this is AI's best role in the room — tireless, unembarrassed devil's advocacy where wrong moves are the exercise, not a cost
• Assist
  Countermove evaluation — participants debate responses; AI pressure-tests each ("here's how the red team punishes that") · the debate builds the shared conviction the exercise exists to produce
• Avoid
  Response playbook commitment — which triggers commit the firm to which responses, with pre-authorized resources · pre-commitments to competitive action are capital and reputation bets executives must own

Thesis & screening criteria → target long-list → target one-pagers → indicative valuation model → diligence request list → diligence findings report → deal memo/IC paper → integration plan

• Automate
  Target scanning & screening — agents sweep company databases and filings against the acquisition thesis and refresh the long-list weekly · high-volume pattern-matching on verifiable criteria; a missed target costs little and a bad one gets filtered next stage
• Draft
  Target profiling & indicative valuation — AI drafts one-pagers and comps-based valuation ranges · comps are checkable, but multiple selection and synergy assumptions are judgment the deal lead signs
• Draft
  Diligence document review — AI extracts change-of-control clauses, customer concentration, liabilities, and anomalies from thousands of VDR documents, every finding cited to its source page · the definitive high-volume extraction task, but a missed material clause is a nondisclosure-grade miss — lawyers verify every flagged item and sample the unflagged
• Draft
  Diligence synthesis & red flags report — AI aggregates workstream findings into a draft findings report with open-questions log · verifiable against the diligence record; the deal team owns what counts as deal-breaking
• Assist
  Deal memo & IC recommendation — AI assembles the fact base and drafts the paper; the sponsor writes the recommendation and the risks section in their own voice · the IC paper is a personal accountability document — the sponsor's judgment is what the committee is buying
• Avoid
  Go/no-go and price — the bid, the walk-away, the board recommendation · irreversible capital allocation under uncertainty; this accountability is the corp-dev job
• Draft
  Integration planning — AI drafts Day-1 readiness checklists, synergy-tracking workbooks, and workstream plans from the diligence record · templated, verifiable against diligence findings, and revised heavily in contact with reality

Opportunity brief → business model canvas → assumption map → validation evidence log → unit-economics model → investment memo

• Automate
  Business model option generation — AI generates and stress-tests alternative models (pricing structures, channel plays, analogues from other industries) · divergent ideation at volume where every option is reversible and cheap to discard
• Draft
  Assumption mapping — AI decomposes each model into testable assumptions ranked by kill-power; humans re-rank against what they know the market rewards · systematic decomposition is mechanical, but which assumption is truly load-bearing is a market-feel call
• Automate
  Evidence gathering & desk validation — deep-research agents hunt for analogue outcomes, pricing benchmarks, and disconfirming evidence per assumption · high-volume, citable retrieval where disconfirming evidence is the deliverable, not a risk
• Assist
  Customer & expert conversations — humans run them; AI drafts discussion guides and structures the notes afterward · the signal is in tone, hesitation, and relationships — rules-vs-relationships lands squarely on relationships
• Draft
  Unit-economics modeling — AI builds the driver-based model with scenario toggles; humans own the three numbers that matter · model mechanics are verifiable; CAC, conversion, and churn assumptions are the actual bet in disguise
• Avoid
  Fund / kill / pivot recommendation — the venture bet and its resourcing · a capital-allocation choice under deep uncertainty whose author must be accountable when it's wrong

Initiative charters → milestone/KPI tracker → monthly status packs → risk & decision log → QBR portfolio review deck

• Automate
  Status collection & normalization — agents pull updates from project tools, finance actuals, and owner check-ins into a standard tracker · repetitive, rules-based aggregation of verifiable internal data, wrong entries surface immediately
• Automate
  Variance & risk detection — AI flags slipping milestones, benefit shortfalls versus charter, and stale risks across the portfolio · threshold logic on structured data at volume; a false flag costs one glance
• Draft
  Status pack drafting — AI writes the monthly narrative per initiative from tracker data and owner notes · verifiable against the tracker, but owners must edit because status language is political and AI can't read the room
• Assist
  Root-cause conversations on red initiatives — sponsors dig into why; AI supplies the timeline and prior-commitment receipts · these are accountability conversations between people — relationship work with career stakes
• Avoid
  Kill / recommit / rescope decisions — the quarterly reckoning on which initiatives live · reallocating people and capital, and un-killing an initiative is organizationally near-irreversible
• Automate
  QBR deck assembly — AI compiles the portfolio view, trend charts, and decision log into the review deck · mechanical assembly from a settled source of truth, spot-checked by the PMO