Sales

ICP definition → account/contact list → research dossiers → personalized sequences → replies → booked meetings

• Assist
  ICP & signal definition — humans define who to target; AI mines closed-won data for the patterns · a wrong ICP silently poisons everything downstream, and closed-won patterns need human interpretation
• Automate
  List building & enrichment — AI agents build and enrich account lists from intent signals, job posts, and tech-stack data · high volume, spot-checkable, and a bad row costs one wasted email
• Automate
  Account research dossiers — AI compiles a per-account brief (priorities, news, likely pain, entry points) · verifiable against sources and exactly the work reps skipped anyway
• Draft
  Personalized sequence drafting — AI writes first-touch and follow-up emails grounded in the dossier · volume play, but "AI-personalized" spam is now pattern-matched by buyers, so humans edit the first touch
• Automate
  Send orchestration & follow-ups — AI runs the cadence, handles no-replies, and books meetings from positive replies · rule-based sequencing with calendar-level verifiability
• Assist
  Reply handling with objections — AI drafts responses to substantive replies for rep approval · the first real conversation is where the relationship starts; a canned reply ends it

form/chat/PLG signal → enriched lead record → qualification score → routing decision → first meeting → CRM record

• Automate
  Instant enrichment — AI fills company size, industry, funding, and tech stack the moment a lead arrives · high volume, verifiable fields, zero relationship content
• Draft*
  Conversational qualification — an AI agent chats with inbound visitors, answers product questions, and captures need/timeline (Draft -grade autonomy) · reversible and monitored, but it is the brand's first conversation, so transcripts get human review
• Automate
  Scoring & routing — AI scores fit and routes to the right rep or self-serve motion by written rules · rules-based, measurable against conversion, and a misroute is recoverable
• Automate
  Meeting scheduling — AI books the meeting directly in the qualification conversation · mechanical, calendar-verifiable, speed-to-lead is the whole value
• Automate
  Pre-meeting brief — AI hands the rep a one-pager on the lead before the first call · compiled from sources the rep can check in seconds
• Assist
  Disqualification decisions — AI recommends, human confirms on borderline accounts · a false disqualify is invisible and unrecoverable — the lead just never hears back

discovery call → call notes/summary → mutual action plan → tailored demo → proposal/quote → redlines → signature

• Automate
  Pre-call research & question prep — AI builds the account brief and suggests discovery questions from similar won deals · verifiable compilation work that directly raises call quality
• Assist
  Discovery calls — the human sells; AI transcribes and captures pain, metrics, and stakeholders in real time · the call is pure relationships-and-trust territory; AI stays in the note-taking seat
• Automate
  Call summary & CRM hygiene — AI writes the summary, updates fields, and drafts the follow-up email · checkable against the transcript, and the single most hated manual task in sales
• Draft
  Mutual action plan & demo tailoring — AI drafts the plan and a demo narrative mapped to stated pains · templated from the call record, but the rep owns what gets promised
• Draft
  Proposal & pricing — AI assembles the proposal from approved blocks and the deal record; pricing beyond guardrails needs human sign-off · a proposal is a commercial commitment with real reversibility cost; discounting stays gated
• Assist
  Legal & redline review — AI flags non-standard terms against the playbook for counsel · contract language carries regulatory and litigation exposure that outranks speed

usage/support/CRM data → account health score → risk & expansion flags → QBR deck → renewal/expansion plan

• Automate
  Health monitoring — AI watches usage, support tickets, champion changes, and sentiment for every account · continuous pattern-watching across more accounts than any human can hold
• Draft
  Risk & expansion flagging — AI raises churn-risk and upsell signals with the evidence attached · signals are checkable, but acting on a false churn flag can itself damage the relationship
• Automate
  QBR deck assembly — AI builds the value-delivered deck (usage, outcomes vs. goals, benchmarks) per account · templated, data-verifiable, and the reason QBRs actually happen for mid-tier accounts
• Draft
  QBR narrative & recommendations — AI drafts the "what's next" story; the AM rewrites it with relationship context no system holds · context availability is the limit — the system doesn't know what was said at dinner
• Assist
  Renewal negotiation — human-led; AI preps comparable-deal terms and concession scenarios · high stakes, low reversibility, and pure relationship judgment
• Draft
  Expansion outreach — AI drafts the expansion pitch tied to observed usage patterns · grounded in verifiable product data, sent under the AM's name after edit

CRM pipeline data → deal inspection notes → stage/commit adjustments → forecast rollup → board number → territory/comp plans

• Automate
  CRM data hygiene — AI fills missing fields from calls and emails and flags stale or contradictory deal records · verifiable against source communications, endless volume, no judgment
• Draft
  Deal inspection — AI scores every open deal on engagement signals and flags happy-ears deals where rep optimism contradicts the activity record · evidence is checkable, but the flag is an accusation that needs human handling
• Draft
  Forecast modeling — AI produces a signal-based forecast alongside rep commits and explains the gap · the model is backtestable, yet the committed number is an accountability artifact a leader must own
• Assist
  Pipeline coverage & scenario planning — AI answers "what happens to Q4 if these three deals slip" on demand · useful arithmetic on assumptions the human must supply and sanity-check
• Draft
  Territory & quota planning — AI drafts balanced territory models from account data · optimization is verifiable, but quota changes hit paychecks — high stakes, low reversibility, human sign-off
• Automate*
  Comp calculation & disputes — rules-based comp runs automated; disputes and exceptions stay human (Automate the math, Avoid autonomous exception rulings) · payroll accuracy is verifiable, fairness rulings are not

call recordings → talk-pattern analysis → coaching scorecards → playbooks/battlecards → practice reps → certification

• Automate
  Call library analysis — AI analyzes every recorded call for talk ratio, question quality, objection handling, and competitor mentions · humans can review 2% of calls; AI reviews 100% with citable clips
• Draft
  Coaching scorecards — AI scores calls against the team's methodology and drafts per-rep coaching notes · scores are clip-verifiable, but delivered coaching is a manager's relationship job
• Automate
  Battlecard & playbook maintenance — AI updates competitor battlecards from won/lost call evidence and competitor releases · sourced, checkable, and perpetually stale when done manually
• Automate
  Objection-handling practice — reps rehearse against AI role-play buyers that mimic real personas and objections · unlimited reps, zero stakes, instant feedback — the ideal AI use case
• Draft
  New-hire onboarding paths — AI assembles personalized ramp plans from the rep's gaps in practice sessions · adaptive sequencing helps, but managers own ramp expectations
• Draft
  Win/loss analysis — AI drafts the quarterly win/loss report from call and CRM evidence · evidence-grounded, but "why we really lose" conclusions steer strategy and need human interrogation