Marketing & Content

market/audience research → campaign brief → messaging matrix → channel plan → asset set → launch calendar

• Automate
  Audience & competitive research — AI agents pull competitor positioning, review-site language, and social sentiment into a research digest · high volume, low stakes, and easy to spot-check against sources
• Draft
  Campaign brief drafting — AI drafts the brief (objective, audience, message, offer, channels) from the research digest and past briefs · well-templated artifact with rich context, but positioning choices carry real strategic stakes
• Draft
  Messaging matrix — AI generates persona-by-pain-point message variants for the team to prune · pure recombination at volume; humans own which claims the brand will actually stand behind
• Assist
  Channel plan & budget split — AI proposes allocation from historical channel performance · budget calls are hard to verify in advance and expensive to reverse mid-flight
• Draft
  Asset production — AI produces the ad copy, landing page copy, and image variants from the approved matrix · high volume and cheap to regenerate, but each shipped asset is a public brand act a human must own
• Automate
  Launch QA & compliance check — AI checks assets against brand guidelines and claim rules before ship · rule-based, verifiable checks against a written standard

content pillars → editorial calendar → post drafts → visuals → scheduled queue → engagement replies → performance recap

• Automate
  Topic mining — AI monitors industry news, community threads, and top-performing competitor posts to propose the week's angles · continuous high-volume scanning where a bad suggestion costs nothing
• Draft
  Calendar planning — AI drafts the weekly calendar mapped to pillars and formats · templated and reversible, but a human should own the mix and timing
• Draft
  Post drafting & repurposing — AI turns one anchor asset (webinar, blog, podcast) into platform-native posts, clips, and threads · this is the canonical volume play; voice drift is the risk, so a human edits every post
• Draft
  Visual creation — AI generates or adapts images and short video from brand templates · fast to verify by eye; brand-inconsistent output is caught before publish
• Automate
  Publishing & scheduling — AI queues approved posts at optimal times · mechanical, rule-based, fully reversible before send
• Assist
  Engagement & replies — AI drafts replies to comments and DMs for one-click approval · public, relationship-driven interactions where a tone miss is screenshot-able; keep a human on send

keyword/topic map → content briefs → drafts → optimized pages → internal links → rank/traffic report

• Automate
  Keyword & topic clustering — AI clusters search and AI-answer-engine queries into a topic map with intent labels · data-heavy pattern work that is fully checkable against the tool's own data
• Automate
  Content brief creation — AI builds per-page briefs (intent, entities to cover, competing pages, structure) · templated, verifiable against SERP evidence, produced in bulk
• Draft
  Article drafting — AI writes the draft from the brief plus the company's own expertise docs and SME interview notes · without proprietary context this produces the same article every competitor's AI writes; the human adds the un-Googleable parts
• Assist
  E-E-A-T & fact review — human expert verifies claims, adds first-hand experience; AI flags unsupported claims for them · accuracy stakes are the whole game here, and hallucinated stats in indexed content are slow to detect and reverse
• Automate
  Technical & on-page optimization — AI fixes metadata, schema, internal links, and cannibalization at scale · rule-based, machine-verifiable, high repetition
• Automate
  Performance reporting — AI compiles rank, traffic, and AI-citation share into the monthly report with anomaly flags · pure data assembly with human review of the narrative

segment definitions → flow map → email copy/design variants → A/B tests → deliverability & revenue report

• Draft
  Segmentation & audience building — AI proposes segments from behavioral and purchase data · verifiable against the data, but segment logic errors silently mis-target thousands of people
• Draft
  Flow architecture — AI drafts the lifecycle flow map (welcome, cart, winback, post-purchase) with trigger logic · well-known patterns with rich in-platform context; human owns the logic before it runs unattended
• Automate
  Copy & subject line variants — AI generates variant sets per segment for testing · high volume, instantly measurable, and a bad variant loses one test rather than the brand
• Automate
  Send-time & content personalization — the platform's AI picks per-recipient timing and content blocks · exactly the per-user optimization loop humans cannot do at all
• Automate*
  Compliance & list-hygiene checks — AI enforces consent status, suppression lists, and CAN-SPAM/GDPR rules (Automate for checks, never bypassable) · rule-based but regulator-facing, so the rules themselves stay human-owned
• Assist
  Test analysis & iteration — AI reads test results and proposes next variants; human approves changes to always-on flows · flows compound over months, so a wrong "learning" propagates at scale

brand strategy → visual identity system → creative concepts → produced assets → brand guidelines → usage QA

• Draft
  Territory exploration — AI generates dozens of visual and verbal directions as moodboards and sample lines · cheap divergence humans could never afford before; choosing a territory is pure judgment
• Assist
  Brand strategy & positioning — humans decide what the brand means; AI stress-tests it against competitor positioning · low verifiability, high stakes, and fundamentally a relationships-and-meaning problem, not a rules problem
• Automate
  Concept-to-asset production — AI produces resized, reformatted, and localized versions of approved hero creative · mechanical derivation from an approved source, checkable at a glance
• Draft
  Original hero creative — AI generates candidate imagery and film; creative director selects and directs revisions · output is instantly inspectable, but IP provenance and taste keep a human firmly in the chair
• Automate
  Brand guideline enforcement — AI reviews outbound assets from every team against the written guidelines · rules codified in a document, applied at high volume, with flagging not blocking
• Avoid*
  Legal/IP clearance — trademark and likeness review of AI-generated creative (Avoid for AI-only) · regulatory exposure and unsettled AI-IP law make this a human counsel call

tracking plan → clean event data → dashboards → attribution model → insight memo → budget reallocation

• Automate
  Tracking plan & data QA — AI audits event taxonomies and flags broken or missing tracking · rule-based checks against a spec, tedious at human speed
• Automate
  Dashboard assembly — AI builds and maintains standard funnel and channel dashboards from prompts · templated, verifiable against the underlying queries, high repetition
• Assist
  Attribution modeling — AI runs MMM and multi-touch models and explains their disagreements · models are unverifiable against ground truth, and the answer moves real budget
• Automate*
  Anomaly detection & alerting — AI watches metrics and explains spikes and drops with likely causes (Automate for detection, Assist for causes) · detection is checkable; causal stories are not
• Draft
  Insight memo writing — AI drafts the monthly "what happened and why" narrative from the data · fast to verify against the dashboards, but the recommendation section is a stakes-bearing human call
• Avoid*
  Budget reallocation decisions — humans decide; AI provides scenario projections (Avoid for autonomous action) · low verifiability plus six-figure irreversibility is the rubric's clearest red zone