The 90-day roadmap for editorial AI capability

A useful editorial AI roadmap does not need to start with a twelve-month transformation programme. Most newsrooms need a disciplined first 90 days: enough time to diagnose the operating model, test a few use cases, build capability, and give leadership a real decision path.

That 90-day structure sits behind my NRCS newsroom AI transformation positioning. It is deliberately practical. A broadcaster should know, after three months, where AI belongs, where it should be blocked, what the team can already use, and what structural work is required before scaling.

The exact outputs will always depend on the organisation. A national broadcaster, a digital-first newsroom, and a regional bureau will not need the same shape of engagement. The common baseline is simpler: after the work, the organisation should understand its own workflow better than it did before.

Here is the shape I would use.

Days 1 to 14: diagnose the workflow

The first two weeks should be observational and specific.

Map story intake, planning, rundown creation, script review, archive requests, production handoffs, digital adaptation, Arabic-English workflows, and shift handover. Identify where AI is already being used informally. Interview the people who carry context: producers, editors, planning leads, archive teams, technical operators, digital editors, and senior sponsors.

The output is a current-state map, not a generic AI strategy.

The map should show where context lives, where it disappears, which workflows have enough structure for AI support, and which ones are too risky or ambiguous. It should also identify the first use cases by operational fit rather than novelty.

This stage often exposes uncomfortable truths. The newsroom may discover that the biggest blocker is not model access. It may be unclear story state, inconsistent approvals, weak archive metadata, or shift handoffs that rely on memory.

Days 15 to 42: run narrow workflow pilots

The second phase should avoid giant pilots. Choose two or three narrow workflows with different risk profiles.

Good early candidates might include shift handoff summaries, archive research packets, planning note synthesis, first-pass script variants for non-sensitive segments, or bilingual comparison notes. Each pilot should have a named owner, review rule, success measure, and failure boundary.

The goal is not to prove AI is exciting. The goal is to learn how the newsroom responds when AI enters a real workflow.

Does it reduce rework? Does it make context clearer? Does it create review burden? Does it make editors nervous? Does it help producers move faster without lowering confidence? Does it expose integration gaps inside the NRCS, archive, CMS, or communication stack?

Those answers are more valuable than a polished demo.

Days 43 to 70: build the capability layer

By the middle of the roadmap, the team should have enough evidence to move from experimentation to capability building.

This means creating practical playbooks: when to use AI, how to review output, what to log, which prompts or workflows are approved, which use cases are blocked, and how to handle sensitive stories. It also means training people on the workflows that survived testing, not giving everyone the same generic AI session.

Different teams need different capability.

Producers need patterns for rundown-adjacent support and handoff clarity. Editors need review and escalation rules. Archive teams need retrieval boundaries. Technology teams need integration priorities. Strategy leaders need a portfolio view of what scales next.

This phase can connect naturally to an AI Integration Workshop when the workshop is used to transfer tested operating patterns rather than introduce AI from scratch.

Days 71 to 90: decide what scales

The final phase should produce an executive decision package and an operational handoff.

The package should include a workflow map, pilot results, governance recommendations, integration needs, training assets, risk register, and a sequenced roadmap for the next quarter. It should separate quick wins from structural requirements. It should also name the workflows that should not scale yet.

That last point matters. A credible roadmap includes restraint.

If a newsroom discovers that archive metadata is too weak, it should not pretend AI archive retrieval is ready. If bilingual review is inconsistent, it should not push AI-written copy into public channels without stronger control. If live production states are unclear, agentic workflows should wait.

What success looks like after 90 days

At the end of 90 days, a broadcaster should have fewer slogans and better answers.

Which newsroom workflows are AI-ready? Which need more structure? What did the team learn from pilots? Who owns governance? Which systems need integration? What capabilities now exist inside the newsroom? What should be funded next?

That is enough to move from interest to operating confidence.

Editorial AI capability should grow through evidence. Ninety days is long enough to learn honestly and short enough to preserve momentum. The trick is to spend the time on the workflow, not on the theatre around the workflow.