Comms: crisis-comms first-draft drafter

A communications director for a SaaS company described her worst moment as "two hours into a security incident, staring at a blank doc, trying to remember the last time we wrote a customer-facing breach notification." The team had a template, somewhere. Nobody could find it under pressure.

The crisis-comms AI employee drafts the first version. The team's first 30 minutes after a crisis go to thinking — what do we know, who needs to be told, when — instead of typing.

The shape of the role

Title. Comms AI — Crisis Communications Specialist.

Mission. When a crisis is declared, produce first drafts of stakeholder communications within 15 minutes. Manage the comms cadence as the situation evolves.

Outcomes. Time-to-first-stakeholder-comm, comm-quality (post-hoc assessed), template availability.

Reports to. Head of Comms or General Counsel.

Tools. Crisis-comms template library, stakeholder-map, severity-tier definitions, legal-review integration, voice eval.

Boundaries. Drafts. The director, GC, and CEO approve. Doesn't send.

Severity tiers

Different crises need different responses. The agent's drafts adapt by severity:

Tier 1 — Major incident. Customer data exposure, multi-day outage, executive misconduct, regulatory action. Drafts go to the CEO and GC immediately. Customer comms within 24 hours.

Tier 2 — Significant incident. Limited-scope outage, partial service degradation, employee issue with public visibility. Director-led comms; customer-facing within 48 hours if relevant.

Tier 3 — Operational issue. Brief outage, fixable bug at scale. Standard incident communications via status page; no executive escalation needed.

The agent classifies based on input from the incident commander. The classification is reviewable; the director can escalate or de-escalate.

Stakeholder map

For each crisis, the agent identifies who needs comms:

Customers (segmented if needed: enterprise vs. SMB; affected vs. not).
Employees.
Investors / board.
Regulators (where applicable).
Press (if/when needed).
Partners.

For each stakeholder group, a separate first draft. Each draft considers:

What this group needs to know.
What they don't need to know yet.
What action they can/should take.
When they'll get the next update.

Draft templates

The library includes templates for common crisis types:

Security incident (sub-categorised: data breach, suspected breach, vulnerability disclosure).
Service outage (sub-categorised: partial, full, regional).
Personnel issue (executive transition, public misconduct).
Regulatory (subpoena, investigation).
Third-party breach (vendor compromise affecting our data).
Operational (data loss, billing error at scale).

Each template has been reviewed by GC and director once, in calm conditions. The agent customises with the specific situation's details. The director and GC review the customisation.

Legal-review hook

For Tier 1 and Tier 2 crises, the agent's drafts route through GC review before any approval is requested. The GC's edits become part of the template library for future incidents.

This is the discipline that keeps companies out of trouble in their crisis comms. Most public crisis-comms failures are not first-draft failures; they're the absence of legal review.

The cadence

A crisis comms strategy isn't one message. It's a sequence:

Initial acknowledgement (we're aware, we're investigating).
First substantive update (here's what we know, here's what we're doing).
Resolution communication (the issue is resolved, here's what happened).
Post-incident report (root cause, prevention).

The agent drafts each in sequence. The director and GC review and adjust. The CEO signs the public versions.

A tabletop exercise

A team using this AI employee should run tabletop exercises:

Hypothetical crisis declared.
Agent drafts the comms.
Team reviews against established standards.
Edits feed the template library.

This is similar to fire drills. The first time it's clumsy. By the third, the team trusts the system.

What we won't ship

Auto-sending crisis comms. Always human approval.

Drafting for crises the team hasn't planned for without immediate director and GC review.

Speculation. Crisis comms states what's known. Speculation gets lawyers involved.

Personality content. A crisis isn't the place for brand voice. The voice is sober.

The KPIs the director watches

Time-to-first-comm for each tier.
Edit volume on agent's drafts.
Post-hoc review rating of comms quality.
Stakeholder-recovery sentiment measured at 30 days post-incident.

If post-hoc rating drops, the templates need GC re-review.

How to start

Build the template library with the GC's involvement. Run tabletop exercises monthly. After three exercises, the team trusts the system. Real crises (which will come) are handled with discipline rather than panic.

Close

The crisis-comms AI employee is a teammate whose job is the first-draft work that lets the human leaders focus on the decisions. The templates exist. The drafts come fast. The legal review is built in. The cadence runs. The director, GC, and CEO sign the comms that go out.