Layoff Comms: What AI Drafts, What Stays Human

The single biggest mistake I see SMB founders make running a layoff is letting AI write the parts that the company needs to see them write — and writing themselves the parts AI would do better. Get the split wrong and you do permanent damage to the trust of the people you're keeping.

Why does the AI-human split matter so much in a layoff?

Because a layoff is a stress test of the founder's voice. The team can tell the difference between a sentence the founder wrote and a sentence the model produced. They will not say it out loud. They will silently update their priors about how much you actually care, and three months later you'll feel it in retention.

Definition: AI-human split — the explicit allocation of which layoff communication artifacts AI helps draft (and which a human edits) vs. which a human writes from scratch with no model assist.

The operational pieces — severance math, COBRA explainer, equity acceleration calculator, FAQ — benefit from AI because accuracy and consistency matter more than voice. The voice pieces — what you said at the all-hands, what you said in the 1-on-1 — are identity-defining moments where any whiff of synthetic language compounds the damage.

What does the 5-day cadence actually look like?

Day -7 to Day 0 — Preparation (private)

The decision is made. Numbers are locked. List of affected roles is finalized with HR and legal. The voice work begins.

AI helps: severance calculator, COBRA/benefits transition guide, equity acceleration math, role-by-role severance package generator, draft FAQ covering the 30 most-likely questions.

Founder writes (no AI): the all-hands script, the personal note that will go to each affected person, the email to the surviving team, the public statement if one is needed.

The founder should physically write these on paper or in a clean document with no model assist. The reason is not productivity — it's that the act of writing them by hand surfaces things the founder needs to think about before saying them out loud.

Day 0 morning — All-hands

The founder speaks. No slides, no AI-polished script reading. Honest reason, what's happening today, what affected people will hear in the next hour, what the surviving team should expect this week. 8-15 minutes maximum.

Day 0 morning — Affected 1-on-1s

The direct manager and an HR partner. Pre-written packet from the AI workflow on the table. The 1-on-1 itself is human; the documents are operational.

Day 0 afternoon — Surviving team check-in

Small-group sessions led by managers. Founder is available. No AI-generated talking points distributed; managers speak in their own voices, prepared by reading the founder's all-hands transcript and a short coaching guide.

Day +1 to +3 — Operational follow-through

AI-drafted FAQ goes live internally. Severance packets are signed. Benefits transition support runs. Manager office hours run twice a day.

Day +4 to +5 — Forward-looking message

The founder's second all-hands. What we're focusing on now. What the smaller team is going to ship. AI helps with structure; the voice is fully human.

Copy/paste template — FAQ generation (AI does this)

You are drafting an internal FAQ for a layoff at a 30-500-employee company.

Inputs:
- Company size before and after: [N → N-K]
- Functions affected: [LIST]
- Geography: [LIST]
- Severance policy: [WEEKS BASE + WEEKS PER YEAR OF SERVICE, with cap]
- Equity treatment: [STANDARD VESTING / ACCELERATION RULE]
- Benefits transition: [HEALTH/COBRA, EAP, OUTPLACEMENT DETAILS]
- Public-statement constraint: [YES/NO, if YES paste statement]
- Tone constraint: factual, calm, no euphemism, no "this difficult decision" filler.

Output: 30 FAQ entries grouped by:
1. Why this is happening (3-5)
2. What it means for affected employees (8-10)
3. What it means for the surviving team (6-8)
4. Operational details — pay, benefits, equipment (5-7)
5. External — clients, press, social (3-5)

Hard rules:
- No euphemisms. "Layoff" not "right-sizing." "Affected" not "impacted" (overused).
- No corporate filler. No "this difficult but necessary decision."
- Every answer 2-4 sentences. No essay-length entries.
- If an answer would require legal review, mark [LEGAL REVIEW] and write a placeholder.
- Numbers must come from inputs; do NOT invent.

This is the prompt that does the operational heavy-lifting. The FAQ goes through legal review, gets distributed internally on Day 0 afternoon, and answers 80% of the questions managers will face in office hours.

Tool tip (AIAdvisoryBoard.me): The hardest part of a layoff comms plan isn't the writing — it's seeing whether the comms are actually landing across the surviving team. A daily Plan → Fact → Gap view of the management layer during the week after the announcement shows which managers held the line, which teams went quiet, and where rumors filled the gap the all-hands didn't close. SMBs that run this measurement in the first 5 days recover trust 2-3x faster than those that wait for an engagement survey six weeks later. See how the 7-day diagnostic works at https://aiadvisoryboard.me/?lang=en.

Manager scan (2-minute digest example)

• Plan: every affected 1-on-1 happens by 11 AM Day 0; surviving check-ins by 3 PM
• Fact: 18 of 20 1-on-1s on time; 2 slipped due to PTO conflict; surviving check-ins all done
• Gap: 2 affected employees were told by their manager before the 1-on-1 slot — process violation, addressed same day
• All-hands attendance: 94% live, recording sent to 6%; no opt-outs flagged
• FAQ readership: 78% of surviving team read by EOD Day 0
• Manager office hours: 11 of 12 managers held hours both days; 1 absence covered by HR
• Rumor signals: 2 Slack channels showed speculation patterns; founder addressed in Day +2 message
• 1-on-1 quality: HR partner debrief flagged 1 manager whose talking points sounded AI-generated; coaching scheduled
• Surviving team retention signal: 0 voluntary resignations in first 14 days (baseline: 1-2 expected)
• Forward-looking message: Day +4 all-hands attendance 91% live, sentiment net-positive in feedback channel

Micro-case (what changes after 7-14 days)

A 180-person SaaS company ran a 12% layoff in March. First instinct was to let AI draft everything for "consistency." Founder caught it in time — pulled back the all-hands script and the affected 1-on-1 notes, rewrote both by hand the night before. Used AI for the FAQ, severance packets, and benefits transition guide. Day 0 all-hands was 11 minutes, no slides, founder voice clearly recognizable. By Day 5 the FAQ had answered 80% of inbound questions; manager office hours covered the rest. Two months later, voluntary resignations among the surviving team were below baseline. The board asked how. The founder's answer: "We wrote the parts that were about us. AI wrote the parts that were about paperwork."

Note on this case: This example is illustrative — based on typical patterns we observe with companies of 30-500 employees, not a single named client. Specific numbers are rounded approximations of common ranges, not guarantees.

Tool tip (AIAdvisoryBoard.me): Daily management visibility matters most in the two weeks after a layoff, when the founder is busy being the voice and managers are quietly figuring out who's at risk. The Plan → Fact → Gap pattern surfaces which teams the founder needs to walk through this week — not in a survey six weeks late. The 7-day diagnostic gives you that visibility before the recovery window closes. See it work at https://aiadvisoryboard.me/?lang=en.

FAQ

Should AI write the all-hands script? No. Use AI as a structural prompt ("am I covering the five things I should cover?") but write the words yourself. The team can hear the difference, and on this one occasion the difference compounds.

Can AI write the personal note to each affected person? No. Hand-write or type each one. Reference something specific. This is the single highest-stakes voice artifact of the entire week. If it sounds templated, the layoff becomes the founding story of how this company actually treats people.

What about the public statement? Founder voice with AI as a stress-test. Draft yourself, run it through AI with the prompt "what's wrong with this," edit, and ship. Do not let AI write the first draft.

What if I can't write the all-hands by myself? Then you're not ready to give it yet. Take an extra day, sit with what you actually want to say, and write. If you genuinely cannot write it, that's a signal worth understanding before standing in front of the team.

Should we mention the AI assist in the FAQ? Disclosure on the FAQ page header is good practice — "this FAQ was drafted with AI assistance and reviewed by our HR and legal teams." Transparency on operational documents costs nothing and earns trust.

Conclusion

A layoff is a voice test the founder cannot outsource. AI does the paperwork — severance, FAQ, benefits, manager packets. The founder does the voice — all-hands, 1-on-1s, surviving team message, public statement. Get the split wrong and the company's identity changes. Get it right and the recovery starts on Day 1.

Pick the next high-stakes comms cycle. Map what AI drafts and what you write from scratch. Hold that line.

If you want a system that surfaces the Plan → Fact → Gap across managers in the days after — automatically, every day — see how the 7-day diagnostic works at https://aiadvisoryboard.me/?lang=en.