Playbook

Firing an AI Employee

Sometimes the role is wrong. Sometimes the model is wrong. Sometimes you just need to reset.

Know when an AI role on your team needs to be swapped, retired, or rebuilt — and have a clean protocol for doing it without losing institutional context or disrupting the rest of your team.

Free to startNo credit card requiredUpdated Apr 2026
For
Any Tycoon operator whose AI team includes a role that's underperforming, duplicating another role, or no longer needed as the business evolves. Also useful as a preventive read before you hire any new role — the criteria clarify what success looks like.
Time to results
A complete role swap takes 30 minutes to execute and 1-2 weeks for the new role to reach prior output quality. Retiring an unneeded role takes under 10 minutes.

The playbook

  1. 1
    Step 1 — Diagnose why the role is failing

    Three failure modes. (A) Context starvation — the role doesn't have the inputs it needs; output is generic. Fix the context, don't fire the role. (B) Autonomy mismatch — you're either micromanaging (role can't contribute) or over-delegating (role makes bad calls). Adjust autonomy. (C) Role-reality mismatch — the shape of the role doesn't match your actual work. This is the only case that requires firing; see steps 2-5. Most 'bad AI hires' are actually (A) or (B) and don't need a firing.

    A blank page and 15 minutesThe role's last 30 days of output for review
  2. 2
    Step 2 — Run the 30-day scorecard honestly

    For genuine role failures, grade against the 30-day scorecard you wrote at hire time (see playbook/first-ai-hire-blueprint). If the scorecard wasn't written at hire time, write it retrospectively now: what were the three outcomes you expected, and did they happen? Be honest. 'Vaguely useful' is not a passing grade. If the role genuinely missed on the scorecard for reasons that are structural (role shape wrong), move to step 3. If the role missed for reasons that are fixable, go back to step 1.

    The original scorecard if writtenA honest 30-minute retrospective
  3. 3
    Step 3 — Capture institutional knowledge before firing

    Before retiring the role, export what it learned: the decision logs, the voice guides, the rules it built, the customer patterns it noticed. This goes into the company brain (see hire/ai-knowledge-manager). Otherwise you'll restart from zero with the new role. This is the single most commonly skipped step — firing without knowledge transfer is how founders end up with new AI hires that repeat old mistakes.

    Tycoon memory exporthire/ai-knowledge-manager if you have oneNotion for the transferred context
  4. 4
    Step 4 — Execute the swap or retire

    Three swap patterns. (A) Role replacement — the AI CMO isn't shipping; hire a different AI CMO configuration. Same title, fresh start with inherited context. (B) Role substitution — the AI CMO is actually not what you needed; you needed an AI Head of Growth. Different role shape. (C) Role retirement — the role isn't needed anymore (you hired an AI AE and realized you don't actually have a sales motion). Delete the role, redistribute its work. In all three cases the swap takes minutes in Tycoon.

    Tycoon role managementhire/ai-hr-manager for the formal off-ramp if you have one
  5. 5
    Step 5 — Onboard the replacement with inherited context

    The new role starts with the captured knowledge from step 3, not a cold context. Run an abbreviated version of the onboarding playbook (see playbook/onboarding-ai-cmo-week1 for the reference pattern) — you can usually compress it to 3 days because the context already exists. First production output in 2-3 days; full cadence within a week. Track the replacement against a fresh 30-day scorecard.

    playbook/onboarding-ai-cmo-week1 as the referenceThe transferred context from step 3
  6. 6
    Step 6 — Retrospective on what you learned

    Why did the first hire fail? Was it hiring order (hired too early), hiring wrong role shape, hiring without clear outcomes, poor onboarding? Write the retrospective — one page. This goes into the company brain so the next hire benefits. Firing an AI hire is expensive in time; firing two for the same reason is self-inflicted. Most founders go through one failed hire in their first year and learn permanently from it.

    A one-page retrospectivehire/ai-knowledge-manager for the filed lesson

Pitfalls to avoid

  • !Firing the role when the real problem is your context or autonomy settings — you'll just repeat the failure.
  • !Skipping the knowledge transfer step — new hire starts cold, you rediscover every lesson.
  • !Emotional firing at week 2 before giving the role a real 30-day runway — you haven't gathered enough data.
  • !Hiring a human replacement out of frustration — you're swapping a fixable AI role for a $100K+ human problem.
  • !Not writing the retrospective — you'll hire the same role wrong again in three months.

Frequently asked questions

How do I know if my AI role is actually underperforming?

Compare to the original 30-day scorecard. If you didn't write one, compare to what a competent human in that role would produce in the same timeframe with the same context. If the AI role is producing less than a junior human would, the problem is almost always context or autonomy configuration, not the AI. If it's producing less than a senior human would, check whether you're giving it senior-appropriate context. True 'this role cannot do this work' verdicts are rare; context and autonomy account for 80% of complaints.

Can I fire the AI CEO?

Yes, though it's the role most founders hesitate on because the CEO coordinates everything. If your CEO is underperforming, the symptoms are usually visible: weak weekly briefings, bad escalation judgment, drift across the rest of the team. The fix usually isn't firing; it's tightening the CEO's context (positioning doc, priority list, autonomy boundary). But if after a genuine 45 days the CEO isn't adding coordination value, swap it. The other specialists continue running during the swap because they have their own context; the coordination layer is what resets.

What if I don't know whether to fire or just re-onboard?

Default to re-onboard first. It's faster (a weekend vs a week of disruption), cheaper (no lost context), and usually sufficient. If after 14 days of intensive re-onboarding the role still isn't clicking, then fire. The re-onboard involves: redo the day 1 context dump, rewrite the voice guide, tighten the autonomy settings, clarify the 30-day scorecard, run an abbreviated version of the onboarding playbook. Most 'firing' impulses resolve during re-onboarding.

Should I tell the AI role why I'm firing it?

Yes — not for its feelings, but for the knowledge transfer step. Have a final session with the outgoing role where it summarizes everything it learned, flags what it thinks went wrong, and hands off context to the new role. This produces a noticeably better transition than a silent swap. It also sometimes surfaces insights you missed — the outgoing role often knows exactly why the role wasn't working in ways that would take you another month to figure out alone.

How often is firing an AI role actually necessary?

Rare if you follow the onboarding playbook. Most Tycoon operators run their original roles for 6-18 months before any swap is needed, and most swaps are because the business shape changed (outgrew the role's level) not because the role was wrong. If you're firing AI roles frequently (more than twice in your first 6 months), the problem is almost certainly in your hiring process — context loading, autonomy settings, or 30-day scorecard discipline — not in the AI roles themselves.

Related resources

Playbook

First AI Hire: Which Role to Start With

The blueprint for picking your first AI hire: decision framework, onboarding plan, and 30-day scorecard. Works for any stage.

Playbook

Onboarding Your AI CMO: Week 1 Playbook

The day-by-day plan for onboarding an AI CMO in week 1: context loading, voice matching, first deliverables. Copy this exactly.

Playbook

Run an AI-First Agency: Service Business Playbook

The playbook for running a service agency with AI staff — how to structure, sell, deliver, and scale. From 1 founder to $1M+ ARR.

Role

AI CEO | Hire Your AI CEO Today

Hire an AI CEO that coordinates your AI team, runs weekly priorities, and escalates only what you should decide. Direct by chat. Ship in 30 seconds.

Role

AI HR Manager | People Ops for Small Teams

Hire an AI HR Manager that handles offer letters, onboarding, PTO, and performance reviews. Direct by chat. Built for teams of 1 to 30.

Role

AI Knowledge Manager | Wiki & Company Brain

Hire an AI Knowledge Manager that curates your internal wiki, company brain, and playbooks. Direct by chat. Set up in 30 seconds.

Pillar

One-Person Company: Run a Solo Business With AI (2026)

A one-person company is a business run by a single founder with AI employees handling execution. The playbook — roles, stack, economics, examples.

Pillar

Hire an AI Team: Build Your AI C-Suite in 30 Seconds (2026)

Hire AI employees — CEO, CMO, CTO, COO, CFO, operators — who run your one-person company by chat. 30-second setup, no configuration, no agents to build.

Run your one-person company.

Hire your AI team in 30 seconds. Start for free.

Free to start · No credit card required · Set up in 30 seconds