CEO tenure just fell to a record-low 7.2 years, and only 40% of CEOs now say they're ready for AI-driven disruption. Boards are responding by reviving a model most of us wrote off as a governance relic: the co-CEO. The numbers aren't crazy — an HBR analysis of 87 co-CEO firms found 9.5% average annual shareholder returns versus 6.9% for their benchmarks, with roughly 60% beating peers.

The model works. It also fails badly. Here's the playbook I'd hand a board.

Decide from capability math, not politics

  1. Map the actual job. If it splits cleanly — external strategy and stakeholders on one side, technical and operational depth on the other — and one hire can't credibly do both, two leaders is rational. Don't frame it as a personality compromise.

Contract the boundaries before the titles

  1. Write decision rights on one page. Who owns capital allocation, product, go-to-market? Co-CEOs die in ambiguity: stalled decisions, conflicting communications, unclear ownership. If you can't write it, don't split it.

  2. Test chemistry in public, cheaply. Run paired operating roles for 90 days before granting titles. The two don't need to agree privately; they must appear unified externally.

  3. Name the tiebreaker up front — a board chair or standing committee that breaks deadlocks. No tiebreaker means drift.

Set a sunset

  1. Treat co-leadership as an instrument, not a permanent state. It works best as a merger bridge, a founder-to-professional handoff, or a predecessor–newcomer pairing. Schedule a 12-month review to confirm, re-scope, or end it.

Two heads beat one only when the division of labor is contractual, not aspirational. That's the difference between a 9.5% return and a leadership vacuum.