Most first-90-days plans for a new CPO are politeness schedules. Meet everyone, listen a lot, find a quick win, do not break anything. I have run that plan. It produces a leader who is well liked at day 90 and operating on borrowed conviction at day 180. I have stepped into the product leadership seat several times, at Commure, Crisis Text Line, SOCi, and Smartcat, and the plan below is what survived. It treats the first 90 days as an audit with a deliverable, not a tour with a vibe.
Day zero, before you walk in
The highest-leverage hours happen before day 1. Become a customer: sign up with a personal email, hit the paywall, file a support ticket, try to cancel, because this is your only unanchored read. Read the unguarded documents: the last four board decks and the last two postmortems, because the gap between the story and the truth it admits under pressure is your first map. And write your ten priors as falsifiable statements, dated, because the first 30 days are designed to break them and you cannot notice your model updating if you never recorded it.
Days 1 to 30: the four audits
Run four audits in parallel. The reality audit is the product on real tasks plus raw customer calls. The money audit gets cost per outcome visible by workflow, and in every AI-era org I have looked at, at least one workflow was quietly underwater and nobody owned the number. The quality audit lists every production AI feature and asks its eval score and trend, and the size of the silent pile is your real quality roadmap. The decision audit reconstructs how the last ten significant decisions actually got made, which is the one nobody runs. Underneath all four runs the trust ledger: log every commitment from hour one, review it every Friday, close or renegotiate everything. New executives bleed credibility through small dropped promises they never noticed making.
Days 31 to 60: instrument and align
Convert the audits into instruments and the conversations into a map. Stand up cost per outcome by workflow and eval scores with trend lines permanently. Build the coalition map: allies, persuadables, blockers, and what each needs to move, because your first bets live or die on this map, not on their merits. Ship one visible decision, not a quick win. Kill one zombie initiative everyone privately agrees is dead. The point is the precedent: decisions here now get made on evidence, in the open, and they stick.
Days 61 to 90: kill list, bets, readout
Write the kill list with what each item costs and what killing it frees up, and kill two publicly with a written explanation. Place two or three evidence-backed bets and name the anti-bets, because an anti-bet list converts a hundred future hallway debates into one paragraph. Then deliver the day-90 readout: one SCQA-shaped memo with five sections (where the product actually is, the two or three constraints that matter, the bets and anti-bets, the kill list, and the scoreboard you want to be judged on), presented to the exec team and then the board. That document is the actual deliverable of your first 90 days.
If you are about to start, write your ten priors as falsifiable statements and date the file this week.