Your first 90 days set the ceiling on everything that follows. Here is the sequence — listen, define, execute — that separates data leaders who build durable mandates from those who spend year one re-earning trust.
A 30/60/90 day plan for a new data or AI leader follows one sequence: listen, define, execute. Days 1–30 are for stakeholder mapping, a current-state audit of the data estate, and an honest assessment of the team. Days 31–60 are for drafting the data strategy, landing one or two visible quick wins, and sharing a roadmap the executive team has actually shaped. Days 61–90 are for shipping the first deliverables, publishing the first metrics, and standing up lightweight governance. Get the order right and the mandate compounds; get it backwards and you spend year one apologising.
Data leadership roles carry a structural disadvantage that most executive roles do not: the organisation has usually been disappointed before. By the time a company hires its first head of data — or replaces one — there is a history. A warehouse project that ran long. Dashboards nobody opens. An AI pilot that produced a demo and nothing else. You are not starting from zero; you are starting from a deficit of trust that you did not create.
That is why the first 90 days matter disproportionately. Executive patience for data investment is a renewable resource, but it renews on evidence. In the first quarter, everything you do is being read as a signal: are you a builder or a talker, do you understand the business or just the technology, will this be different from last time? The 30/60/90 structure exists to send the right signals in the right order — comprehension first, direction second, delivery third.
There is also a practical reason. The observations you can make in your first month — the candid interviews, the unguarded complaints, the ability to ask naive questions without embarrassment — are only available while you are new. A leader who skips the listening phase does not get to run it later. The window closes.
The first month has three workstreams, and none of them involves committing to a build.
Stakeholder mapping. Book thirty to forty-five minutes with every executive and every senior operator who consumes data — not just the ones who report to you or asked for you. Ask the same four questions in every session: what decisions do you make on a recurring basis, what data do you use to make them, where does the current setup fail you, and what would you fix first if you owned my role for a week? Write the answers down verbatim. By interview twelve, patterns emerge that no architecture diagram will show you: the metric two departments calculate differently, the spreadsheet the CFO's analyst maintains because nobody trusts the dashboard, the report that takes four days to produce and is read by no one.
Current-state audit. In parallel, inventory the estate. What platforms exist, what they cost, what actually runs on them. Where the pipelines are fragile, where the data quality problems concentrate, where access is either dangerously open or uselessly locked down. What reporting exists, and — separately — what reporting is used. Which vendors are embedded and which contracts renew soon. This does not need to be exhaustive; it needs to be honest. A two-page findings memo beats a forty-page inventory.
Team assessment. Meet every member of the team individually. You are assessing three things: capability, morale, and where the informal knowledge lives. Most inherited data teams contain at least one person who quietly holds the entire estate together; find that person in month one, because your roadmap depends on knowing what they know. Resist reorganising in the first 30 days — you do not yet know enough to redraw the boxes.
The month-one deliverable is a diagnosis, not a plan: here is what I found, here is what it costs us, here is what I will bring you by day 60. Presenting a diagnosis before a strategy feels slow to ambitious leaders. It is the opposite — it is the only way the strategy will be believed.
Month two converts evidence into direction, in three moves.
Draft the data strategy — with the business, not at it. The strategy document should be short enough to survive an executive read: the three to five business outcomes the data function will serve, the capability gaps between here and there, the sequence of investments, and what you are explicitly choosing not to do. Circulate the draft to the stakeholders you interviewed and let them mark it up. A strategy the executive team has edited is a strategy the executive team will defend. The interviews from month one are what make this possible — every priority in the document should trace back to something a named leader told you.
Land the quick wins. Your month-one interviews surfaced irritations that are small in engineering terms and large in trust terms: the broken report, the reconciliation that eats two days of every close, the definition dispute that derails every monthly review. Pick one or two. Fix them properly, ship them by day 60, and tell the story — which stakeholder raised it, what it cost before, what it looks like now. The point of a quick win is not the win; it is the demonstration that listening leads to shipping.
Share the roadmap. Close month two by presenting the sequenced roadmap: what lands in the next quarter, the next two quarters, the next year, with owners and dependencies visible. Include the AI question honestly — most organisations at this stage need foundation work before model work, and saying so out loud, with the readiness evidence to back it, builds more credibility than promising a pilot you know the data cannot support.
Month three is where the plan either becomes real or becomes a slide deck.
Ship the first roadmap deliverables. Not the quick wins — those were month two. This is the first item from the actual roadmap: the first certified data product, the first rebuilt executive scorecard, the first pipeline moved onto stable footing. It should be visibly connected to the strategy you circulated, so the organisation sees the document producing outcomes on schedule.
Publish the first metrics. Instrument your own function before anyone asks you to. Report a small set of numbers monthly: delivery against the roadmap, data quality on the domains you have touched, adoption of what you have shipped, and spend against plan. A data leader who measures the data function earns the standing to measure everything else.
Establish governance — lightweight and real. By day 90 you need three things running: named ownership for the critical data domains, a definition process for the metrics that matter (one governed definition per KPI, with a change process), and a decision forum — a monthly data council is enough — where prioritisation happens in the open. Governance introduced in month three, attached to real deliverables, reads as professionalism. The same governance introduced in month one, before any delivery, reads as bureaucracy.
Everything above is the same sequence we run as a firm. Our engagement model is built around a structured 30/60/90 Roadmap — diagnose the current state in the first month, co-author the strategy and land early wins in the second, and ship governed, measurable deliverables in the third. We run it for organisations that do not yet have a senior data leader, and alongside newly hired leaders who want an experienced partner to accelerate the audit and pressure-test the roadmap before it goes in front of the board.
The reason we productised the sequence is the reason this article exists: the failure modes are predictable, and so is the fix. If you are stepping into a data or AI leadership role — or hiring for one — the roadmap page lays out the full methodology, deliverable by deliverable, and our free assessments are a fast way to get the current-state evidence your first 30 days depend on.
Spend the first 30 days listening, not building. Run structured interviews with every executive stakeholder, complete a current-state audit of the data estate — platforms, pipelines, reporting, data quality, and spend — and assess the team you have inherited. Resist announcing a strategy before you have evidence; the credibility of everything you propose later depends on the accuracy of this diagnosis.
Pick one or two visible, low-risk fixes surfaced in your first-month interviews — a broken executive report, a metric two departments define differently, a slow pipeline that delays month-end. Ship the fix by day 60, name the stakeholder it helped, and tie it explicitly to the larger roadmap so it reads as a down payment on the strategy rather than a one-off favour.
Committing to a re-platforming or a large build before finishing the diagnosis. Leaders who announce a technology decision in week three are anchoring the organisation to a guess. The second most common mistake is the opposite: auditing for six months and shipping nothing, which spends the honeymoon period without banking any credibility.
The plan should assess AI readiness from day one but should rarely ship AI in the first 90 days. Month one establishes whether the data foundations — quality, access, governance — can support AI at all. By day 90 you should have an honest readiness position and one or two scoped candidate use cases, which is far more valuable than a rushed pilot built on data nobody trusts.
Listen, define, execute — in that order, on that clock. Diagnose in month one, co-author the strategy and land a quick win in month two, ship governed deliverables and publish your own metrics in month three. If you want the full methodology with deliverables and checkpoints mapped out, start with The 30/60/90 Roadmap — or book a discovery chat with Maria to talk through your first quarter.