# What the first ninety days look like > Justas Butkus is a fractional AI officer based in Vilnius, Lithuania, founder of AINORA, MB and of Impetora, and a graduate of ISM University of Management and Economics. Justas Butkus is a fractional AI officer based in Vilnius, Lithuania, working with mid-market companies across the UK, EU and US at two to four days a month. Every engagement starts with a fixed-scope diagnostic, runs on written milestones rather than hours, and includes documented handover as a deliverable. **A fractional AI officer engagement starts with a fixed-scope diagnostic, then runs on ninety-day milestones. The first month establishes what AI is already in use and what is worth doing; the second produces a decision-ready plan and a governance structure; the third puts the first work into production.** Canonical: https://justasbutkus.com/fractional-ai-officer/how-it-works/ Last updated: 2026-08-01 --- ## It starts with a diagnostic, always No ongoing arrangement begins before both sides know what the work actually is. That is not caution, it is arithmetic: the most expensive mistake available in this category is committing to a direction before anyone has looked at the data, and it costs far more than the diagnostic does. The diagnostic is fixed in scope and has a defined end date. It produces a written roadmap you keep regardless of whether anything follows. If the honest conclusion is that you do not need an ongoing arrangement, that is what it will say. ## Month one: find out what is actually true Almost every company overestimates how much it knows about its own AI usage, and underestimates how much is already happening without approval. 1. **Inventory what is already running** — Every AI tool in use, sanctioned or not, and what data has passed through it. This is usually the uncomfortable part. 2. **Map the repeating processes** — The expensive, frequent, unloved ones. Not the exciting ones, the boring ones with volume behind them. 3. **Establish who is accountable for what** — Usually the answer is nobody, which is the finding rather than a failure. 4. **Test the data** — Not whether it exists, but whether it is good enough to build on. This kills more projects than the technology does. ## Month two: decide, and write it down The output is a plan someone can be held to, not a strategy document that reads well and commits to nothing. - **A ranked shortlist**, with the reasoning for what was excluded. What you decided not to do matters as much as what you chose. - **A governance structure** proportionate to the company. Who approves what, what happens when a system produces a wrong answer, and who is accountable for each system. - **A build-versus-buy position** for each item, with the reasoning attached. - **Something the board can read.** If it needs translating before it goes upstairs, it was written for the wrong audience. ## Month three: put something in production The first thing shipped is deliberately not the most ambitious item on the list. It is the one that proves the pipeline works end to end: specification, build, evaluation, oversight, rollback. Companies that begin with their most ambitious use case almost always stall, because they discover the hard parts of operating AI systems while simultaneously attempting the hard parts of the use case. Doing them in that order is what produces the pilot graveyard. ## How the engagement ends It should end. An arrangement that never does has quietly become a permanent role that nobody scoped properly. - **You hold the keys throughout.** Accounts, infrastructure, prompts, runbooks and documentation are yours from day one, not handed over at the end. - **Handover is a deliverable**, written into the engagement rather than negotiated when you want to leave. - **The success condition is redundancy.** Either an internal person can now hold the mandate, or the role has been defined clearly enough to hire for permanently. ## Frequently asked questions ### How many days a month does a fractional AI officer work? I work two to four days a month. The wider market ranges from about two days a month up to three days a week, typically with a more intensive first ninety days that scales back once the strategy is embedded and internal people can carry it. ### What does a fractional AI officer deliver in the first 90 days? An inventory of AI already in use and its risks, a ranked roadmap with reasoning for exclusions, a governance structure proportionate to the company, and at least one thing actually in production with monitoring and a rollback path. ### How long does an engagement last? Long enough to embed a working function and no longer. Typically the arrangement runs through a first ninety days and then continues at a lighter cadence, ending when an internal person can hold the mandate or the permanent role has been defined well enough to hire. ### What happens to the systems if the engagement ends? They are yours and always were. Accounts, infrastructure and documentation sit under your control from the start, and documented handover is a contractual deliverable rather than something negotiated at the end. ### Do you write the code yourself? Yes, where building is in scope. That is the distinguishing feature of this arrangement: most engagements in the category pair an advisor for strategy with separate engineers for delivery, and the handoff between them is where most of the value leaks away. ## Related - [What a fractional AI officer does](/fractional-ai-officer/) — The role, and when a company needs one. - [How to choose one](/how-to-choose/) — What to ask before you engage anybody. - [When pilots never reach production](/ai-implementation/) — Why AI projects stall, and what actually unblocks them. ## If you want to know what the first ninety days would look like here The honest way to answer that is to look at your actual processes rather than describe a generic plan. Contact: justas@ainora.lt · [LinkedIn](https://www.linkedin.com/in/justas-butkus/) ## Other languages - EN: https://justasbutkus.com/fractional-ai-officer/how-it-works/ - LT: https://justasbutkus.com/lt/dirbtinio-intelekto-vadovas/kaip-vyksta/