Fractional AI Leader: The First 90 Days, Step by Step

This guide is for the CEO, COO, founder or board chair of a mid-market company who has engaged, or is about to engage, a part-time AI leader and wants to know what good looks like before the first invoice. Policy content, such as risk tiers and approval rules, sits in our guide to AI governance for mid-market companies; the roles and engagement models are on the fractional AI leadership service. This page is the sequence that connects them.

In this guide

• What should a fractional AI leader deliver in the first 90 days?
• Days 1-30: how is the mandate set, and what does the leader inventory?
• Days 31-60: how does the leader build the portfolio and the governance baseline?
• Days 61-90: what does a handover plan contain?
• The 90-day plan as a checklist
• Who should own AI in the first 90 days?
• Where AI fits
• What delivery record exists, and what does not
• Independence: what to check when the leader's firm also builds
• Alternatives: when a 90-day fractional mandate is the wrong tool
• Limits of this guide
• Frequently asked questions
• Next step

What should a fractional AI leader deliver in the first 90 days?

Four workstreams run in parallel, but each one has a clear finishing point.

Netbase structures fractional AI work as outcome-based milestones, in which each milestone has a written outcome and acceptance criteria agreed before it starts. Outcome-based here means scoped deliverables, not a promised business result or a price guarantee.

The handover plan appears in the first 90 days on purpose. A fractional role exists to fill a seat for a period, and Netbase's fractional technology leadership engagements are written with a handover to a permanent hire. If nobody knows who inherits the work, the company is renting a leader rather than building capability.

Days 1-30: how is the mandate set, and what does the leader inventory?

The first month is about authority and facts, not building.

The mandate. It answers five questions in writing: who the leader reports to, which decisions the leader can take alone, which ones need the sponsor, what is out of scope (for example hiring, contracts or vendor payments), and how often the leader meets the sponsor.

The inventory. The leader lists every AI tool, pilot, vendor contract and data flow already in use, including tools staff adopted on their own. Each item gets an owner, or the note "no owner". This list is the raw material for both the portfolio and the governance baseline.

The stakeholder map. One conversation with each department head that uses or wants AI, recorded as problems and data, not as tool requests. A RACI chart shows early where two people believe they own the same decision.

Day-30 checkpoint: the mandate is signed, the inventory exists, and the leader can name the three questions that most need a decision. Failure signal: the leader presents a vendor shortlist before the inventory.

Days 31-60: how does the leader build the portfolio and the governance baseline?

The second month turns the inventory into choices; portfolio and baseline are built together, because risk changes rank.

The use-case portfolio. Every candidate gets the same fields: the business problem, the owner, the data it needs and whether that data is reachable, the risk level, and the reason for its rank. The leader also writes down what was left out and why. A portfolio without exclusions is a wish list.

The governance baseline. The minimum set of rules the company needs before any AI use case moves past a pilot: who approves a new use case, which actions need a human decision, how incidents are reported and who can stop a system. The NIST AI Risk Management Framework organises this work in four functions (govern, map, measure and manage), and its Playbook offers voluntary suggested actions that a company can adopt selectively. A mid-market baseline starts with govern: named roles and decision rights.

The first pilot decision. By day 60 the leader recommends whether to start one pilot from the top of the portfolio, and on what terms. Our first-90-days plan for agentic AI describes how such a pilot is gated once it starts.

Day-60 checkpoint: the sponsor accepts the ranked portfolio and signs off the approval points. Failure signal: the governance baseline is a policy copied from elsewhere that names no person in your company.

Days 61-90: what does a handover plan contain?

The third month proves the plan survives without the leader in the room.

Owners. Every portfolio item, governance role and open decision has an internal owner. Our guide to AI agent operating models sets out the owner and escalation roles that agent work needs once it runs.

The decision log. A dated record of each decision taken in the 90 days, who took it and what alternatives were rejected.

The next-term plan. The leader recommends one of three paths: continue the fractional role for a defined term, hire a permanent AI leader (with a role profile and interview support), or close the mandate because an internal owner can carry it.

Day-90 checkpoint: the sponsor accepts the handover plan and the decision log. Failure signal: the recommended next step is "continue as is" with no change in scope.

Who should own AI in the first 90 days?

A fractional leader is one of four ways to staff this period. The right choice depends on whether the company needs a decision-maker, a report or a builder.

Decision rule: if the company cannot name who would accept the four deliverables, start by naming a sponsor; no option works without one.

What delivery record exists, and what does not

Netbase offers fractional AI and digital-transformation leadership as part-time and per-project roles, including a fractional Chief AI Officer, an AI transformation advisor, a project-based AI programme lead and an independent board or advisory seat, delivered remotely first. Huy Nguyen (David), Netbase's Founder and CEO, takes part personally as advisor, board member or independent expert.

The engagement models are a part-time retainer, a per-project sprint and outcome-based milestones. Most Netbase projects are delivered on fixed-price contracts agreed after discovery. Netbase's delivery lifecycle includes discovery, outcome-based milestones and training and rollout, which is the practice a fractional leader draws on.

The AI delivery behind the advice is the recommendation engine built for 4over4's online printing store, the one named delivered AI feature, plus AI projects for clients we do not name, including retrieval-augmented knowledge assistants, document AI and MLOps pipelines.

What does not exist: no named fractional AI or board engagement is published, and we state no count of mandates, no duration and no result. The 90-day plan on this page is the way we structure the work, not a record of past outcomes.

Independence: what to check when the leader's firm also builds

A leader from a firm that also builds software knows what delivery takes, but the conflict has to be managed in writing.

In Netbase's advisory and board roles, any Netbase interest in a recommendation is disclosed in writing before the client decides, and the leader takes no part in selecting or evaluating Netbase as a vendor for the same client unless the client asks and the interest is disclosed. Formal board directorships follow the client's governance and applicable law. For the company, accountability starts with knowing who gains from each recommendation.

Alternatives: when a 90-day fractional mandate is the wrong tool

• A single AI feature to build. If the decision is already taken, a fixed-scope build is faster than a leadership mandate.
• Only a readiness score is needed. An AI strategy and readiness assessment gives a baseline without an ongoing role.
• No internal sponsor. A fractional leader without a sponsor who can accept deliverables produces documents nobody adopts; fix sponsorship first.

Limits of this guide

• The plan assumes a mid-market company with some AI activity already under way; a company with none may spend the first month on education instead of inventory.
• The 30-60-90 split is a planning frame; a regulated company may need the governance baseline first.
• Nothing here is legal, audit or investment advice, and a board seat always follows the company's own governance.

Next step

Planning AI leadership without a full-time hire? Book a solution review to agree a mandate and acceptance criteria for the first 90 days.

Related services and solutions