Why the sequence matters more than the speed
The instinct in most companies is to start building immediately — pick a tool, automate something visible, show quick motion. I understand the instinct and I resist it anyway, because building before you've mapped the business almost always means automating the wrong thing first: the loud problem instead of the expensive one. The 90-day structure exists to fix that ordering problem before a single system goes into production.
I run this exact arc for every client I take on as a fractional AI officer, and I run a version of it inside my own portfolio companies whenever we take on a new initiative. It's not a sales framework — it's the sequence that actually works.
Days 1-30: Audit and roadmap
The first 30 days are entirely diagnostic. No building yet. The work breaks into three parts:
- Workflow mapping. Every meaningful process in the business — sales, marketing, operations, customer-facing work, internal admin — gets mapped for time cost, error rate, and how much it actually matters to revenue or margin.
- Automation scoring. Each workflow gets ranked on feasibility: how automatable it actually is today, not in theory, given current tools and the company's data.
- The roadmap. A ranked, P&L-anchored build list — the two or three highest-leverage systems to build first, with an honest estimate of what each will save or generate.
The output isn't a vision deck. It's a prioritized list any operator in the business could read and understand why item one comes before item four.
Days 30-60: Build and measure
This is where the top items from the roadmap actually get built and shipped into production — most commonly the AI marketing engine (content, SEO, and automation), sales and proposal automation, or internal reporting and research copilots, depending on what the audit surfaced as highest-leverage for that specific business.
Every system gets a metric attached from day one — hours saved, error rate, throughput, pipeline velocity, whatever's actually measurable for that workflow — so results get reported as numbers, not impressions. If a system doesn't move its metric within a reasonable window, it gets adjusted or replaced rather than left running on faith.
Days 60-90: Train and hand over
The last third of the engagement is where a lot of AI initiatives quietly fail elsewhere — the outside expert leaves and nobody inside the company can run what got built. I structure this phase specifically against that failure mode:
- Documentation. Every system gets a written playbook — not a wiki page nobody opens, but a document the team actually uses when something breaks or needs adjusting.
- Training sessions. Working sessions with whoever owns each system internally, not a one-time demo.
- Operating cadence. A set rhythm — weekly or biweekly — for reviewing what's working, what needs fixing, and what to build next, so the systems keep compounding instead of decaying the month after the engagement's initial phase ends.
My job, done well, includes making myself replaceable on a schedule. If a client still needs me to run the same system in month twelve that we built in month two, the training phase failed — regardless of how well the system itself performs.
What happens after day 90
Two legitimate paths, and I tell clients this directly before they sign anything:
- The retainer continues, shifting from build mode to a sustaining and expanding cadence — new systems get added as the highest-leverage opportunities change, and I stay the accountable owner of the function.
- The company takes it in-house. Once the systems are built, documented, and the team is trained, some clients are ready to run it themselves. That's a legitimate outcome, not a failure of the engagement — see what a fractional AI officer actually does for how that handoff decision usually gets made.
What determines whether 90 days is realistic
The arc holds for most companies in the $1M-$50M range, but a few things stretch or compress it: how much of the audit work already exists internally, how many stakeholders need to sign off on new tools, and how much data hygiene has to happen before automation is safe to turn on. I flag this honestly during the intro call rather than promising a uniform timeline that doesn't survive contact with a messy CRM.
If you want to see how this maps to what actually gets run in your business, here's how the engagement works and what it costs.
What clients underestimate going in
Two things surprise most founders about the first 90 days. The first is how much the audit phase surfaces that has nothing directly to do with AI — broken handoffs between sales and delivery, data that lives in someone's head instead of a system, a marketing calendar that exists only informally. AI doesn't fix organizational debt; it exposes it faster than anything else, because you can't automate a process that isn't actually a process yet. Part of the audit's value is simply making that debt visible before it becomes the reason a later automation fails.
The second surprise is how little of days 30-60 is about the AI tools themselves. Choosing a model or a platform is usually the easiest decision in the whole build. The actual work is in wiring the system correctly to the business's real data and real workflow, and in getting the humans who'll use it to trust it enough to actually adopt it instead of quietly working around it. A technically perfect automation that the team routes around isn't a win — adoption is part of the deliverable, not an afterthought.
Frequently asked questions
What happens in the first 30 days with a fractional AI officer?
A full operations and marketing audit: mapping where hours and dollars go, scoring workflows on automation feasibility, and producing a ranked, P&L-anchored roadmap. No building happens yet.
What gets built in days 30 to 60?
The two or three highest-leverage systems from the audit get built and put into production, each with a metric attached from day one so results can be measured against a baseline rather than described anecdotally.
What happens in the last 30 days, days 60 to 90?
The internal team is trained to run and extend the systems, the playbooks get documented, and an operating cadence is set so the systems keep running and improving without the fractional officer in the room.
What happens after the first 90 days?
The retainer either continues to sustain momentum and build the next layer of systems, or the company takes over running what was built. Both are legitimate outcomes, and a good fractional AI officer makes either path work.
What if the audit finds nothing worth automating?
It happens rarely, but it's a possible and honest outcome. If a business is small enough, specialized enough, or already lean enough that the audit doesn't surface a clear P&L case for building anything, I say so directly rather than manufacturing a project to justify the engagement. That kind of honesty is part of what makes the audit phase worth paying for in the first place — a diagnosis you can trust, whichever direction it points.