AI Onboarding

From zero to a running AI pilot in 90 days — safely.

Most "AI strategy" engagements end at a slide deck. We end at AI doing real work in your business — with policy, governance, secure tooling, and trained users in place. A structured 90-day path so the pilot you launch is the foundation, not the experiment that gets killed.

Book a discovery call See the framework

The path no one runs the same way twice — until they run it right.

Most companies trying to adopt AI bounce between two failure modes. Mode one is the slide deck: an outside consultant produces a "transformation roadmap" with three swimlanes and an executive summary, leadership nods, nothing ships. Mode two is the rogue pilot: a smart manager builds something in a free tool over a weekend, it does something useful, it spreads through the team, IT discovers it three months later, and now there is a compliance problem.

Neither produces durable AI capability. The first never gets to running code. The second gets there, but on shaky ground that creates exposure when it scales. The structured 90-day onboarding we run is built to land between them: deliberately fast, deliberately small in scope, deliberately serious about the governance posture so the pilot becomes the foundation for everything that comes next.

The day-one promise is straightforward. At the end of ninety days you have one AI project doing real work — picked together for low risk and high learning — plus an AI usage policy your legal counsel signed off on, vendor contracts on the right tier with the right terms, access controls that scope who can do what, audit logging your auditors will accept, and a documented playbook your team can use to evaluate the next opportunity without us in the room.

The four phases.

Ninety days broken into four phases, each with a specific deliverable. The pace is calibrated to be aggressive without being reckless — slow enough to do the work properly, fast enough that the program does not stall in a quarterly review.

Phase 01

Weeks 1–2

Readiness assessment

Audit existing AI tool usage across the org
Map data classifications & regulatory scope
Review identity, MFA, and access posture
Draft acceptable-use policy v1
Pick the pilot project together

Phase 02

Weeks 3–5

Tool selection & secure setup

Evaluate vendors for the pilot use case
Negotiate enterprise tier & data terms (BAA where needed)
Configure SSO, role-based access, audit logging
Establish governance rhythm (who owns what)
Document the architecture decisions

Phase 03

Weeks 6–10

Pilot build & user training

Build the pilot workflow (with your team in the room)
Train the staff who will use it
Stand up monitoring & usage analytics
Iterate based on first-two-week feedback
Run a security & compliance review of the live system

Phase 04

Weeks 11–12

Handoff & playbook

Deliver final documentation set
Train internal owner on the operating runbook
Identify next 2-3 candidate projects
Establish renewal cadence for vendor & policy review
Optional: monthly advisory engagement going forward

Sample pilots — to make the abstract concrete.

The right pilot is project-specific, but here is what they tend to look like in practice. These are anonymized versions of engagements we have actually run.

Internal knowledge assistant for a regional accounting firm.

Private chat over the firm's policies, procedures, and client engagement standards. Staff get instant answers without paging a partner. Built on Claude with documents in a contained workspace; SSO via the existing IdP; usage audited monthly. Time-to-pilot: 8 weeks. Saved an estimated 4–6 hours of senior time per week within the first month.

Inbound ticket triage for a regulated services group.

Incoming support requests classified by type, urgency, and routing target. Replaces a manual queue review that one of the operations leads was doing for an hour each morning. Built on the existing ticketing platform with Claude API. Audit log captures the classification reasoning. Time-to-pilot: 6 weeks.

Contract first-pass review for an SMB legal team.

Vendor contracts pre-reviewed against a standard playbook before human review. Flags missing clauses, unusual terms, and renewal cliffs. Reduced average human review time by an estimated 40%. Time-to-pilot: 10 weeks. Required the most careful data governance work of any pilot — contracts are sensitive.

Monthly close acceleration for a mid-market manufacturer.

Automated variance commentary on the monthly P&L, with the controller reviewing and editing rather than writing from scratch. Built on a combination of Excel data, Claude, and a small n8n workflow. Time-to-pilot: 9 weeks. Reduced close cycle by 2 days within the first quarter.

What's included.

Every onboarding engagement includes the following deliverables. Custom scope items get added in the discovery call — these are the baselines.

AI readiness report. A documented current-state assessment covering tool usage, data classifications, identity posture, regulatory scope, and the gaps that need closing before broader AI rollout.
AI usage policy. A drafted, legal-reviewable policy your staff can read in five minutes, scoped to your actual environment. Not a template downloaded from the internet.
Vendor selection & contracts. We evaluate, negotiate, and document the vendor tier and contractual terms for the AI tools you adopt. BAAs where regulatory scope requires them.
Access & governance configuration. SSO, role-based access, audit logging, monitoring. Configured for the tools we onboard, with the configuration documented for your records.
One running pilot project. Built with your team in the room so they understand how it works and can maintain it after handoff.
Operating runbook. Who runs what, what breaks how, when the vendor contract renews, when to do the next access review. The day-two operational playbook.
Pilot user training. The staff who will use the pilot get trained on it before launch. Continuing reinforcement is offered through our team training programs or on-demand courses.
Next-project recommendations. A documented shortlist of the 2-3 highest-value AI projects to consider next, with rough scope and the ordering rationale.

Common questions

Three things. First, identity and access: every user has a managed identity, MFA is enforced, you can revoke access cleanly. Second, data hygiene: you can articulate what's confidential, what's not, and what regulatory class each category falls into. Third, basic policy: you have at minimum an acceptable-use statement for AI tools that staff can read in five minutes. Most companies fail one or two of these. Getting them addressed is week one of onboarding.

A running pilot project — meaning AI is doing real work in your business — plus the documentation around it: AI usage policy, vendor contracts, access control configuration, governance plan, audit logging, and training records for the staff using it. You should be able to point to it in your next audit, your next board meeting, or your next insurance renewal and explain how it works in two minutes.

We pick the pilot together in week 2, after the readiness assessment. The criteria are deliberately conservative: low data-sensitivity exposure, clear business case, contained scope, fast feedback loop. The first pilot is rarely the highest-value AI project in your business — it's the project that proves the model so you can do bigger ones with confidence. We'd rather a "small but undeniable" result in 90 days than an ambitious thing that doesn't ship.

Optional, and structured so you don't have to. The 90-day engagement is designed to leave you operating it yourself, with documentation strong enough that a new internal owner can take it over. We offer ongoing monthly advisory engagements for companies that want continuity — typical scope is 4-8 hours/month of strategy, vendor evaluation as new tools emerge, and audit support — but it's not required.

Engagements scope by environment complexity and regulatory burden — a 25-user professional services firm and a 250-user multi-state claims operation are different conversations. The 20-minute discovery call covers actual numbers; we don't publish a price sheet because every meaningful engagement we've quoted from one has been wrong.

Twenty minutes to scope it.

Tell us where you are and what you'd want to ship. We'll tell you whether 90 days is realistic and what it'd look like.

Book a discovery call Back to AI Services