Straight answers before we talk.

A virtual CAIO is the accountable owner of your AI decisions before a full-time executive hire makes sense. Concretely, I decide what starts, what waits, and what gets killed: I build a ranked portfolio of AI opportunities tied to real workflows, metrics, and owners; set the governance guardrails before anything ships; oversee the people who implement; and run a 90-day cadence that measures value honestly. I am not a committee, a vendor demo parade, or a deck-and-leave strategist. I'm the person whose job is to make AI produce defensible value and to tell you the truth about what won't.

Three differences. Ownership vs. advice: a consultant ships a deck and leaves; I own the decisions, the success metrics, and the cadence until the work is producing value or has been honestly killed. Neutrality: an agency sells you the build it happens to do; I am driven by your ROI and the primary needs identified by your business assessment. Dose: a big firm brings many bodies and a board-safe brand at a price and pace that rarely fits a mid-market firm — I bring one accountable operator at a fractional cost. I'm the leadership layer above your tools and partners.

I'm accountable for the outcome; the implementation partner is accountable for the build. That split is deliberate, and it's in your favor. I own the strategy, the success metric, the baseline measurement, partner selection, and the oversight that holds delivery to the plan — so there's a single throat to clear on whether this is working. You're not managing two vendors: I manage the partner against the metric we agreed on, and you deal with me. I'm also up front about my relationships with the partners I bring in — I'd rather you know exactly where I stand than wonder. What you're buying is one accountable owner who won't hide behind “that was the vendor's job.”

Everything I do is priced in plain numbers, in USD. The AI Strategy Assessment is $25,000–$35,000, one-time, delivered in 2–3 weeks. The ongoing virtual CAIO retainer runs $5,000 / $10,000 / $15,000 per month (Foundation / Growth / Transformation), three-month minimum, scalable up or down on 30 days' notice. Workshops are $6,000 (half-day) or $10,000 (full-day). On-site adds travel. Those fees cover the advisory work — strategy, governance, partner selection, and oversight. The build itself is quoted separately by whichever implementation partner does it, so you always see that cost on its own and know who's being paid for what. No “request a quote” wall — if a number is going to be a problem, you should know on the first visit, not the third call.

We set the baseline before anything is built — current cycle time, cost per matter, hold time, whatever the real metric is — and we agree what “better” means in numbers. Then the 90-day cadence checks it honestly. If the pilot hits its kill criterion, we stop; sunk cost is not a ranking criterion. You'll never get a vague story from me — you'll get “intake went from 45 minutes to 8,” or you'll get a clear-eyed decision to shut it down before it costs you more. About a quarter of AI projects don't pan out; the discipline is deciding that early and cheaply, on evidence, instead of defending a doomed pilot for a year.

Yes to an NDA, as a matter of course, before you share anything sensitive. For firms handling protected health information — clinics, and the PI and med-mal lawyers who touch it — yes to a Business Associate Agreement as well. Practically: I work in your environment and your tools where possible; I don't run your data through consumer AI products; and any tool I'd recommend for production has to not train on your data, full stop. Data classification, retention, and a clean hand-off at the end of the engagement are part of the governance work, not an afterthought.

Let me be precise, because “neutral” gets oversold. Two things are true and I'll stand behind both. First, I'm not paid to send you anywhere: I take no commissions, referral fees, or finder's bonuses — from software vendors or from implementation partners — for bringing them your work. That incentive simply doesn't exist in how I'm paid. Second, I won't pretend I have no relationships. I do ongoing fractional work for some of the partners I might bring in, and in at least one case I have a financial interest in a partner I'd recommend. So when a partner I'm connected to is genuinely the right fit, I'll say so — and I'll disclose the relationship in the same breath, every time, so you're never finding it out later. The guardrails that make that safe for you: I always disclose a relationship before I recommend anyone — every time, without being asked; you always see more than one option and you make the call; I'll bring in a vendor I have no relationship with whenever they're the better fit, or simply because you'd prefer one — your asking is reason enough on its own; and every partner, including the ones I'm closest to, is held to the same metric, the same oversight, and the same kill criterion as anyone else. Independence here isn't “I know no one.” It's “you always know my interest, you can always choose someone I'm not tied to, and you always keep the decision.”

That's the wrong frame, and it's usually how these projects fail. The work I target is the unglamorous throughput drain — intake, routing, document prep, triage — so your people stop spending billable-grade time on clerical motion. In practice that restructures the work, not the headcount: the law firm freed 40 paralegal hours a month for billable work; the clinic moved 2.5 staff to patient care. If your only goal is cutting heads, I'm probably not your guy — the durable wins come from making the team you have more valuable, with humans kept in the loop on every decision that matters.

Because the failures I'm hired to prevent are operational, not mathematical. Most mid-market AI projects don't die on the model — they die on process, permissions, data quality, adoption, and measurement, which is exactly the unglamorous territory I've worked in for 30+ years. I don't train models; I decide which problems are worth solving, whether the data and process can support it, and how to ship it without creating your next compliance mess — then I bring in specialists for the build. The best AI implementers tend to be operators who work alongside engineers, not engineers looking for somewhere to apply a model. And the framework I use is written down, in a book you can read before we ever talk.

Yes, start small. Most firms begin with the fixed-scope Assessment (2–3 weeks) or a single workflow before any ongoing retainer. The virtual CAIO engagement has a three-month minimum — below that there isn't enough runway to set a baseline, ship something bounded, and measure it honestly — and it scales up or down on 30 days' notice. No long lock-in, no annual contract required to get started.

My sweet spot is professional-services and manufacturing firms roughly $3M–$100M+ in revenue — big enough to have real throughput pain and a budget, small enough that one accountable operator can move the needle fast. I work in depth with legal, healthcare, accounting, engineering/AEC, consulting and agencies, and manufacturing/distribution, with playbooks for each. If you're far outside that range, I'll tell you on the first call and point you somewhere better.

One 30-minute call. No agenda except getting honest about where you are and what's realistically movable in the next 90 days. If there's a fit, the next step is a scoped plan — metric, owner, constraints, risk, timeline. If there isn't, I'll say so in the first ten minutes and point you to a better path; I've got nothing to sell into the wrong fit. Assessments usually start within two weeks; retainers can start right after the first call.

A working operating rhythm, not a slideshow. Days 1–15 (Assess): score readiness, pick one workflow, map where the work goes, name the sponsor and owner, and agree the metric before any tool gets chosen. Days 16–45 (Illuminate): build one boring, bounded pilot with an error log and a human gate before go-live. Days 46–75 (Redesign): measure honestly, run error analysis on real traces, redesign the workflow around what actually changed. Days 76–90 (Decide): scale, adjust, or stop — and set the quarterly re-validation cadence before attention drifts. You leave the first quarter with a ranked use-case portfolio, governance guardrails, one shipped pilot, and a measured decision.

Every pilot gets a kill criterion written down before it starts — the metric and threshold that, if unmet, means we stop. At day 90 the only honest outcomes are scale, adjust, or stop, and “we've already spent money on it” is not a reason to continue. My credibility is the real thing I'm selling, and the fastest way to lose it is to defend something that isn't working — so I won't. That holds even when the partner who built it is one I have a relationship with: those get held to the same kill criterion as anyone else, because the moment I protect a partner over your results, I'm not worth hiring.

You need full-time when AI becomes core to your product and you've got a growing applied-AI team and portfolio that justify a permanent executive — at which point the role pays for itself. Before that, a full-time CAIO is expensive too early and the job goes vague. The honest sequence is usually: fractional to set strategy, governance, and the first wins; full-time once there's enough real work to fill the chair. Part of doing this job well is telling you when you've outgrown needing me this way.

You own your work product and your data. Strategy documents, roadmaps, governance policies, prompts, configurations, and anything built specifically for you during the engagement are yours, and that's stated in the agreement so there's no ambiguity later. I keep only my own general methodology and frameworks — the material that's in my book anyway — never your data, your workflows, or your deliverables. One honest nuance on tooling: IP depends on what you're buying. You'll own your data and your process whether a solution is off-the-shelf or custom — but you don't own the IP of an off-the-shelf product you're licensing, and you only own the IP of a custom solution if your contract with the implementation partner assigns it to you. Getting that assignment in writing in the partner's statement of work is part of what I oversee, so you're never surprised later about what you do and don't own.

Yes. I carry professional liability (E&O) and cyber coverage, and engagement terms spell out liability and indemnification so both sides know where they stand before we start.

Compliance is built into the governance work, not bolted on after. I translate the frameworks that actually apply to you — NIST AI RMF, ISO 42001, the EU AI Act, OWASP's LLM risks — into operator-grade rules: acceptable-use policy, vendor and data-handling controls, and human-review gates for high-risk outputs. For healthcare that means designing around HIPAA from day one; for firms touching EU users or markets it means tracking the AI Act's phased obligations; for US firms it means watching a fast-moving patchwork of state laws. Where you need formal audit evidence, that becomes its own evidence-led workstream rather than a checkbox.

Lightweight enough to ship, strict enough to hold up. The minimum viable version is four things: a short acceptable-use policy, a vendor-intake rule, a data-classification rule, and a human-review gate for high-risk outputs. From there it scales to your risk profile, grounded in NIST AI RMF and ISO 42001. The goal is guardrails that let useful AI ship — not a policy museum that freezes adoption while your people quietly route around it with shadow tools.

Design, not hope. Hallucination risk drops when you point AI at bounded, well-defined work and keep a human gate on anything irreversible or client-facing — which is the whole “boring AI” thesis. Data leakage is controlled by tool selection (no training on your data, no consumer products for sensitive work) and by classification rules about what can go where. Shadow AI — people using unapproved tools — gets solved by giving them sanctioned tools that are good enough that they don't need to sneak around, plus a clear policy. The failure mode isn't usually the model; it's an ungoverned free-for-all.

You get me. There is no junior I hand you off to for decisions, though others may assist with information gathering or administrative tasks. Availability scales with tier: Foundation includes async Slack access and monthly strategy time; Growth adds Slack and email plus bi-weekly syncs; Transformation is daily access with weekly check-ins and board-level reporting. The retainer hours are real working hours, not a placeholder for a team you never meet.

Work with, not replace. I'm the AI leadership and strategy layer that makes your existing people and vendors more effective — I set direction and governance, they keep running what they run. Your CIO keeps the lights on; your MSP keeps doing tickets; I make sure the AI decisions on top of all that are owned, prioritized, and measured. Nobody loses their job because I showed up.

Start with the Results page — anonymized engagements with the actual before/after numbers: intake 45→8 minutes, clinic call volume down 80%, proposals from 3 days to 3 hours, MSP triage 2 hours to 8 minutes. I hold those to the standard I hold my book to: real, dated, and specific — not a vague “we drove change” story. As an engagement gets serious, and where a past client is willing and their confidentiality allows, I can look at connecting you directly. I don't hand out a reference list on day one — I protect my clients' privacy exactly the way I'll protect yours. If proof is what you need to take the next step, tell me what would convince you and I'll show you what I can.

I keep the roster deliberately small — this is hands-on work with real accountability, and that doesn't scale to a dozen logos. I won't take a direct competitor of an active client in the same market, and anything I learn inside your firm stays there. If a conflict ever looked possible, I'd raise it before signing, not after.

If we get on a call and it isn't a fit for what I do, I'll tell you in the first ten minutes and point you to a better path. No forms, no funnel, no pressure. I have nothing to gain if I sell into the wrong fit, so there's no quota pushing me to talk you into something. Honest conversations sometimes end in “no,” and that's fine.

I work in depth across seven verticals — healthcare clinics, law firms, accounting firms, engineering/AEC, consulting and agencies, manufacturing and distribution, and professional services broadly — each with its own playbook drawn from patterns that actually recur in that work: document-heavy review for legal, intake and prior-auth for clinics, RFP and submittal work for AEC, reconciliations and AP for accounting. Same operating principles, different failure modes. If your industry isn't on the list, the framework still applies — I just won't pretend I've seen your exact workflow fifty times if I haven't.

Remote by default — I'm based between Toronto and São Paulo and most of this work is done virtually without losing anything. When it helps (a leadership alignment session, a kickoff, a workshop that needs everyone in a room), I travel; on-site adds a facility surcharge plus travel and accommodation. Time-zone gaps are routine and easy to manage; I've worked with firms across 12+ countries.

Default to buy; build only where it's a genuine differentiator. Most mid-market wins come from configuring proven tools to a well-understood workflow, not from commissioning custom models that need a team to maintain. Part of the assessment is exactly this call — what to build, what to buy, what's open-source, what's SaaS — made before anyone gets attached to a vendor. The bar is business value, not demo sparkle.

Deliberately, I'm not loyal to any. The right tool depends on your workflow, data, risk profile, and what you already own — and the market shifts every quarter, with every platform suddenly claiming to be “agentic.” My job is to separate practical capability from sales theatre against one bar: does it move your metric. I'll name specific tools once I understand your problem, never before.

Three formats: a leadership session (where AI fits, what to avoid, which metrics matter), an operator session for the people who actually run intake/documents/approvals, and a governance session on acceptable use and guardrails. Half-day or full-day, usually remote, up to 40–50 people, and you leave with decisions, ranked use cases, templates, and next actions — not a motivational talk with better lighting. Generic prompt training is where enthusiasm goes to become a PDF; this isn't that.

The free one is a directional self-check — useful, fast, do-it-yourself. The paid Assessment is a working engagement: stakeholder interviews, evidence, a ranked use-case portfolio, risk and vendor direction, and a first-90-days roadmap your leadership and board can act on. One tells you roughly where you stand; the other gives you a plan you can fund.

No — and be wary of anyone who does. An outcome guarantee in this work usually means the bar is set so low you'd clear it anyway, or the person is overpromising to close. What I guarantee is method: a baseline before we start, honest measurement, a written kill criterion, and the discipline to stop fast if it isn't working. That protects your money better than a guarantee that's really just marketing.

Plan for it honestly: beyond my fee there's tool/subscription cost, some of your team's time, integration, and training — and after go-live, ongoing maintenance. I'd rather set that expectation on day one than have it surprise you in month six. Part of my job is making sure the running cost is worth the return before you commit, and that what we build is something your team or a partner can sustain without me.

You deal with me; I deal with the partner. I scope the work, hold them to the metric and timeline we agreed, and keep one decision log so there's no “the strategist said X, the builder said Y” gap. If the partner isn't delivering, that's mine to fix, not yours to mediate.

Expected, and easy. Most clients' needs change as they move from wanting to deploy AI to using AI as part of their operations. The retainer scales up or down on 30 days' notice, and scope changes get handled as an explicit, agreed adjustment — not a silent re-prioritization. The quarterly cadence exists precisely so the plan keeps matching reality instead of drifting.

Fair question. The operations background is why the AI work lands, but the AI work is its own record: AI-first company operations end-to-end for clients and partners, shipped intake automation, call triage, proposal and RFP systems, and ticket routing across legal, healthcare, engineering, and IT-services firms — each with measured before/after results, each governed, each with humans in the loop. The book documents the framework; the Results page documents the outcomes. I'd rather be judged on systems that shipped than on certifications that expire.

Fast and structured. We confirm the sponsor and owner, agree the first workflow and its metric, get access sorted under the right data rules, and stand up the cadence. Within the first two weeks you should feel the difference between “we're talking about AI” and “someone owns this now.”

Toronto and São Paulo, remote by default, comfortable across North American, European, and Latin American hours. Clients in 12+ countries so far.

English primarily; given the São Paulo base, Portuguese-language engagements can be accommodated with local intermediaries on my team.

Yes. Many firms have their leadership read it before we start so we're speaking the same language on day one.

Selectively, usually on the “boring AI that pays for itself” thesis and operator-grade adoption. Reach out with the audience and date.

Either. I have a standard agreement that covers scope, IP, confidentiality, and liability, and I'm fine working from your paper when your firm requires it.

Pricing is in USD but can be charged in other currencies to avoid FX where needed; terms are set out in the engagement agreement (typically staged for assessments, monthly in advance for retainers).

For the right interim full-time or long-term fractional CAIO/CTO situation, a cash-plus-equity structure can be discussed — but I don't do equity-only. Someone who won't charge cash isn't confident they'll deliver.

Read the book, grab the free Operator's Toolkit and readiness assessment, and subscribe to The AI Operator newsletter. No spam, unsubscribe anytime. When you're ready to talk, the call's there.