Services

Scoped, priced, and finished.

Four core engagements, plus a long list of specific projects underneath them. Every one is quoted as a fixed fee against a written scope before anything begins, so you know the cost and the finish line before you commit.

Why this is worth paying for

The research says the same thing twice.

Where AI is deployed into a workflow that was designed for it, the measured gains are large. Where it's bought as a pilot and bolted on, the return is close to zero. Same technology, both times.

66%

Average productivity gain across three controlled studies of business users given generative AI for their real work.

Nielsen Norman Group meta-analysis, 2023 — nngroup.com

14%

More issues resolved per hour by 5,179 customer support agents given an AI assistant — rising to 34% for the least experienced staff.

Brynjolfsson, Li & Raymond, NBER Working Paper 31161 — nber.org

95%

Of enterprise generative AI pilots produce no measurable return, against $30–40 billion in spending. MIT calls the cause organizational, not technical.

MIT NANDA, The GenAI Divide: State of AI in Business, 2025 — report PDF

What that looks like in your numbers

A rough way to size a single workflow: one person, one recurring process, five hours a week. At a fully-loaded cost of $60 an hour, that one workflow is roughly $15,000 a year of capacity. Most businesses I look at have three or four of them, and the build is a one-time cost against a recurring return.

Which is exactly why I scope first

That arithmetic only holds if the process is genuinely repetitive, the data is usable, and the output doesn't need a human checking every line. Working out whether all three are true, before anyone spends money, is what the Readiness Sprint is for.

A note on these figures, because you should be suspicious of consultants quoting statistics: the 66% and 14% come from controlled studies of specific tasks, not whole businesses, and your results will depend on the process. The $15,000 above is illustrative arithmetic, not a projection — it's there to show you the shape of the maths, and I'd rather run it against your real numbers on a call than have you take mine.

Core engagements

Start here.

01 · typically 2 weeks

AI Readiness Sprint

The front door. Two weeks to answer the only question that matters: where does AI actually pay off in this business, and what should you do first? I map how the work really gets done — the spreadsheets, the copy-paste, the person who "just knows" — then rank the opportunities against effort and risk.

Workflow map of how the work actually moves today
Opportunity map, ranked by value against effort
A recommended first build, scoped and costed
Tool and model recommendations with real monthly run costs
Data readiness assessment — what's usable, what needs work
Risk, privacy, and acceptable-use notes
90-day roadmap you could hand to somebody else

Best for Anyone who keeps having AI conversations that don't end in a decision.

02 · typically 3–8 weeks

Build — agents & automations

I design and ship the thing. A working agent or automated workflow, running against your real data, in your stack, on a schedule or on demand. This is the part most AI consultants hand off to somebody else. I don't have a somebody else.

A working system, deployed and running
Evaluation set, so you can see when it's right and when it isn't
Monitoring, logging, and a sane failure mode when the model is wrong
Cost controls and usage limits, set before you get a surprise bill
Plain-English documentation and a live handoff session
All source, prompts, and configuration — owned by you
30 days of support after handoff, included

Best for A specific, repetitive, expensive process you already know is worth fixing.

03 · half day or full day

Team enablement

Half your AI value is locked up in people not knowing what's possible, and a quiet amount of your risk is in people not knowing what's unsafe. A working session built on your actual workflows — your documents, your emails, your process — not a generic prompting webinar.

Half-day or full-day session, in person or remote
Prompt and workflow patterns built around your real tasks
A written playbook the team keeps and can hand to new hires
A one-page acceptable-use policy sized to your business
Tool recommendations by role, not one tool for everybody
Follow-up session 30 days later to fix what didn't stick

Best for Teams already using AI informally, inconsistently, and occasionally unsafely.

04 · monthly retainer

Fractional AI lead

Senior judgment on retainer, for companies that need an AI person but can't justify hiring one. I sit in the decisions, review the vendors, keep something shipping, and keep the team current as the tools change underneath them.

Standing working sessions plus async access between them
Vendor and tool evaluation before you sign anything
A build in flight at all times, on the higher tiers
Ongoing enablement as models and tools change
Quarterly written review of AI spend against value delivered

Best for Companies where AI decisions keep coming up and nobody owns them.

Ongoing support

Three levels of retainer.

Month to month after an initial three-month term — long enough to be worth doing, short enough that you're never locked into something that isn't working.

Advisor

A standing brain on call.

Two 60-minute working sessions a month
Async questions answered within one business day
Vendor and tool decisions reviewed before you commit
A short monthly note on what changed and what it means for you

Good fit Solo operators and small teams who mostly need someone to check their thinking.

Most common Operator

Advice, plus something always shipping.

Everything in Advisor
Weekly working session
One build in flight at all times — automations, agents, integrations
Quarterly team enablement session
Ongoing maintenance of anything I've built for you

Good fit Companies with a real backlog of things AI should be doing and nobody to do them.

Embedded

Roughly two days a week. Two clients at a time, maximum.

Everything in Operator
In your standups, your planning, and your vendor calls
Multiple builds running in parallel
Hiring support — scoping the role, screening for it, onboarding them
Board and investor-ready reporting on AI initiatives

Good fit A funded startup or established business making AI a real line item.

Everything here is quoted, not listed. Sprints, builds, enablement sessions, and retainers vary too much by scope and by business to put a number on honestly before I understand yours. So we talk first, and you get a fixed fee against a written scope before any work begins — the number is settled before you commit to anything, and it doesn't move unless the scope does.

Embedded availability is limited by design. That role only works when it gets real attention, and there's only one of me.

Project types

Whatever is causing the most pain.

Engagements get shaped around your situation rather than picked off a list, but these are the shapes the work usually takes.

Custom agent build

An agent that researches, monitors, drafts, or decides on a schedule — with tools, memory, and a written record of why it did what it did.

Document and email automation

Intake, extraction, summarization, drafting, and routing for the paper that moves through your business. Usually the fastest payback available.

Internal knowledge assistant

Ask your own documents a question and get a cited answer. Built on your files, scoped to who's allowed to see what, honest about when it doesn't know.

Lead generation and prospect research

Automated research that produces a qualified, deduplicated list on a schedule, with the angle for each one written out. Fit-qualified, not intent-qualified — I'll explain the difference.

Customer support triage

Classification, routing, draft responses, and escalation rules, with a human check where being wrong would be expensive.

Monitoring and alerting agents

Something that watches a market, a competitor, a regulation, or a data feed and tells you only when it matters.

AI features in your product

Where AI genuinely improves the product versus where it's a checkbox for the pitch deck. Design, prototype, and a build plan your engineers can take from there.

Reporting and dashboard automation

The recurring report that eats a day every month, assembled and written for you — with the numbers pulled from the source instead of retyped.

Vendor and tool selection

Independent evaluation of the AI tools you're considering, with no reseller relationships and no vendor's thumb on the scale. Including when the answer is "none of them yet."

Prompt library and workflow standardization

Turn what your best person does with AI into something the whole team does the same way, stored somewhere findable.

AI policy and governance

A practical acceptable-use policy, data handling rules, and an approval path for new tools — sized to your business, not copied from an enterprise template.

Data readiness

Most AI projects fail on data, not models. An assessment of what you have, what shape it's in, and the smallest amount of cleanup that unlocks the use case.

Cost and model optimization

Already running something that works but costs too much? Model selection, caching, and prompt work that usually takes a meaningful bite out of the bill.

Stuck project review

You built something and it's been 70% right for a month. An outside read on whether it's a prompt problem, a data problem, an evaluation problem, or the wrong idea.

AI due diligence

A straight read on whether a company's AI claims are real — for investors, acquirers, or anyone about to sign a large contract.

Workshops and training

Role-specific sessions for teams, leadership briefings, or a one-off session for a board or offsite.

Not sure which one you need?

That's normal, and it's what the first call is for. Describe the problem in plain language and I'll tell you which of these fits — or that none of them do.

Start a conversation Who you'd be working with →