Amber McKenzie Advisory Fractional Data & AI Leadership · AI Governance
Amber McKenzie, Ph.D. · Fractional Chief Data & AI Officer

Executive data and AI leadership, one or two days a week.

I run the data and AI function for companies that need that level of leadership but not full time — strategy and roadmap, governance and risk, team design, and the technical judgment to evaluate what the team is actually building.

amber@amckenzieadvisory.com LinkedIn
When to call

Most companies don't go looking for this. They hit a moment.

  • A customer or prospect sends an AI questionnaire, and nobody inside can answer it.
  • A board member or investor asks who owns AI risk, and the honest answer is nobody.
  • Someone asks what AI is actually running in the company, and there's no list.
  • You know what you want built, but not whether it's worth building, or whether your data can support it.
  • You're deploying AI in hiring, healthcare, insurance or lending, where it stops being a policy question and becomes a legal one.
  • The data function has outgrown the setup that got it started, and you need a leader before you're ready for a full-time one.
How I work

Three ongoing engagements

Three-month minimum, then month to month. The more of my time you commit, the lower the effective day rate.

Advisory

About two days a month

A standing working session with your executive team, async access in between, review of your team's work, and board and investor preparation.

Fractional Lead

About a day a week

Everything in Advisory, plus ownership of a workstream: governance program stand-up, data and AI strategy and roadmap, technical due diligence on vendors and architecture, hiring and team design.

Embedded Lead

About two days a week

Functional leadership: standing in your team's ceremonies, owning the roadmap, accountable for delivery against it. Governance approvals stay with a named owner on your side.

Starting smaller

Two-week sprints, fixed fee

If you continue into Fractional Lead or Embedded Lead, either sprint is the first two weeks of that engagement. You don't pay for it twice.

Scoping Sprint

For governance questions

A preliminary inventory of the AI actually running in your company, which regimes attach to what, the three to five gaps that hurt first, and a 90-day sequence — delivered as a live readout with your leadership in the room.

Technical Discovery Sprint

For “we need someone to build this”

A written problem statement with the number that says it worked, an honest read on your data, a recommended approach — including what not to build — and a working baseline on your own data, with the code. Delivered as a live readout.

Fixed-scope work

AI Governance Readiness Assessment. Full AI system inventory, risk profiling, gap analysis against NIST AI RMF 1.0 and ISO/IEC 42001, and a phased roadmap with named owners and approval checkpoints. Four to six weeks.
Regulatory Applicability Review. EU AI Act classification plus US federal, state and sector-specific exposure: what applies to you, what it requires, and what it costs to get there. Two to three weeks.
AI & Data Due Diligence. For boards and investors: whether a company's AI claim is real, whether the architecture is sound, and where the actual risk sits.
Executive Workshops. AI governance or AI literacy for a leadership team, built around your own systems rather than generic material. Half or full day.
Track record

From the model to the boardroom

Fifteen years building NLP and machine learning systems, then eight leading the teams that ship them. Results from that work:

~$4M / yr
saved by automating review of ~98% of several million posts a week
+13 pts
net expansion revenue across two DS/ML products
−10%
customer churn on the same products
5×
addressable market ($3.2B → $16.3B) opened by multilingual models in 80+ languages
~500K hrs
saved firm-wide by automated extraction from tax documents
Architected and led implementation of an ISO/IEC 42001 AI management system, and chaired the committee that surfaced shadow AI use and vetted AI vendors.
Built model monitoring and human-in-the-loop retraining for production classifiers, and led teams of up to 16 data scientists, ML engineers and analysts.
About

Amber McKenzie, Ph.D.

Greenville, SC · working remotely

I came up as a practitioner. Fifteen years in natural language processing and machine learning — document AI at PwC, systems that removed protected health information from call transcripts, a U.S. patent on automatic call analysis, and research at Oak Ridge and Sandia National Laboratories — before eight years leading data and AI organizations, most recently as VP of Data Science and Analytics at FAMA and VP of Data Science and Machine Learning at Bombora.

Today I'm the AI governance subject matter expert behind NCompas, an AI governance platform, and I run this practice. My sector depth is in employment screening and healthcare, where bias and privacy scrutiny came with the job long before AI governance had a name.

The combination is the point. Governance people usually can't evaluate the engineering, and AI people usually start from the technology instead of the business problem. I can tell you whether your team's answer is true.

NIST AI RMF 1.0 ISO/IEC 42001 & 42005 EU AI Act US federal, state & sector rules
Ph.D. in Computer Science and M.A. in Linguistics, University of South Carolina.

If you're deploying AI and not certain what you'd say about it to a regulator, a customer or your board, let's talk.

The first conversation is short and free.

Email me LinkedIn