What does a Head of AI Enablement do? I read ten job postings to find out

One title, two jobs

J.D. Power says it in a line: “This is a builder role, not a steward role.” Deepgram calls its seat build-first and not a training role. Shine wants someone who builds tools and automations for each team and owns them end to end, with nobody to hand them to. Unity Advisory wants a lead who builds smaller tools personally and calls in engineering for production work. Spectris expects hands-on building on low-code platforms.

The other half reads differently. Synchrony wants an enterprise change leader and asks for 10 or more years in fields like communications, marketing, learning and change management. Flutter wants an architect of adoption and change who designs training. HPE leans on change management methods and prefers Prosci certification. Johnson Controls calls its seat a capability-building and standards role rather than a delivery role. Harvard’s lead is the service owner for ChatGPT, Anthropic and similar tools, which means access, licenses and billing.

By my sort that is five and five. Builders: Deepgram, J.D. Power, Shine, Unity and Spectris. Change, standards or service leads: Synchrony, Flutter, Johnson Controls, HPE and Harvard. Unity and Spectris straddle the line. I put them with the builders because both list hands-on building as something the candidate must be able to do.

If I were hiring, I would pick the builder who likes sitting with a team. A lead who has never built anything cannot tell a use case that will work from one that only sounds good in a meeting. Second Talent argues the role is non-technical in delivery and technical in judgment. I agree the lead should not build the production integration. I doubt you get the judgment without having built something small. Unity’s version is the one I would copy: build the small thing yourself, and give the production work to engineering.

The builder version has a trap, and Johnson Controls names it outright. Its posting says the seat will be pulled toward execution constantly and has to resist. A lead who becomes the company’s most overworked automation engineer is not enabling anyone. Spectris draws the line in writing: anything that needs production-grade architecture goes to its Enterprise Architect.

Five things the job can own

I sorted the duties in all ten postings into five areas. The sort is mine, and so is every mark in the grid. A full circle means the posting lists the duty plainly. A half circle means it names the area but only in part, for example as support for someone else. An empty circle means I found nothing.

Read it by rows. Measurement is full in all ten. Adoption and skills is at least half in all ten. After that it comes apart. Only four postings (Unity, Shine, Spectris and J.D. Power) give the lead the job of choosing what to work on, and Flutter and Deepgram do not mention it.

That is the row I care about most. Spectris asks its lead to keep one list of opportunities with a business owner, expected value, risk and status. Second Talent says the same job includes running an intake and being willing to reject most requests. Without that list, the training and the champions go to whatever tool someone happened to buy. Flutter’s lead is handed a product suite (Flutter Assistant, AI Hub, Workato) and asked to drive adoption of it, which may be why that cell is empty. I would still want the right to say no.

Standards are the quiet row. Johnson Controls puts them at the center: which tools are approved, what qualifies someone for a license, and how licenses are reclaimed and reviewed. It is the only posting that writes reclaiming into the job. I think that is right. A license nobody has touched in 30 days is a decision somebody avoided making. The same posting asks its lead to hold a standard, including by saying no to exceptions that would erode it, without owning the delivery pipeline.

Who it reports to

Six of the ten postings state a boss. Two involve a Chief AI Officer. The rest name a VP of IT, a client services director, a chief product and technology officer and a head of business performance. Flutter, Deepgram, Johnson Controls and HPE say nothing in the text I read.

Betts Recruiting says the seat typically reports to a Chief Revenue Officer, Chief Operating Officer or Chief Product and Technology Officer, but it is writing about go-to-market teams. Second Talent lists a transformation office, IT, operations or a Chief AI Officer. Nobody agrees, so here is my answer.

I would not put the job under IT unless the job is mostly tools. Harvard’s is, and its line into IT fits, because the lead runs a service. Spectris is different. Its own posting names Sales, Human Resources, Service and Operations as the places the work happens, then puts the lead under the VP of IT, so the boss sits outside every place the work is supposed to change.

The two lines I find most interesting are Shine’s and Unity’s. Shine puts a small team in the CEO’s office under the head of business performance, and the lead works with that boss to prioritize where AI adoption will have the most business impact. Unity splits the line between its Chief People Officer and its Chief AI Officer on purpose, because it sees the job as much about how people work as about the technology. Both put the lead next to the people who decide where work changes. That is where I would want to sit.

What it pays, where anyone says

Five of the ten postings state pay. Harvard gives a salary grade (058) and points to a table I did not open. Flutter, Unity Advisory, Shine and Johnson Controls say nothing. That is how thin the evidence is, and it is why I will not quote you a market rate.

Titles will not sort this out. Only four of the ten say “Head of”: Synchrony, Deepgram, Shine and J.D. Power. The rest say Lead, Leader, or Adoption & Enablement Lead, and Second Talent adds AI Integration Manager, AI Adoption Manager and AI Workflow Lead for the same work. The spread inside the data is more than threefold, from the bottom of Spectris’s band to the top of Synchrony’s.

Second Talent, quoting ZipRecruiter’s July 2026 data, reports a 25th to 75th percentile of about $51,000 to $141,500 for AI enablement roles, and explains the width: the same search catches hourly support work and senior leadership. Its warning is the useful part. The title covers everything from a coordinator running training sessions to an executive who owns enterprise-wide adoption, and those are not the same hire.

Measure the work, not the login

Every posting asks for measurement. They disagree about what to measure. HPE lists training completion and user sentiment. Synchrony lists participation and sentiment. Flutter lists NPS and time-to-competency, and asks candidates for a record of hitting adoption targets, with over 80% user adoption as its example. Those are the numbers of a program that runs well. They do not say the work changed.

Johnson Controls, Deepgram and J.D. Power go the other way. Johnson Controls says its draft KPIs are meant to measure enablement outcomes rather than program activity, and wants a lead who can tell the difference between activity and adoption. Deepgram wants productivity and quality, not activity. J.D. Power says it measures the hire on outcomes, not activities.

I side with them, plus one rule of my own: no outcome number without a baseline. Second Talent puts it bluntly. Without a measurement of how long a task took beforehand, every claimed gain is an anecdote. Here is how I would sort the usual measures.

If you let me keep only one number, it is this: for each workflow you chose, the share of the team still working the new way after 30 days, next to time per task against the baseline. That pair is my pick, not any posting’s. The Johnson Controls list is marked in its own posting as a draft for discussion, to be finalized against what the company can instrument, so treat it as one company’s first try. For how to take a baseline, see how to measure AI impact and the entry for baseline.

A first 90 days I would defend

No posting I read publishes a 90-day plan. Unity Advisory lists what year one should look like, and Second Talent lists methods. The plan below is my synthesis of what those sources name, with my opinions added. It is not any source’s plan.

Here is what I would not do in month one: launch company-wide training. Second Talent says a generalist running a company-wide program produces broad, shallow adoption, and that the strongest results come from someone who knows how one function really works. I would not buy another tool either. Betts Recruiting lists different teams settling on different tools for similar tasks as one symptom of a missing leader.

Three jobs it gets mistaken for

The training job. Deepgram says its seat is not a training role. Unity asks its lead to work with its Performance & Development Lead so AI skills end up inside onboarding and development, and not in a separate program. Training is part of the work and a poor definition of it.

The IT job. Johnson Controls calls its seat a capability-building and standards role rather than a delivery role. Spectris sends production-grade work to its Enterprise Architect. Harvard is the exception, and its lead sits in IT because the lead runs a service.

The forward-deployed engineer. AI Talent on Demand says the role complements forward-deployed engineers and does not replace them (see what is a forward-deployed engineer). Shine’s posting reads to me like an internal version: sit with a team, learn how it works, build for it. That is a good job. It does not scale by hand. One person across Finance, HR, Sales and Customer Support is one very busy person, which is why I also want the lead to share what works from one team to the next.

Two smaller confusions. A Center of Excellence is a group, and the lead is a person who may run one. Unity’s lead builds its CoE from scratch, Flutter’s works alongside one that already exists, and at J.D. Power the Hub coordinates while champions deliver inside their functions. And an AI product manager owns a product built for customers, a line Second Talent draws in its FAQ.

What I could not find

I found no 90-day plan in any posting, no pay range at five of the ten, and no reporting line at four. I also cannot tell you whether any of these hires worked out. A posting is what a company hopes for. It says nothing about the person who took the job.

Second Talent expects some consolidation. Where adoption matures, it says, enablement folds into operations, learning or a permanent AI function, and the standalone title may not last everywhere. I think that is right. It is a reason to hire for the work in the grid above, and to stay relaxed about the title on the door.

Sources

1. The Betts Team. “What is a Head of AI Enablement?” Betts Recruiting, July 20, 2026.
2. “AI Enablement Lead: Key Skills & Responsibilities in 2026.” Second Talent, August 14, 2026.
3. Melissa Bridge. “What Does an AI Enablement Manager Actually Do?” AI Talent on Demand, July 22, 2026.
4. “VP, Head of AI Enablement (job posting, Job ID 2602496).” Synchrony, posted September 25, 2026.
5. “AI & Automation Enablement Lead (job posting).” Flutter Group, accessed October 15, 2026.
6. “AI Enablement Lead (job posting).” Unity Advisory, via Ashby, accessed October 14, 2026.
7. “AI Enablement Lead (job posting 002871SR; expired).” Harvard University Information Technology, accessed October 15, 2026.
8. “Head of AI Enablement Engineering (job posting).” Deepgram, via Ashby, accessed October 15, 2026.
9. “Head of AI Enablement, CEO Office (job posting).” Shine, accessed October 15, 2026.
10. “AI Enablement Lead (job posting, Particle Measuring Systems).” Spectris, via Workday, published August 24, 2026.
11. “AI Enablement Leader (job posting).” Johnson Controls, accessed October 15, 2026.
12. “AI Adoption & Enablement Lead (job posting).” Hewlett Packard Enterprise, via Workday, published August 11, 2026.
13. “Head of AI Enablement (job posting R-100190; expired).” J.D. Power / Autodata, via London Tech Jobs, posted June 30, 2026.

Plain-text copy: /blog/what-does-a-head-of-ai-enablement-do.md