I’ve stepped into a newly created operating role three times at three major companies. The mandate was different in the details, but the shape of the work was remarkably consistent.
Understand how the company operates. Find where AI could make the work better. Redesign workflows. Connect systems. Put appropriate controls in place. Help people adopt new ways of working. Show that any of it produced a meaningful result.
Each time, I ran into the same structural problem.
The company had hired a person to lead AI Transformation, but the work required a capability built across the organization.
One person can lead that capability. They can’t sustainably perform every discipline inside it.
The job description quietly contains a company
An AI Transformation leader is often expected to discover opportunities across the business, understand the company’s data and systems, redesign workflows, prioritize investments, build agents and integrations, establish governance, drive adoption, measure outcomes, and support whatever reaches production.
That’s product, data, engineering, IT, operations, security, change management, and program leadership bundled into one position.
I understand why companies hire this way. The work crosses traditional boundaries, so they look for someone who can move between business problems and technical possibilities. A strong generalist can ask the right questions, connect people who don’t usually work together, design the operating model, and keep the work moving.
But working across disciplines isn’t the same as replacing them.
That person can’t also be the dedicated engineer, data analyst, systems administrator, security partner, functional expert, and support team for every workflow. When the operating model assumes they will be, the role becomes a queue of dependencies disguised as ownership.
Sometimes the constraint is financial
I don’t think leaders always misunderstand what the work requires. Sometimes they know several capabilities are needed and can fund only one new position.
That’s a real constraint. The expected outcome still has to match the investment.
If a company funds one role, provides no committed technical capacity, limits system access, and relies on teams with other priorities, it has made a legitimate operating choice. That choice can support discovery, a focused set of experiments, and perhaps a few lightweight improvements.
It probably can’t support company-wide workflow redesign at speed.
Too often, the business case assumes the output of a team at the cost of one person. The AI Transformation leader becomes product manager, analyst, engineer, change lead, governance owner, support desk, and program manager. Every technical dependency means finding someone with capacity, explaining the work, requesting access, negotiating priority, and waiting for an opening.
The work doesn’t slow because nobody owns it. It slows because the person who owns it has no reliable way to move the dependencies.
A surprising amount of AI Transformation becomes calendar management.
The minimum viable capability is cross-functional
If the goal is to change how work happens across a company, then the company is building something closer to an internal product capability than a standalone AI role.
The exact structure will depend on the organization. Not everyone needs to report to AI Transformation, and not every specialty needs a dedicated person on day one. But the work usually needs committed capacity from a few places.
Product or Operations needs to understand the problem, redesign the workflow, prioritize the portfolio, and stay accountable for the outcome.
Data needs to make the company’s information understandable and trustworthy. That includes figuring out what the business actually knows, which definitions matter, where the information lives, and whether the result can be measured credibly.
Engineering needs to turn the workflow into a reliable system. That may involve integrations, orchestration, evaluation, monitoring, and the controls required to run it in production.
IT and Security need to provide identity, permissions, provisioning, and a safe way for systems to act across company tools.
Functional experts need to explain how the work really happens. They know the exceptions, the consequences, the customer impact, and the places where an elegant design would fall apart by Tuesday.
The organization doesn’t necessarily need a large centralized AI team. It does need a real commitment from the people whose expertise and systems make the work possible.
“You can ask Engineering when you need something” isn’t committed engineering capacity.
The leader’s job is to connect the capability
The AI Transformation leader creates the most value by connecting these pieces.
They find the right problems, understand how the work currently happens, challenge processes that no longer serve the business, and decide where AI is actually useful. They help determine what the system may do on its own, where a person needs to review the work, who owns the outcome, and how the company will know whether the change helped.
They should be able to move between business and technical conversations without throwing the problem over the wall in either direction. That takes range, but the value of the role comes from judgment, orchestration, and operating design.
Making that person perform every specialized function leaves less time for the work only they are positioned to do.
At a software company, my discovery work surfaced more than 100 go-to-market tasks and hundreds of subtasks spread across teams, systems, and institutional knowledge. Finding the opportunities was only the beginning.
Turning that picture into a working AI-enabled operating model required data access, system architecture, process design, permissions, evaluation, implementation, adoption, and clear functional ownership. No amount of personal efficiency could remove those requirements.
Access is part of the operating model
Companies sometimes hire someone to understand how the organization works, then limit access to the systems that reveal how it works.
The caution makes sense. Customer records, financial systems, internal documentation, analytics, identity tools, and production systems contain sensitive information. Access shouldn’t be broad simply because a role has AI in the title.
But the alternative can’t be preventing the person from seeing the work they’ve been hired to improve.
A useful access model might include read-only permissions, sandbox environments, audited activity, progressive write access, explicit prohibited actions, human approval for important changes, and named owners for each system.
The details will vary. The principle is that governed access is infrastructure for the job, not an administrative favor granted one request at a time.
Authority matters too. An AI Transformation leader may find a valuable cross-functional opportunity but have no way to commit the teams needed to build it. They own the outcome while every dependency belongs to someone else.
That arrangement can work only if leadership actively resolves priorities and protects the capacity required to deliver.
Measure the operating model honestly
When AI work stalls, companies often look first at the person leading it. Sometimes that’s fair. The quality of the leadership matters.
The operating model deserves the same scrutiny.
Did the role have enough access to understand the work? Was engineering capacity actually committed? Could the data team support the project? Did functional leaders own implementation in their areas? Were IT and Security involved early enough to create a safe path to production? Could someone resolve competing priorities across teams?
If the answer to most of those questions is no, the company didn’t fund an AI Transformation capability.
It funded a person to negotiate for one.
Match the capability to the mandate
Not every company needs the same structure.
If the goal is to help employees learn new tools, run a portfolio of lightweight experiments, and improve a few contained workflows, one strong generalist may be enough.
If the goal is to redesign work across the company, connect systems, automate parts of important decisions, establish governance, and move production workflows toward greater autonomy, the minimum viable unit is cross-functional.
Hire one accountable leader, then give that person committed capacity from Product, Data, Engineering, IT, Security, and the functions where the work lives.
AI Transformation can have one leader. It can’t be a one-person capability unless the mandate is one-person sized.