About a month ago, I started using AI much more seriously to help me build and operate a business.
At first, I thought I was simply using AI to get more done.
I expected it to help with research, writing, planning, process design, and some of the repetitive work that can consume hours without necessarily moving the business forward.
It did all of that.
But somewhere along the way, I realized I wasn’t just using a collection of tools.
I was learning how to manage a workforce that wasn’t entirely human.
And honestly, it has been one hell of an education.
AI Is Impressive — Until It Isn’t
AI can do extraordinary work.
I’ve watched it analyze complex problems, identify issues I missed, design workflows, organize large amounts of information, and complete in minutes what might otherwise have taken me hours.
When it works well, the productivity gain is real.
But I’ve also watched AI confidently move in the wrong direction.
I’ve seen systems lose context, repeat work, make assumptions, contradict one another, and solve the wrong problem with remarkable efficiency.
Sometimes the output looked polished and convincing while still being operationally useless.
That is where the excitement around AI starts colliding with reality.
The technology may be capable, but capability alone does not produce a reliable operating environment.
The Real Problem Is Operational
I’ve spent more than 20 years leading people and operations.
That has included large teams, customer escalations, service delivery, vendor relationships, SLAs, process improvement, workflow design, and the daily challenge of delivering consistent results across complex organizations.
What surprised me was how much of that experience applies to AI.
The same operational questions still matter:
Who owns the outcome?
What can be handled independently?
What requires approval?
When should work be escalated?
How is performance measured?
Where does the source of truth live?
What happens when two contributors disagree?
And who is accountable when something goes wrong?
Those questions do not disappear simply because the work is being performed by software.
In many cases, they become even more important.
”Human in the Loop” Is Not an Operating Model
One of the most common phrases in conversations about AI is “human in the loop.”
It sounds responsible.
It suggests that a person remains involved and that automation will not be allowed to operate without oversight.
But the phrase is usually far too vague.
Which human?
At what point in the process?
What triggers the intervention?
Does the person have enough context to take over?
Can they override the system?
How quickly does ownership transfer?
What happens when the AI fails to recognize that escalation is needed?
Simply stating that a human is involved does not define how the work is controlled.
A real operating model needs clear boundaries.
It needs to specify where automation ends, where human judgment begins, and how responsibility moves between the two.
AI Needs Guardrails, Ownership, and Escalation Paths
The strongest AI implementations will not be the ones with the most tools.
They will be the ones with the clearest operating discipline.
That means defining:
- what the AI is authorized to do
- what requires human approval
- what data it may use
- where decisions are recorded
- how errors are identified
- how exceptions are escalated
- who owns the final outcome
- how performance is monitored over time
Without that structure, AI can create the illusion of progress while quietly increasing risk, inconsistency, and rework.
The output may be faster, but the operation may not be better.
That distinction matters.
AI Does Not Eliminate Management
There is a popular idea that AI will dramatically reduce the need for management.
In some areas, that may be true.
But my experience so far suggests that AI does not eliminate management.
It changes what must be managed.
Instead of only managing people, organizations will increasingly need to manage a combination of people, automated systems, AI agents, data sources, vendors, workflows, and decision rights.
That requires a different kind of leadership.
It requires people who understand operations, customer experience, risk, service delivery, process design, and accountability.
Technical expertise will matter, but technology alone will not be enough.
Someone still has to design how the work actually gets done.
Operations Leaders May Be More Important Than Ever
Much of the public conversation around AI focuses on engineers, models, platforms, and technical capabilities.
Those things are important.
But the next phase of AI adoption will depend heavily on people who know how to operationalize change.
That includes leaders from operations, customer experience, service delivery, corporate travel, procurement, finance, and other functions where technology must perform reliably in real-world conditions.
These are the people who understand exceptions.
They understand customers.
They understand what happens when a process breaks at 2:00 a.m., when a traveler is stranded, when a vendor misses an SLA, or when an automated decision creates a financial or reputational risk.
AI will need that experience.
Better Models Are Not Enough
I am still learning all of this in real time.
I am testing what works, identifying what does not, and building better structures around the technology as I go.
But one conclusion is becoming difficult to ignore:
The future of AI is not just about better models.
It is about better operating models.
The organizations that succeed will not necessarily be the ones that adopt AI first or use the most tools.
They will be the ones that establish clear roles, strong guardrails, defined escalation paths, reliable sources of truth, and real accountability.
AI may change how the work is performed.
It does not remove the need to manage the work well.