Agentic AI by Industry: Where the Pain Sits in Each Sector

Agent adoption changes less by sector than people assume, and it changes in a very specific way. The technical pattern is always the same: map a process, encode its rules, decide what happens with exceptions, and measure the outcome. What differs from one industry to the next is not the method but where the pain concentrates: which processes weigh on the P&L, which systems run the day, which regulation binds you, and which slice of the margin the game is played on. This page shows how that pattern looks in each vertical and links the analysis of each.

I write it because boards ask me for the opposite of what they need. They ask for “cases from my sector” expecting a different recipe, when the useful thing is to see that the mechanics repeat: if you can read them, you recognize the automatable work in any industry without waiting for someone to publish the case identical to yours.

The pattern that does not change

An agent executes one well-defined kind of work: processes with volume, with rules you can write down, and with exceptions you can typify. Whether we are talking about a factory or a distributor, the adoption sequence is identical.

First, the process: inputs, steps and expected outcome, written down. Second, the rules: what gets decided at each step and on what basis, which is almost never documented and usually lives in the heads of two people. Third, the exceptions: which cases leave the script and who resolves them, because real operations live on exceptions, and that is where pilots that only worked on the happy path die. And fourth, the outcome: how you measure that the agent does it well, with test cases and thresholds, before it gets real work. That cycle, with the business mechanics by function, is developed in AI agents in operations, and the general frame of what agents are and how they get adopted sits in the pillar on AI agents for business.

Everything else rests on that fixed pattern. So the sector question is not “does this work for what I do?” — it does — but “where do I start so the first result shows up in my P&L?”. That answer does change.

The reason this matters is that most people go sector-shopping for reassurance and end up delaying. They wait for a published case that matches their industry, their size and their systems, and while they wait, the process that would have paid off first keeps burning hours. The faster route is to learn the pattern and apply it to your own operation, because the person who knows your process best already works for you. What a sector view adds is not a different recipe. It is a shortlist of where the pain usually concentrates, so you look in the right place first.

What does change by sector

Four things move the answer from one industry to the next. The mix of processes: where the administrative volume sits that today eats the hours of people with judgment. The dominant systems: an industrial ERP, an e-commerce stack with its payment gateway, a project management system. The agent works inside them, so their integration sets the difficulty. The regulation that weighs: document traceability, data protection, quality requirements. In Europe, the AI Act applied to agents sets obligations by use. And where the margin is: in a thin-margin business the administrative cost per transaction is the lever; in a service business it is the expert time lost on low-value tasks.

Those four variables explain almost everything that separates one sector from another. What follows walks them vertical by vertical.

Manufacturing and industry

In manufacturing the pain sits in the paperwork around production, not in the machine. The production order, the sourcing of materials and its delivery notes, the documentary tracking of quality, and the coordination between plant and office generate an enormous administrative volume that today is absorbed by qualified people moving data between the ERP, email and spreadsheets. That is where an agent takes you from the order to production with no manual jumps.

It is worth separating this from real-time industrial control. Running a line, a PLC or a vision system is another discipline, with its own safety guarantees. It is not what I am talking about. The territory of process agents is the administrative layer that connects the order with the factory. I detail it in AI agents in manufacturing.

Distribution and retail

Distribution lives on a thin margin and a high volume of orders, and that combination turns the administrative cost per order into the number to attack. The work concentrates in multichannel order management, delivery incidents, inventory synchronization across systems that do not always talk to each other, and communication with the customer when something goes wrong. It is a process with many exceptions and little room to spend expensive hours on it.

That is why the pattern here is “from the order to the delivery with exceptions governed”: the agent resolves the patterned case and escalates the rest with context, leaving a trace. The full analysis, with the relationship between margin and cost per order, is in AI agents in distribution and retail.

B2B professional services

In professional services the scarce asset is expert time, and much of it goes into administrative work around the project: preparing proposals, managing contracts and their renewal, reconciling milestone billing, keeping client documentation current. It is not work that needs the senior, but today it eats them anyway. The agent frees those hours back to billable work.

The dominant system is usually a services ERP or a project manager, and regulation weighs on the side of client confidentiality and data protection. When I publish the analysis of this vertical I will link it here. Meanwhile, the mechanics by function are in operations.

Media and content

In media and content the volume sits in the editorial chain and the paperwork around it: asset management, rights and licensing control, production tracking, adapting pieces to formats and channels. It is a sector where the creative part is not automated, but the logistics that hold it up are, and they consume more time than they seem to.

The relevant regulation here is rights and data, and brand risk demands strict controls and traceability over what the agent publishes or modifies. It is a vertical I will cover in its own article. The frame of controls and escalation applies just as it does everywhere else.

Mid-market companies

The mid-market is not a sector, it is a cut by size that runs through all of the above, and it deserves a separate mention because it changes the structural decision. A company of 10 to 100 million has the process volume for agents to pay off, but rarely has a product and technology department able to configure, evaluate, integrate and maintain agents at the pace this field moves. That gap, not the sector, is usually what decides where to start and with whom.

For this company the question is not only what to automate, but who absorbs the technical cycle without forcing it to stand up a new function. I will cover it in detail in its article. The underlying logic is in the pillar and in the managed operation model.

Where to start in your sector

Whatever the industry, I start from the same place: a process with volume, tied to revenue or margin, with rules you can write and exceptions you can typify. The sector only tells me which process that is and which controls it needs. It does not change the method. Before deploying anything I require a mapped process, defined permissions and a way to measure the outcome. The risk and audit frame for all of this is in AI agent governance. Pick the point where it hurts most, not the flashiest one.

Frequently Asked Questions

Is agent adoption very different depending on the sector?

The method is the same everywhere: map the process, encode rules, typify exceptions and measure the outcome. What changes by sector is where the administrative volume concentrates, which systems dominate, which regulation applies and which slice of the margin the game is played on. Do not look for a different recipe. Look for where it hurts most in your operation.

Which sector should I start with if I operate in several?

The one with a process that has volume tied directly to revenue or margin, with writable rules and typifiable exceptions. The first deployment has to show up in the P&L. Pilots disconnected from the business are the ones that end up on the shelf.

Do agents control machines or production lines?

Not what I am talking about here. Real-time industrial control of machines, PLCs or lines is another discipline with its own safety guarantees. Process agents work in the administrative layer: orders, documentation, coordination and communication between systems and people.

What matters most when automating in a regulated sector?

Traceability and controls. In sectors with document or data requirements, every agent decision must be recorded and auditable, and sensitive actions pass through explicit permissions and escalation to a person. In Europe, the AI Act adds obligations depending on the agent’s specific use.