Automation Is Still Better Than AI. Let Me Show You Why.

By Matt BeranShare

AI is getting most of the attention right now. Generative AI, agentic AI, AI business automation, AI-driven automation…pick your favorite phrase and someone is probably building a pitch deck around it.

Some of that attention is deserved. AI can do useful things. It can summarize, classify, recommend, draft, detect patterns and help people move faster when it is placed in the right part of the workflow. But in the rush to make everything “AI-powered,” a lot of organizations are skipping over something boring, proven, cheaper and still wildly underused.

Automation. Just…plain, old, dependable automation.

And when we compare AI vs. automation in practical IT and Service Management environments, automation still wins more often than people want to admit.

The difference between AI and automation

People get AI and automation mixed up constantly. They overlap, but they are not the same thing.

Traditional automation follows a defined rule, trigger or workflow. Something happens, and the system performs the next step. A ticket is submitted with a certain category, so it routes to the right team. A password reset request comes in, so the system validates the user and starts the reset process. A laptop enters inventory, so the asset record updates and the onboarding workflow continues.

AI, by definition, works differently. AI uses models to interpret, predict, generate, classify or recommend based on data and patterns. It can help when the input is messy, ambiguous or hard to define with simple rules. And that is useful. It is also where things get expensive, harder to govern and even dangerous.

The difference between AI and automation matters because they solve different problems. Automation is strongest when the organization already knows what should happen. AI is strongest when the organization needs help understanding, interpreting or deciding what might need to happen. (That word “might” is doing a lot of work.)

Dependability is still the best feature

The strongest case for automation is dependability. “If this happens, we know we want this next thing to happen.” That is the entire appeal.

When an automated workflow has run ten thousand times a day for the last 576 days days without failing, you do not need a philosophical debate about its intelligence. All you need to do is protect it, monitor it and improve it.

Automation does not need to “feel confident,” to infer intent or to be prompted into giving the right answer. It just executes. That makes it incredibly valuable in IT.

IT work has plenty of tasks where ambiguity has already been removed. A new employee needs access, a device needs to be assigned, a software request needs approval, etc. For these scenarios, the goal is simple: consistency.

The Pareto principle gets thrown around a lot in AI conversations. You hear some version of “If AI can do 80% of the work, that is good enough.” Maybe that works for a draft, a summary or an internal suggestion. It does not apply cleanly to automation.

If a workflow provisions access correctly 80% of the time, that is more of a security problem than any kind of productivity miracle. If a workflow updates assets correctly 80% of the time, your inventory becomes inaccurate. Your reports and dashboards are a bunch of lies.

Automation earns trust because it does the same thing the same way every time. That matters more than novelty in most operational environments.

AI is powerful, but comes with a cost

The second major advantage of automation is cost.

AI is getting expensive in ways many organizations feel directly. Tokens cost money. Usage-based licensing costs money. Premium AI features cost money. Specialized models, integrations, oversight, training, governance, testing, monitoring and security reviews all cost money.

A lot of companies are discovering that “AI-powered” also means “metered in several creative ways.” That makes AI a business decision.

Meanwhile, a well-designed automated workflow can keep delivering value long after implementation.

Self-hosted automation, point solutions and mature workflow tools can become very attractive when the alternative is paying every time a model reads, writes, thinks, summarizes, classifies or hallucinates. This is where automation and AI often get confused in budget conversations.

AI can make automation smarter in some places. It can identify patterns, summarize context, suggest next actions and help build or improve workflows. But organizations should still ask a basic question before introducing AI into a process: Do we actually need AI here?

Sometimes the answer is yes. Often the answer is no; we need a working form, a clean approval path, a better knowledge article, a maintained asset record and one person to finally decide who owns the process. That is less glamorous, yes, but it's also cheaper.

Most organizations still have not maxed out automation

Most organizations are not sitting on a fully automated operation and wondering what frontier to conquer next. They still have manual handoffs everywhere, people copying information between systems, approvals happening in chat, onboarding checklists floating around as spreadsheets, documents, emails and tribal knowledge.

So, before asking “how is agentic AI different from traditional automation,” it may be useful to ask whether traditional automation has even been implemented properly. A lot of IT teams have barely scratched the surface.

This is especially true in Service Management. ITSM platforms usually have workflow engines, routing rules, approvals, notifications, SLAs, templates, catalog items, asset relationships and integrations. Many teams use a fraction of that capability. Then the organization gets excited about AI because it promises to leap over the boring work.

But the boring work is often the work. If the service catalog is confusing, AI will not magically make the service experience coherent. If the knowledge base is stale, AI will confidently recycle stale knowledge. If nobody owns the workflow, AI will accelerate the consequences of nobody owning the workflow.

Automation forces organizations to define what should happen. That discipline creates value by itself.

The best answer is usually AI and automation

The question should not be AI or automation in some dramatic steel cage match. A better question is: where should we use AI, and where should we use automation?

AI can help before automation begins. It can analyze ticket history, identify recurring requests, spot incident patterns, summarize common resolutions and show where manual work is consuming time. That is useful because many organizations do not have a clear map of their own operational friction.

AI can also help during automation design. It can draft workflow logic, suggest routing rules, help write knowledge content, generate test cases and propose improvements. Then automation can take over the reliable execution.

Use AI to find the repetitive work. Use humans to validate what should happen. Use automation to make it happen consistently. That is a much healthier model than handing an agent a vague objective and hoping it wanders toward governance.

Agentic AI still needs guardrails

Agentic AI can be impressive. It can plan, act, use tools and complete multi-step tasks with less human intervention. It can also introduce risk if the organization has not defined the boundaries clearly.

Traditional automation is explicit. The trigger is known. The conditions are known. The output is known. The audit trail is usually easier to understand.

Agentic AI adds interpretation. That may help with complex work, but it also creates questions. What actions can the agent take? Which systems can it access? When does it need approval? How do we test it? How do we audit the decision path? Who owns the outcome when it does the wrong thing?

If the answer is, “the vendor says it is safe,” have fun with your future incident review.

Agentic AI needs limits, supervision, clean data, escalation paths and process ownership. In other words, it needs many of the same things automation needed all along.

Automation still deserves respect. AI is exciting. Automation is dependable. In IT, dependable wins a lot.

The opportunity is not to reject AI. The opportunity is to stop treating automation as yesterday’s technology just because it does not give keynote demos with dramatic music.

Automation remains one of the most practical, affordable and underused ways to improve IT operations. It reduces repetitive work. It enforces consistency. It improves speed. It supports governance. It makes service delivery easier to understand and easier to trust.

AI has a role. A meaningful one. But when the process is clear, the decision is known and the next step should happen the same way every time, automation is still the better tool.

And once you look closely at many AI solutions, you may find that many of them are just automated automations, anyway.

10 Customer Service Tips I Learned During the World Cup

Susan Smith's team provided technical support for first responders and others at FIFA World Cup events in Dallas this summer. It took months of planning, new equipment and new software. As National Customer Service Week approaches, her blog is a good reminder that world-class service isn't built in one big moment, but around ordinary ones, repeated daily.

Can We Really Reduce "Work About the Work?

How To Get Your AI Project to Deliver Measurable ROI

Bridging the Gap: 5 Tips for Cross-Functional Collaboration That Enables AI Transformation

Ask ten executives who owns AI at their company, and you’ll get ten different answers. IT says it’s not their call. Legal gets blamed for slowing everything down. HR figures it’s someone else’s department. Meanwhile, teams are buying tools nobody signed off on, duplicating work and hoping it all sorts itself out. Sound familiar?

Lisa Duerre spent the last year studying why that happens. As part of an applied research project for her leadership consulting collective, RLD Group, she studied where AI adoption breaks down inside organizations, and where it works. The findings from RLD Group’s research helped inform a collaboration on the CONVERSATIONS WORTH HAVING®: The Human Accelerator for Artificial Intelligence Quick Start Guide, which is available as a digital download.

Duerre views organizations through what she calls an I–WE–US leadership framework, defined like this:

I: individual judgment and accountability
WE: workflows and cross-functional coordination
US: governance, decision rights and organizational measures

“All three levels are contributing to the breakdown or the alignment, whether people realize it or not,” Duerre says. “AI is amplifying whatever’s already true in your system. The teams that were disconnected before AI showed up are more disconnected now. The ones that talked to each other are moving faster, together.”

If your company is ready to collaborate better with AI tools, Duerre shared the following tips. Take a look.

Form a cross-functional AI committee

Duerre’s background is in HR, and she says most HR leaders assume AI ownership belongs to IT. It doesn’t, at least not exclusively.

“Ownership needs to sit at the system level,” Duerre says. “Each function carries a piece of it, based on what they do, how well they understand that part of the business and how their work depends on everyone else’s. AI is flattening how we work. You can’t just keep it in your own business unit anymore. You have to look all around you.”

For starters, she suggests building a cross-functional AI committee instead of having one department make all the AI decisions. Legal, IT, cybersecurity and HR should be on the committee, Duerre says.

“If you have a C in front of your title, you should be on that committee,” Duerre says. “That’s how I look at it, because it’s a system-level solution.”

During these meetings, Duerre says you’ll find out that some departments are racing ahead with AI and others are holding back.

“Both sides need to name the trade-offs aloud,” Duerre says. “With teams moving too cautiously, you have to talk about the opportunity cost of falling behind. With teams sprinting ahead, you have to ask them what happens if they don’t bring everyone else along with them.”

Figure out how to use AI strategically

Most companies spent the past two years telling employees to use AI with anything, without much strategy behind it. Duerre says that’s starting to catch up with organizations as finance teams scrutinize the cost.

Her rule of thumb: if you can’t articulate the goal and how you’ll measure success, don’t roll it out yet.

“Teams that use AI well have a strategy behind it,” Duerre says. “They’ve kicked the tires on what they’re trying to solve it for. You need to ask yourself, ‘Which business outcome are we trying to improve, and what must be aligned for AI to create measurable value?’”

Here are a few examples of how to use AI strategically:

A company could select a workflow that regularly creates delays, redesign it with AI and test the new approach. Then, measure whether it improves time, cost, quality or capacity.
Use AI to support early sales outreach and qualification across markets and languages. AI can help a business reach and assess more potential opportunities, while people remain responsible for understanding the customer and building trust.
Flag patterns in customer complaints across multiple channels with AI, so leadership can see recurring problems before it shows up in satisfaction scores.

Check-in regularly during an AI rollout

Duerre recommends a minimum weekly check-in during any AI rollout, sometimes daily depending on complexity. But the format matters more than the frequency. Status updates don’t cut it.

“Ask, ‘What are we learning and what are we surprised by?’ That’s a question that helps you with your check-ins, versus, ‘It’s in three products now and we’ve tested six,’” Duerre says. “That doesn't help, because you’re having these meetings to figure out what’s working and why. If you ask more strategic questions, you can move even faster.”

Publish AI guardrails

Employees who don’t know what’s allowed with AI will either freeze or go around the system entirely, Duerre says. She recommends publishing clear, specific guardrails on what’s okay and what’s not. Come up with some real examples, and pair them with an intake process that doesn’t require writing a thesis to get an approval for using it.

"The approval path should be lightweight, not bureaucratic,” Duerre says. “Something like, ‘If you’re going to use AI, here’s the path. And if it needs approval, here’s three or four quick questions for you to answer.’”

Take employee anxiety about AI seriously

“AI is just a tool” is a phrase Duerre hears at nearly every conference she attends, but she doesn’t buy it.

“Saying it’s a tool is underselling what’s happening at companies right now,” Duerre says. “AI is changing how we work. It’s changing how we lead teams.”

Duerre wants leaders to remember that a lot of employees are fearful of AI. Pew Research Center found 52% of U.S. workers are worried about the future impact of AI in the workplace.

Employees who feel AI is being “done to them,” instead of built alongside them are especially anxious, she says.

“Leaders need to recognize that anxiety is contagious,” Duerre says. “As a leader, this is your opportunity to show up as the safe, steady person who is showing what you’re learning with AI. And don’t be afraid to show how you’ve failed using AI, too.”

Duerre asks every executive she works with: “Who am I with AI?” and encourages them to pass this mindset question along to their employees, too.

“AI is now your teammate,” Duerre says. “Phrasing it as, ‘who am I with AI?’ is different than, ‘what’s going to happen to me with AI?’ You really want your team to feel empowered with AI and show them how it can help accelerate their career.”

Put these ideas into action

Rewiring your organization for AI requires more than the right tools. It takes shared language, practical frameworks, and a willingness to keep learning. Here are a few resources to help you take the next step.

Enterprise AI Playbook: Practical frameworks and executive discussion questions to help IT, HR, and business leaders align around AI that delivers measurable value.
Work-First AI Use Case Assessment: Identify the workflows where AI can have the greatest impact before you invest in new tools.
The REWIRED Brief: Get weekly insights, real-world case studies, and practical advice on leading AI transformation.

Informa helps businesses and professionals in hundreds of ways.