Artificial intelligence has quickly become one of the biggest priorities in business. Companies are investing in copilots, chatbots, AI agents, automated workflows, and a growing list of tools that promise to make work faster, reduce costs, and improve the customer experience.
There is nothing wrong with that. AI has enormous potential to improve how organizations operate and how customers interact with them.
The problem is that many companies seem to be starting in the wrong place.
Instead of first asking how work should be designed, where customers are struggling, or what is getting in the way of employees delivering a good experience, the conversation often starts with a much simpler question:
Where can we use AI?
That can quickly turn into a solution looking for a problem.
A better starting point is to ask where the organization is already losing value. Where are customers getting frustrated? Where are employees spending time on work that should not be necessary? Where are handoffs breaking down? Where is information getting lost? Where are policies, processes, or systems creating problems employees have to fix manually?
Once those questions are understood, AI may absolutely be part of the answer.
But it should not automatically be the first answer.
AI Adoption Is Not the Same as AI Value
A company can deploy dozens of AI tools and still fail to improve the customer experience, employee experience, or economics of the business.
Research on AI in organizations has increasingly shown that technology alone does not create transformation. Value comes from how technology works with the processes, knowledge, capabilities, employees, and decisions surrounding it.
Deploying the technology is often the easy part. Redesigning how the organization works is harder.
That creates what I think of as the AI Value Gap: the distance between implementing AI and actually producing a better outcome.
An AI system can work exactly as designed while the organization around it continues operating with confusing policies, disconnected systems, unnecessary approvals, fragmented information, and poorly designed workflows.
When that happens, the company has adopted AI without necessarily improving the experience.
Before You Automate the Workflow, Understand the Workflow
Imagine a customer contacts a company because there is a problem with their account.
The employee opens one system to find the customer, another to locate the transaction, and another to review account history. They search a separate knowledge base for the policy, contact another department because the information is unclear, enter notes into another system, and wait for an approval before giving the customer an answer.
A company looking at this process might immediately see an opportunity for AI. Maybe AI could retrieve the information, summarize the account, recommend the policy, or prepare the response.
All of that could be useful.
But before automating those steps, I would want to know why so many of them exist in the first place.
Why are the systems disconnected? Why is the policy unclear? Why does another department need to become involved? Why does the employee need approval? Why isn’t the information needed to resolve the customer’s problem already available?
There is a major difference between automating a workflow and improving a workflow.
If a process contains seven unnecessary steps, using AI to perform those seven
steps faster may improve efficiency, but it does not necessarily solve the problem.
The better answer might be to eliminate several steps, integrate systems, simplify a policy, or give employees greater decision authority before AI is added.
That part of transformation is less exciting than announcing a new AI agent, but it may be where much of the real value is hiding.
Garbage Process In. Garbage Experience Out.
Technology professionals have used the phrase garbage in, garbage out for years to describe what happens when poor-quality information enters a system.
The same principle applies to processes.
Garbage process in. Garbage experience out.
If the information feeding an AI system is outdated, AI can provide an outdated answer more efficiently. If policies conflict, AI may surface those conflicts more frequently. If departments disagree about ownership, AI cannot automatically solve the organizational problem underneath it.
AI can make a strong process more scalable. It can also make a weak process more scalable.
Imagine an AI customer service assistant repeatedly gives customers the wrong information about a return policy. The immediate reaction may be to blame the AI.
But trace the problem backward.
The AI retrieved its answer from the knowledge base. The knowledge base contained an outdated policy. The policy had changed, but the team responsible for the change assumed another department would update the information.
Customer service employees already knew the policy was wrong because customers had been complaining about it. During live conversations, those employees had been manually correcting the answer.
Then the company automated more of those conversations.
Now the workaround is gone, and the underlying break becomes visible.
What looked like an AI failure was actually an organizational failure that AI exposed.
The technology did not create the problem. It removed the person who had been quietly compensating for it.
Before You Automate the Employee, Understand What They Are Fixing
This is why I think organizations should be careful when AI business cases begin primarily with headcount reduction.
Employees carry a significant amount of operational knowledge that rarely appears in formal process documentation. They know which policies cause problems, which systems cannot always be trusted, which departments need to be contacted when something unusual happens, and when following the official process will actually create a worse outcome for the customer.
Over time, employees develop workarounds that keep the experience functioning despite weaknesses elsewhere in the organization.
That does not mean those workarounds should remain. It means they should be studied.
An employee manually moving information between systems may look inefficient. But before automating that activity, the organization should understand why the employee has to do it.
The employee may not be creating the inefficiency. They may be the person preventing an inefficient system from failing the customer.
Research also suggests that AI can create significant value when it augments people rather than simply replacing them. A large field study published in The Quarterly Journal of Economics examined more than 5,000 customer support agents using a generative AI assistant and found meaningful productivity improvements, with some of the largest gains among less experienced workers.
The value was not created because the employees disappeared. It was created because the technology helped them perform better.
That creates a different question for organizations. If AI gives an employee back time by helping them retrieve information, summarize documentation, or perform repetitive administrative work, should that capacity automatically be treated as an opportunity to reduce staffing?
Or could employees use some of that time to investigate recurring problems, talk with customers, improve processes, identify patterns in complaints, or handle complicated situations that require human judgment?
Efficiency is valuable.
But it is not the only type of value.
Customers Don’t Need Humans for Everything. But They Don’t Want Bots for Everything Either.
There are many customer interactions where automation makes perfect sense.
If I need to reset a password, check an order status, update an address, or find basic information, I usually do not need another person. I need a fast and accurate answer.
AI can make those interactions dramatically better.
The problem starts when organizations assume that because customers prefer automation in some situations, they prefer it in every situation.
They do not.
Customer needs change depending on the complexity of the problem, the amount of money involved, the level of risk, emotional stakes, urgency, and individual circumstances.
Research in service and consumer behavior supports this more nuanced view. Customers can be more comfortable with AI when interactions involve routine or objective information, while human involvement becomes more valuable when a situation requires judgment, empathy, interpretation, or reassurance.
That distinction matters.
Checking an account balance and disputing a significant financial transaction may both technically be customer service, but they are very different experiences.
One primarily requires information.
The other may require investigation, trust, judgment, and reassurance.
The better question is not whether customers prefer AI or humans.
It is:
What does this customer need in this situation?
The AI Problem May Not Be an AI Problem
This is where AI begins to expose a much larger experience problem.
Companies organize themselves into departments, systems, products, policies, channels, and technologies. Customers do not experience those boundaries in the same way.
They experience one company.
A customer may begin with a chatbot, move to a customer service representative, require information from another department, and eventually interact with a completely different system. Internally, those may be separate areas of the business. To the customer, it is one continuous experience.
When those internal boundaries become visible, they become part of the customer experience.
We see this clearly when a customer spends several minutes explaining a problem to an automated system, eventually reaches a person, and immediately hears:
“Can you tell me what happened?”
The customer already did.
The company failed to maintain continuity.
The problem is not simply that automation failed. The information failed to move with the customer.
This Is Where Experience Continuity Matters
This is one of the ideas behind the Experience Continuity Chain™ at Signal & Journey.
Customer experiences are rarely created by one department or one moment. What happens at the point of interaction is often the downstream result of decisions and conditions elsewhere in the organization.
Policies affect what employees can do. Processes determine how work moves. Technology affects what information is available. Training influences employee readiness. Decision authority determines whether an employee can resolve a problem. Knowledge management affects whether the information being provided is accurate. Customer expectations and circumstances influence how all of those things are ultimately experienced.
By the time a problem becomes visible to the customer, several parts of the organization may already have influenced the outcome.
That is why it can be dangerous to fix an experience problem only where it appears.
A poor customer service interaction becomes a training issue, an abandoned digital journey becomes a website issue, and an incorrect AI response becomes a technology issue.
Sometimes those diagnoses are right and the break happened much earlier.
The more useful questions are:
Where did continuity break?
What caused the break?
Follow the Signal Backward
A customer complaint is a signal. So is an increase in repeat contacts, abandonment, escalations, employee frustration, or manual workarounds.
Rather than treating each signal as an isolated problem, organizations can trace it backward through the experience.
At Signal & Journey, we think about that process as:
Signal → Break → Cause → Impact → Intervention → Metric
Start with what became visible. Determine where continuity broke. Trace that break back to the condition that caused it. Understand what impact the break created, intervene at the point most likely to change the outcome, and then measure whether the experience actually improved.
This becomes especially important with AI because an AI failure can easily look like a technology problem even when the real cause is outdated information, unclear ownership, a broken policy, poor governance, or a flawed workflow.
The right intervention depends on finding the actual break.
AI Should Remove Friction, Not Humanity
None of this is an argument against AI.
Organizations should use AI aggressively where it genuinely improves the experience. Employees should not spend hours searching for information that technology can retrieve in seconds. Customers should not sit on hold for simple questions that can be answered immediately. Companies should not manually analyze thousands of customer comments when AI can help identify patterns.
Those are real opportunities.
But automation should not become the objective by itself.
Sometimes the right intervention is AI. Sometimes it is process redesign, better information, clearer ownership, improved employee authority, or stronger coordination between teams. And sometimes the human interaction is not the inefficiency, it is the value.
Follow the Break, Not the Buzz
There is enormous pressure right now for companies to prove they are doing something with AI. That pressure is understandable. The technology is changing quickly, competitors are investing heavily, and no organization wants to fall behind.
But urgency should not replace diagnosis.
Before asking where AI belongs, organizations should understand where the experience is already breaking. Listen to customers. Listen to employees. Pay attention to repeated escalations, workarounds, abandoned journeys, lost information, and places where employees are forced to choose between following the process and doing what actually makes sense.
Those are signals.
Follow them backward.
Sometimes the answer will be AI. Sometimes it will not.
The companies that ultimately create the most value from AI may not be the ones that automate the most. They may be the organizations that are more disciplined about understanding where technology improves the experience, where processes need to be redesigned, and where human knowledge and judgment should be strengthened rather than removed.
AI is becoming an increasingly powerful part of the experience system.
But it is still only one part.
Because you cannot automate your way out of a broken experience.
And in the rush to adopt AI, companies should be careful not to automate away the people, knowledge, judgment, and human connection that were creating value all along.
Research Behind This Perspective
Brynjolfsson, E., Li, D., & Raymond, L. (2025). Generative AI at Work. The Quarterly Journal of Economics, 140(2), 889–942.
Huang, M.-H., & Rust, R. T. (2018). Artificial Intelligence in Service. Journal of Service Research, 21(2), 155–172.
Huang, M.-H., & Rust, R. T. (2021). A Strategic Framework for Artificial Intelligence in Marketing. Journal of the Academy of Marketing Science, 49, 30–50.
Ding, Z., Zhang, Y., Sun, J., Goh, M., & Yang, Z. (2026). Harmonizing Human Touch and AI Precision in Customer Service. Journal of Service Research, 29(3).
AI or Human: How the Type of Information to Be Disclosed Alters Customer Service Agent Preferences. (2026). Journal of Retailing and Consumer Services, 89, 104621.
