Table of Contents
What CCW Nashville revealed about buyer fatigue, and the workforce decision hiding beneath every AI pitch.
Last week at CCW Nashville, customer contact leaders came together to explore the future of our industry. I joined a panel discussing how agentic AI is changing the role of the human agent in customer experience and to what degree this is changing the ways we have worked for decades.
One of the strongest themes I heard from attendees had little to do with excitement about the future. The common theme from most buyers was exhaustion.
Similar promises. Similar terminology. Similar demonstrations. Everyone seems to be selling “agentic,” “seamless,” “intelligent,” and “transformational.”
When every vendor sounds the same, buyers have to work harder to understand what is actually different. Their fatigue is understandable. Their confusion is consequential. Because beneath those interchangeable marketing messages sits a decision that will shape the enterprise for years:
Are you buying AI that helps humans deliver service or AI that can deliver service? With humans providing expertise, guidance, and better outcomes?
Those are fundamentally different workforce models. They deserve fundamentally different buying decisions.
A better-equipped human workforce is still a human-first workforce
AI that summarizes conversations, recommends responses, retrieves knowledge, and automates after-call work can create meaningful value. But if a human still has to interpret every request, navigate the systems, execute the actions, and own the resolution, the operating model remains human-first.
You have improved how the work gets done. You have not fundamentally changed who does it. That distinction matters when an enterprise believes it is purchasing an autonomous service capability but receives another productivity tool.
If every customer interaction still requires a human to finish the job, are you really getting the most from today’s advanced technical capabilities? An impressive assistant and an accountable AI service agent should not pass the same procurement test.
AI-first changes the starting point
An AI-first workforce starts by asking which customer needs AI can safely resolve from beginning to end.
Within defined permissions, AI understands the request, accesses relevant information, executes authorized actions, and verifies the outcome. Humans establish policies, observe performance, guide critical decisions, and only assume ownership when absolutely necessary to optimize customer and business outcomes.
Consider a billing dispute. In a human-first model, AI might summarize the complaint, retrieve a policy, and suggest a credit. The employee still investigates, decides, executes, and confirms. In an AI-first model, AI can investigate the charge, apply approved policy, issue an authorized adjustment, update the account, and confirm resolution. When the situation exceeds its authority, it brings in a human with the evidence and action history intact. The human guides the AI towards the optimal decision, but the AI never relinquishes ownership.
The difference is responsibility for completing the work. AI-first should mean accountable execution, not to be confused with simply another chatbot at the front door. And it should never mean trapping customers in automation. An explicit request for a person, a sensitive circumstance, or a decision requiring human authority must have a workable path to human help.
Buyers need to stop grading the pitch
Buyer fatigue is a reason to demand better evidence. The due diligence process should expose the operating model behind the presentation. Start with six questions:
1. What can your AI actually finish?
Ask for a real workflow in your systems, with your policies and permissions. Require the vendor to distinguish answering a question, recommending an action, and completing a resolution.
2. Where does human labor remain?
Map every approval, manual task, exception, and transfer. Include the people maintaining knowledge, monitoring quality, and repairing failures. Hidden labor belongs in the business case.
3. What happens when the interaction gets difficult?
Test ambiguous requests, conflicting information, unavailable systems, policy exceptions, and customers asking for a person. A successful demonstration should reveal boundaries as clearly as capabilities.
4. Can we see, and control, what it did?
Require traceable actions, enforceable permissions, auditable decisions, and a practical way to pause or correct behavior. Accountability cannot depend on accepting the vendor’s explanation after something goes wrong.
5. How will we prove the customer’s problem was resolved?
Measure verified resolution, repeat contacts, customer effort, quality, and total cost per resolved issue. Containment alone cannot tell you whether the customer succeeded or simply gave up.
6. What does our workforce become?
Ask the vendor to describe the future operating model: roles, staffing assumptions, observability and optimization responsibilities, training, and the transition required to get there. The agentic workforce will look nothing like the workforce of today.
If the answer to that last question is simply “your agents will be more productive,” examine whether the solution matches your ambition.
The human role deserves a more ambitious conversation
During discussions about AI, we often reassure employees that humans will handle “the complex interactions.” That is an incomplete workforce strategy.
An AI-first enterprise needs people who can supervise AI performance, define policies, improve workflows, investigate failures, and provide guidance where automation reaches its limits. Those responsibilities require skills, authority, career paths, and investment.
Leaders also need to be honest. Changing who performs the work can change staffing needs. Credible workforce planning must address both new opportunities and the realities of transition. The human role should evolve deliberately, alongside the technology.
Choose the workforce before you choose the vendor
Enterprises differ in their readiness, risk tolerance, customer needs, and operational constraints. The right deployment sequence will differ too. But those differences do not excuse ambiguity about the destination. A phased implementation can support an AI-first strategy. A collection of disconnected assistants can also leave an enterprise permanently dependent on the same human workflows it intended to transform.
Before selecting a vendor, decide which future you are building and demand evidence that the solution can get you there. The most expensive AI purchase may be the one that makes your current operating model slightly better while persuading you that you have transformed it.
Buyers have heard enough promises. Show them who does the work. Show them who owns the outcome. Show them what their workforce can become. If you would like to connect on this topic, drop me a message on LinkedIn.
