A funny thing (not funny “ha-ha”, funny weird) has happened to customer experience technology lately.
The business problems have stayed the same: we all still want to reduce cost-to-serve, improve retention, increase customer satisfaction, shorten resolution times, and stop answering the same question 10,000 times a month.
The conversations, however, have changed a lot. More often than not, a CX technology evaluation can feel less like a discussion about actual business outcomes and more like an AI vocabulary competition. Autonomous agents, agentic workflows, AI orchestration, AI-first support. For the lady, perhaps an AI-enhanced AI for your other AI?
At some point, the industry became so fascinated with what the technology could do that it stopped spending enough time discussing what the business was actually trying to achieve in the first place.
That's a problem because "becoming AI-first" isn't an outcome. Reducing onboarding churn is an outcome. Increasing retention is an outcome. Lowering cost-to-serve while maintaining service quality is an outcome. Improving self-service success is an outcome.
One is a means to an end; the others are reasons a business would break its piggy bank in the first place. The distinction sounds obvious, but it can become really easy to forget once the really sexy demo starts.
At least, they shouldn't. Most CX leaders don't walk into a quarterly planning session with their most important question being “how can we increase AI utilization by 20%?”
They're usually dealing with much more practical concerns, whether that's support costs growing faster than the business, onboarding journeys that aren't converting as expected, customers repeatedly contacting support about the same issues, or renewal rates that have started moving in the wrong direction.
Technology then enters the chat because it's supposed to help solve those problems. Somewhere along the way, though, many buying conversations start treating the technology itself as the destination. The questions change to:
Those aren't inherently bad questions (although they sure are on their way there), but they are absolutely incomplete ones.
A company can automate 80% of its contacts and still create a worse customer experience than a competitor automating 40%. A chatbot can achieve excellent containment rates while frustrating customers into opening a second ticket through another channel. Metrics tell stories for sure, but they don't always tell the whole story.
Customer experience isn't immune to technology fashion trends. At various points, businesses convinced themselves they needed an app, live chat, omnichannel support, and now an "AI strategy”.
Some of those investments created real competitive advantages, and others probably left behind painful and expensive reminders that adopting a capability and creating value are not the same thing. CX is littered with examples of companies launching something they felt they needed because “everyone else is doing it”, only to discover later that customers hadn't been asking for it in the first place.
The pattern is usually the same: a new channel, platform, or technology becomes available, the market gets excited, and organizations start asking how much of it they can implement before they've figured out where it actually belongs. A solution looking for a problem.
You can usually spot it when any conversation shifts from customer outcomes to adoption metrics. Instead of asking whether customers are finding answers faster, resolving issues more easily, or staying longer, teams start focusing on containment rates, deflection percentages, and automation volumes.
One reason AI-first thinking becomes so attractive is that automation produces the kind of metrics executives tend to love. Containment and deflection rates, automation percentages, and cost savings are all easy to measure, easy to benchmark, and easy to drop into a quarterly presentation.
Customer outcomes are much less cooperative. Trust doesn't show up neatly in a pretty dashboard, and neither does the customer who decides to renew because a difficult conversation was handled exceptionally well six months earlier.
As a result, organizations can end up optimizing for the thing that's easiest to count instead of the thing they're actually trying to achieve. That's how perfectly sensible goals like improving customer experience somehow turn into heated discussions about deflection rates.
Take a cancellation request: in one business, the ideal outcome may be helping the customer leave quickly and efficiently. In another, that same interaction may be one of the most valuable retention opportunities in the entire customer journey. The customer is asking the same question, but the business objective is completely different.
Yet this is where outcome-first thinking often gets replaced by capability-first thinking. The conversation goes from "What are we trying to do?" to "What can the technology do?" Those sound similar until you realize one starts with the customer and the other starts with the software.
Treating both scenarios the same just because the technology can automate them misses the point entirely. A cancellation flow isn't successful because it's automated, it's successful if it achieves the outcome the business actually wanted, whether that's efficiency, retention, insight, or some combination of the three.
Before evaluating any AI-first CX platform, suite, automation initiative, or outsourcing partner, it's worth stepping back and answering this: what are the two or three outcomes you're actually trying to create? (And yes, sometimes one of those outcomes is saving money. That's perfectly fine; just be honest about it. Not every initiative needs to be wrapped in a story about delighting customers and reinventing experiences).
You shouldn’t try to focus on every metric on the dashboard here; most CX teams already have more metrics than they know what to do with. The challenge is identifying the handful of outcomes that would actually make a meaningful difference to the company.
Once those priorities are clear, the buying process gets a lot easier. Instead of getting lost in feature comparisons and debates about which flavor of AI is currently fashionable, you can evaluate tools based on their ability to support a specific objective.
It becomes much easier to identify where automation creates value, where human expertise creates value, and where both need to work together. Without that clarity, companies can spend months building a remarkably sophisticated operation only to discover they've become exceptionally good at optimizing something that wasn't especially important to begin with.
Many AI-first strategies are built around an assumption that rarely gets stated explicitly: if an interaction can be automated, it probably should be.
There are plenty of situations where that really is true; most customers don't want a meaningful human connection with a password reset flow, and few people are hoping to spend extra time discussing a shipping address update, so sure, automation is the better experience.
The yucky part comes when that logic gets applied to everything, indiscriminately. A customer considering whether to renew, a complex troubleshooting issue that spans multiple systems, or a high-value account that's already frustrated after several service failures is a very different challenge.
In those moments, the value isn't always found in efficiency alone, it’s in context, judgment, flexibility, and the ability to understand what the customer is actually trying to accomplish.
Technology gets most of the attention because it's the part people can see. Governance tends to get far less excitement despite being the thing that determines whether the investment actually works or not.
Questions around ownership, knowledge management, content reviews, escalation paths, and failure handling aren't very glamorous, but they're often what separates successful implementations from expensive experiments.
At some point, every organization has to decide who maintains the knowledge base, who reviews automation failures, and how customers move between automated and human experiences when things become more complicated.
Those decisions sound operational because they are operational. They're also the decisions that shape the customer experience long after the implementation team has moved on to the next project. A workflow doesn't become self-governing because AI is involved, and customers are remarkably good at finding the gaps between what a process was designed to do and what it actually does in practice.
One of the stranger behavior patterns in CX is spending six figures on sophisticated automation while leaving the knowledge base in a condition best described as "good luck, I guess."
Technology has a habit of amplifying whatever system already exists. If the underlying workflows, knowledge, and governance are strong, that's great news. If they aren't, automation can increase confusion just as efficiently as it increases output.
That's also one of the reasons "AI-first" can be a misleading strategy. There's a tendency to talk about automation as though it arrives, packs itself out of the box, solves a problem, and takes care of itself from that point forward.
Somebody still needs to maintain the knowledge, review failures, update workflows, and make sure the experience is producing the outcomes the business actually cares about. The technology may be new, but the need for operational ownership sure isn't.
Buying technology is the easy part. Deciding what success should look like is considerably harder because it forces organizations to make choices. Which customer journeys matter most? Where does human involvement create disproportionate value? Which outcomes deserve investment, and which just sound nice in a strategy document?
Making choices is hard, and none of those conversations is as exciting as the promise of an AI tool that will deflect 80% of your volume. They're also far more likely to determine whether the investment creates meaningful business value six months later.
That's why the organizations seeing the strongest results from AI in customer experience aren't the ones deploying the most automation. More often, they're the ones with a clear understanding of what they're trying to achieve and enough discipline to resist solving every problem with the same tool.
Once the destination is clear, decisions about technology, workflows, governance, and staffing become much easier to make. The opposite approach is how companies end up with a very impressive AI strategy and a lingering suspicion that nobody can quite explain what it was supposed to accomplish in the first place.
Customers don't wake up hoping their next support interaction contains the optimal ratio of humans to AI. They want their issue resolved accurately, efficiently, and with as little effort as possible.
The businesses getting the most value from AI tend to understand that distinction; they're not chasing automation for its own sake. They're using technology, people, process, and governance intentionally to achieve specific outcomes.
That's a very different strategy from just becoming “AI-first”, and probably a much more useful one.