As a CMO, I pour serious money into attracting customers. Websites, ads, market research, demand programs — all designed to create conversations with the right people. Getting someone to choose you is hard work.
But what happens after that hard work pays off and they become a customer? Can they easily call you with a question? If something goes wrong, are you there to help? Or are you hoping they’ll quietly self-serve their way to an answer?
We know customer expectations are rising fast, with AI making every other interaction feel instant and frictionless. We know happy customers buy more and refer more. We know the technology to deliver great service experiences exists. What we didn’t know was whether the Global 2000 — the world’s largest companies — were actually providing it.
We decided to find out.
In the first-of-its-kind study, Parloa’s research team deployed AI agents to mystery shop 10,000 enterprise websites, conduct 4,000 chat interactions and dive into 100 phone trees across the world’s largest companies.
We walked away with four uncomfortable truths, and they point to a current state that’s even worse than we anticipated.
Uncomfortable Truth #1
We’re Deflecting Instead Of Connecting
The first thing we looked for was simple: When a customer wants to reach you, can they?
Shockingly, almost half (43.3%) of the websites we evaluated don’t give customers a clear path to reach support. No phone number listed, no obvious chat functionality. Just an FAQ page, a search bar and a quiet hope that the customer figures it out on their own.
That’s a deliberate choice companies are making. Enterprise CX has been architected for containment: Get customers to self-serve, keep them off the phone and minimize agent time.
I get the cost logic. But a customer who can’t reach you won’t stay a customer for long. Every interaction they can’t complete is lost revenue. But the cost of deflection isn’t obvious, so companies keep doing more of it.
Uncomfortable Truth #2
The Smart Chat Illusion
Most companies believe their chat solutions work when what they’ve actually built is a system optimized to look like it’s doing its job.
Almost all the chatbots we tested are built on rule-based flows in the form of rigid decision trees that can only handle questions they were explicitly programmed to answer. Ask something outside that narrow band, and the experience falls apart. At some level, companies know this — which is why the metrics they track are designed around the customers with simple, predictable questions. When you measure only the easy cases, the numbers look great.
Measure everyone else, and you get the real picture: only 8.9% of chat conversations achieved what the customer was trying to do. The other 91% left without help, probably more frustrated than before they started.
Uncomfortable Truth #3
Decades-Old Voice Experiences
Ninety-nine percent of the voice systems we evaluated rely on decades-old automation, or none at all. Not because the technology doesn’t exist to change this — it does. It just hasn’t been prioritized.
When customers are frustrated enough to call, they’ve already decided they need to talk to someone. They found the number, even though it’s buried on the website. They dialed. They’re willing to work for help.
Here’s what they get for their efforts:
We mapped the voice journey at large enterprise accounts. In most industries, customers face three to four interactive voice response menu levels before reaching a human. In banking and insurance, six to eight path options per menu. Callback options are almost nonexistent. The longest hold time we observed was 90 minutes.
Ninety minutes.
Voice AI has been layered into a lot of these systems, but as cosmetic renovation rather than substantive change. A customer hears a friendlier prompt, states what they need and gets routed to the same rigid decision tree that existed before. Context disappears the moment a handoff happens, so they have to repeat everything.
While the AI sounds more modern, the experience is as old school as it gets.
Uncomfortable Truth #4
Not Ready For Personal AI Agents
AI is already changing how people interact with the world: booking travel, researching purchases, managing accounts. The next wave is personal AI agents. These systems reach out directly to companies to resolve issues, negotiate and transact on behalf of consumers.
Early adopters are already deploying them, but companies are woefully unprepared.
We tested readiness for agent-to-agent interaction across the Global 2000 and found that only one percent of enterprises are prepared for it. Authentication frameworks assume a human caller. IVR menus require button presses. APIs don’t exist.
The companies spending the most on CX automation today are building systems that will be incompatible with how customers want to interact in two years. They’re optimizing for a world that’s already going extinct.
The Revenue Being Left On The Table
When we started this research, I suspected enterprise CX was underperforming. I didn’t expect it to be this bad.
Each one of these uncomfortable truths has revenue consequences. A customer who can’t reach you picks a different brand. A customer stuck in a broken chat loop doesn’t buy more. A customer who waited 90 minutes on hold tells her friends.
Treating CX as a cost center to contain is the wrong perspective. There are billions of dollars in revenue potential sitting inside the service channel, in customers who want to buy more, renew faster and refer enthusiastically. You just have to make it easy for them to get help when they need it.
The technology to do this is here. Voice agents that actually resolve issues, chat that works seamlessly and systems built for agent-to-agent interaction are available. Parloa customers are already delivering these experiences.
Are you building toward connection with your customer service approach? Or are you still burying your phone number, hoping to avoid talking to those customers we worked so hard to win?
