Post by Christian Schacht

A telco client expected their AI chatbot to cut service costs immediately. Instead, it spiked contact center volume by 14% in month one. Here's what we learned about the translation problem. The company launched a customer service chatbot with the best intentions. Lower barriers to support. Faster answers. Reduced workload. What actually happened: Customers flooded into the new low-friction channel. That part worked. But the chatbot couldn't connect information across legacy systems. Questions that needed cross-system context got bounced to specialists. Support load increased instead of decreased. They had solved half the problem. They removed the barrier to entry. They didn't solve how to actually answer customers without human escalation. **The counterintuitive insight: Accessibility without the right infrastructure is a liability, not an asset.** Three things had to change: → The database was split across legacy systems. A customer asking about billing needed data from Account, Usage, and Payment History. But these lived in separate places. The chatbot couldn't bridge them. Specialists had to manually hunt for answers. The fix: integrate the data layer so the translation layer can actually translate. → Not every customer question should be automated. Some need nuance, context, judgment. The shift: clearly define which request types can resolve end to end without human input. Route everything else intelligently instead of forcing the customer or the system to figure it out. → When automation hits a boundary, escalation has to be smart. Not a dead end bounce back to the customer or a generic queue. Route with context. Preserve what the customer and the system have already figured out. Make the specialist's job easier, not harder. Only after these three pieces aligned did workload actually drop while satisfaction improved. Adding accessibility without solving the translation problem doesn't reduce support burden. It exposes how broken the underlying system already was. What's the translation problem in your organization? Where are customers hitting friction because your systems can't talk to each other?