Tip #2: Your Call Volume Is Plowing Upward Because Residents Are Being Told to Wait
Consider this.
Have you ever dialed a municipal hotline, endured the hold music, at last spoken to a live agent, and then been told that the agent would have to return your call because the answer wasn’t in their head?
That is not a hiring crisis. It’s not a software glitch.
That is a knowledge management crisis.
And across the nation, agencies are currently rushing to launch AI chatbots without addressing that underlying failure first. They are attempting to construct a skyscraper on a compromised bedrock. Residents are footing the bill for it daily.
The numbers do not misrepresent.
Last year, in a 2025 survey of over 300 executives and AI specialists, 61% of respondents cited unreliable, misleading, or inaccurate answers generated by AI as their top worry when adopting AI in the workplace. The warning issued by Gartner was even more explicit — “100% of AI virtual assistant projects which do not integrate with modern knowledge management will not meet their intended goals for improved customer experience and reduced cost.”
100%.
This is not a gamble. This is the predictable result of skipping this crucial step.
Here is exactly what occurs when you give AI a broken database to mine.
Someone uses your chatbot or digital portal for assistance. They have a question. Are you eligible? Which documents will I need? When can I expect my permit to be issued?
The AI searches its database for that answer — and it cannot locate a definitive, up-to-date, or factual source of truth. It hesitates. It directs the issue upward. It tells the resident to try another time or expect a callback later.
This is why so many chatbot projects stall before they ever start. They were designed to answer questions, not to solve the problems that caused those questions in the first place. Those are two distinct things.
And resolving them costs money.
Approximately 45% of people express anger when it takes just 5 to 15 minutes for a service agent to reach them and resolve the issue. And 71% of individuals state they expect to receive the answer immediately after making that phone call.
The public is not being unreasonable. They are simply stating what they need to receive. A quick, precise response upon the very first contact.
If you fail to provide that solution, they pick up the phone again. They use a different digital channel. They tell their neighbors. And your call volume increases — not decreases.
Here is what that issue typically isn’t.
I want every government official reading this to hear this message loud and clear.
When your AI is delivering late, incorrect, or partial responses, the technology is usually not to blame. That issue stems from the information sitting behind that technology.
Important data is trapped in archived threads of email correspondence. Procedural information known only to a select few within your organization. Eligibility criteria are divided across three separate online repositories. Policy revisions that simply never made it to the public-facing website.
When the knowledge base an AI system uses is not synchronized with the latest, correct information, residents are left feeling irritated and disbelieving. When someone follows the guidance of an AI agent — and is subsequently given incorrect directions — their original issue worsens. So, too, does their faith in your agency.
We work in a trust-based sector. Each poor interaction costs us more than we think.
Here is what effective knowledge management actually seems like.
Call centers that record the fewest number of escalations are three times more likely to have their official resolution procedures properly published and easily available to the front line — not locked inside a management team member’s digital inbox or on a shared folder that everyone forgets to touch.
This same logic applies specifically to the implementation of artificial intelligence. The more thoroughly and factually documented your knowledge becomes, the better the AI will function. Without fail.
Before you begin your search for a software application that incorporates artificial intelligence, ask these three questions before anything else:
Are we keeping our knowledge in one single place that is available for both staff members and AI systems?
Is our data up-to-date, reliable, and phrased in everyday language, not bureaucratic speak?
Does our organization have a defined protocol to keep this resource up to date when rules or services are altered?
If you answered no to any of the above questions, that is your core issue. And no AI software on the market will solve it on your behalf.
“AI will only function better when combined with existing work patterns and bolstered by reliable knowledge protocols. It does not fix bad information; it magnifies it.” – Contact Center Benchmark Insights, 2025-26
My final recommendation to you.
Slow answers almost always hide knowledge management issues dressed up like a technology issue.
Prioritize expanding the scale of your artificial intelligence. Examine the information you truly possess and the methods you use to preserve it. Tidy it up. Bring it together. Keep it fresh.
Because improved response times don’t begin with a smarter chatbot.
They begin with smarter information.
Rosetta Carrington Lue is the Founder and President of GovCXP Digital Partners and Co-Founder of the National 311 Executive Council. She is one of the foremost experts on government customer-centric leadership and a highly sought-after speaker and advisor.
Building Welcoming, Connected, & Innovative Government Customer experience 