Bad Data Is Preventing You From Realizing AI’s Potential

• September 1, 2025

By Kasey Connors

You decided to try out the new artificial intelligence (AI) assistant that everyone’s talking about. You upload two policy documents and ask it to summarize the key differences for your client—something that normally takes your staff 30 minutes or more.

The tool delivers a clean, well-written summary in seconds. There’s just one problem: It compared the current policy to one that was canceled a year ago. The wrong document was attached to the client file. It wasn’t labeled correctly. And now, you’re back to double-checking everything manually.

Trying to use AI on top of disorganized, outdated or unreliable data doesn’t create efficiencies, it creates more work. However, AI isn’t the problem. The problem is that you skipped a step.

AI has the power to transform how agencies operate, serve clients and grow. But before you can realize that future, you need to make sure your foundation is ready. That means getting serious about data.

The Big “I” Agents Council for Technology (ACT) data workgroup recently dug into this problem—and the finding should be a wake-up call. They identified common red flags that indicate agency data isn’t ready for AI, including: duplicate records due to inconsistent naming; misused or blank fields, like ZIP codes entered as “00000” or policy types filed under the wrong category; outdated contact information that blocks client outreach or causes missed renewals; inconsistent formatting, such as five different versions of phone numbers; and notes or documents attached to the wrong account.

These issues aren’t just frustrating; they’re costly. Bad data leads to operational inefficiencies, like wasted time re-verifying information and lost revenue from missed renewals and incorrect commissions. It can also lead to stress across departments, client churn due to poor service and communication, and even errors & omissions exposure due to incorrect or missing documentation.

AI thrives on patterns. It learns from data. But if that data is wrong, outdated or inconsistent, it will make bad recommendations.

Only 28% of insurance executives say AI is more technologically challenging than previous tech waves, according to a report from Genpact. The real blockers in implementing AI are governance, risk management and poor data quality. Meanwhile, just 2% of insurance executives say their teams are fully AI-fluent. That means most agencies aren’t just lacking tools—they’re lacking the structure to use them well.

Here are five things to do before you implement AI in your agency:

1) Clean and structure your data. Standardize contact formats, remove duplicates and validate key fields like policy numbers and effective dates.

2) Review your workflows. Identify where data is being entered inconsistently or where key steps are skipped. Fixing one broken handoff could save hours each week.

3) Establish governance. Create policies around naming conventions, field usage and where documents are stored.

4) Invest in training. You don’t need everyone to be an AI expert. But you should understand how tools like ChatGPT or Copilot work and how to spot when something looks “off.” AI-fluent agents save up to 12 hours a week, according to the 2024 “Agent-Customer Connection Study” by Agent for the Future.

5) Set realistic goals. Are you aiming to improve turnaround time? Reduce manual entry? Identify cross-sell opportunities? Define success so you can measure progress.

Kasey Connors is executive director of the Big “I” Agents Council for Technology.