Andy Watts is the Chief Financial Officer for Brown & Brown, Inc.—ensuring financial discipline, value creation and operational excellence.
Every organization wants to talk about artificial intelligence (AI). Yet despite record-breaking investments in AI and data platforms, many are seeing underwhelming returns. When results don’t materialize, the assumption is often that the technology isn’t advanced enough. But what if the real issue isn’t the algorithm, but the data beneath it or the willingness to truly embrace AI?
Since earlier this year, I’ve had a unique seat at the table as both CFO and interim CIO—responsible for how our organization invests in AI and how those tools are built and deployed. This perspective has given me a firsthand view of why so many AI initiatives fail.
The reality is that a centralized AI strategy means very little if it doesn’t translate to what front-line teams need to drive incremental value. Companies often build impressive, centralized data systems to feed AI’s machine learning, neural networks and natural language processing, only to realize they haven’t answered the right questions—the ones their business units and customer-facing teams are asking every day.
Inside large organizations, the questions change dramatically depending on where you sit. An executive team may want to know how many new accounts were written with a specific company this year. But a vertical market leader might want to know how many policies in a sub-line of business were quoted versus bound. Those are fundamentally different queries that require a unique approach to data collection and analysis. When only questions from leadership are answered, the disconnect between spend and perceived value grows.
When AI investments are designed only to satisfy leadership’s dashboards, millions can be spent on platforms that generate reports but never help teams make better decisions. The deeper cost shows up on the technology side. Tools that don’t match the realities of how data is captured and used quickly lose credibility with the people who need them most. Once front-line teams stop trusting the data—or can’t see how the tools help their day-to-day—the adoption curve flatlines.
Best Practices For Bridging The AI Tool Gap
So, how do you avoid spending millions on AI tools that no one uses? And, more importantly, how do you invest in the right AI tools that will move the needle for your teams? The lesson I’ve learned in balancing investment decisions with deployment realities is that AI only delivers ROI when data strategies connect the enterprise view with operational execution. These answers can be found by anchoring your AI efforts in a thoughtful, scalable data strategy that accounts for needs at both the macro and operating levels of the business.
1. Build dual-level data strategies.
It’s not enough to have a single “master plan.” You need a deliberate data strategy that cascades from enterprise goals to front-line realities. The goal is to create a throughline between company-wide insights and the information each team needs to act, whether they’re building products, serving customers or managing risks.
This dual-layer strategy ensures that a top-down vision doesn’t drown out the bottom-up needs that power the business.
2. Define a common data language and a translation layer.
One of the most overlooked challenges in data strategy is taxonomy, or the classification/naming of fields that lead to specific data extraction. For example, what does “customer” or “industry classification” mean in each of your businesses? Without a common taxonomy, you will end up with siloed teams speaking different languages, even when they’re technically mining for the same data.
Think of it like this: You’re speaking French, I’m speaking German and someone else is speaking Mandarin. The solution isn’t forcing everyone to switch languages—it’s about building a translation layer. A well-defined data taxonomy acts as that interlinking bridge. It helps you identify what’s truly shared across the business versus what’s unique to a line or function.
That’s also a governance issue. Without a clear data definition and ownership model, accuracy suffers. And, if your data isn’t accurate, your AI tool will probably not solve it.
3. Make data accessible, or AI can’t help you.
Even the smartest AI can’t fix inaccessible data. Many companies still store critical operational information in emails, spreadsheets and static files. If you’re asking an AI to deliver insights from fragmented or buried data, it’s like asking an intern to do your taxes without any instructions.
AI can help solve this challenge, but only if you direct it properly. With “smart prompting,” AI tools can extract data from static documents, pre-populate forms and reduce manual entry. But it requires human oversight, ongoing validation and the infrastructure to ingest that data into the right systems.
Data inaccessibility and reliability are business problems with real consequences. If your front-line teams don’t trust the data—or can’t use it easily—the ROI of your AI investment will stay theoretical.
Data First, Then Intelligence
Before investing too much in AI tools, ask yourself if you have the right data capture in place to feed it. Also, have your business leaders clearly defined what data is critical or key to running their business? If not, you will need to unpack this topic so you know what you are trying to solve.
AI isn’t a silver bullet. In fact, it can amplify the disconnect between leadership and operations if you’re not careful. A data strategy that balances big-picture ambition with front-line usability can be a force multiplier.
Get your data house in order, then let AI do what it does best.
Forbes Finance Council is an invitation-only organization for executives in successful accounting, financial planning and wealth management firms. Do I qualify?