The killer AI app: helping customers navigate their account data.

The world runs on data, but many people struggle with it. As a universal data-to-language interface, AI can create loyal customers and more productive service teams.

About a decade ago, I attended an event organised by my Wealth Management client for their customers. Alongside presentations on markets and regulations, we set up a booth to showcase an early version of the client portal under development.

I was ready to discuss the interactive charts, mobile-friendly layout and security checks, but virtually no one asked about that. What they wanted to know was: Will this portal answer my questions? Is my account on track? If not, why not and should I take action?

After the event, we discussed this reaction with my client. Were we solving the wrong problem?

I. Not another dashboard

Many people struggle with data literacy, but most value the convenience of reporting apps. This is especially true in finance, but I have spotted the same trend in commerce, utilities, and even B2B. Show me a dashboard, and I will show you some frustrated customers.

Customer service teams mirror that frustration. Given the commercial pressures and the ever-increasing admin workload, many are growing impatient with having to explain the same things again and again. In a world that runs on data, the data literacy gap between professionals and their customers is getting wider.

Closing that gap is a massive challenge in customer experience. Explaining data is key to customer loyalty and team productivity. It is not a feature - it’s the whole game.

II. Data-to-language converter

From an experience design perspective, AI is a real-time, data-to-language conversion engine. Automated language generation grabs the headlines, but it is merely a delivery mechanism for the magic that is computerised comprehension. Training an AI to understand a customer’s account and then talking to them about it unlocks an entirely new type of experience.

“Language generation is merely a delivery mechanism for the magic that is computerised comprehension“

Dashboards may be redesigned as personal news reports, each “story” providing a trigger to discuss the underlying data in plain language. Customers can explore further by running calculations and projections based on simple instructions rather than fiddly inputs and sliders.

And that’s just the start. Trained on live datasets, AI-enabled apps can go beyond the interface and deliver proactive, predictive experiences.

III. Proactive and predictive service

Helping users understand and explore their accounts is an excellent start. But a great experience does not just give customers great tools when they log in; it tells them when to log in, and reassures them that everything is OK when they don’t. To build trust, you must be transparent. To build loyalty, you must be proactive.

You can instruct an AI to simulate user interactions regularly, alerting service teams if irregularities are detected. These irregularities can be investigated, discussed and solutions deployed quickly, reinforcing the impression of a smooth and responsive customer support.

Agents with some autonomy can conduct deeper searches and compare customer data, uncovering novel insights and early warnings that humans might easily miss. Predictive recommendations can help the service team address issues before they become problems. There is no better incentive to remain loyal than the absence of problems.

IV. Soft and hard guardrails

Soft guardrails are the first thing to consider when building AI customer apps. Output must be labelled and caveated as “generated by AI”. The model should be forced to remain within strict, predefined limits. Responses should be logged and reviewed by a human crew that is readily available for escalation.

To ensure data integrity, “hard” guardrails should be used. The app might block the AI if a particular word pattern is detected (e.g., “should I…” often denotes advice, usually a no-go). Similarly, non-AI checks should be performed to authorise access and verify data accuracy.

For sure, creating AI apps - especially customer-facing ones - is not trivial. But the technology is maturing rapidly, exposing a wealth of opportunity to early adopters.

V. Scaling up the human touch

Back in the pre-AI, pre-COVID world of the Wealth Management event, the portal was successfully launched. Despite great feedback, however, adoption remained modest. With the benefit of hindsight, we provided a delivery solution (“Here is a mobile-friendly, secure portfolio app”) to a data comprehension problem (“What does it mean? What do I do next?”).

“Dashboards are still an impersonal way to deliver data. AI can help you apply a human touch at machine scale.”

This case study is typical of the limitations of the digital toolkit and why we are so excited about AI experiences:

• Dashboards can visualise data but not explain it, failing to address customer and service team frustrations.

• Data-to-language AI can help customers understand their accounts while freeing up teams for more high-value tasks.

• Proactive querying can check for data irregularities and alert customers and service teams before problems arise.

• Soft and hard guardrails must be put in place, as any application using personal data must adhere to the strictest safety rules.

How far we have come. I can still remember switching from letters in the post to “paperless” customer portals. They were just folders of print-ready PDFs, but at least I didn’t have to file them.

When dashboards arrived, my accounts were more accessible, secure and interactive, but it was still an impersonal way to deliver data. AI will go the full length and help companies apply a human touch at machine scale.