Skip to content
Fast Company Executive Board

The Fast Company Executive Board is a private, fee-based network of influential leaders, experts, executives, and entrepreneurs who share their insights with our audience.

AI fatigue happens if we strap rockets to horse carriages

While good at inventing new technologies, humans are less effective at implementing them.

[Photo: Getty Images]

Profile photo of Beaumont Vance

Senior Vice President, Data, Analytics, and AI · Paychex, Inc.

Over the past few years, AI has exploded into our work lives, giving rise to a phrase we now hear more often: AI fatigue.

For more than 30 years, technology has promised to make work easier. In many ways, it has. But every breakthrough has also brought more applications, more platforms, more logins, and more workflows.

Think about your workday. Most people move constantly between email, messaging, customer relationship management (CRM) databases, calendars, expense systems, analytics platforms, support portals, and documentation repositories. The modern workplace is drowning in software.

Now AI is being layered on top of that same fragmented environment. We have AI tools for many of these applications, adding another layer of complexity. People were already overwhelmed by the number of applications to navigate, and now they’re asked to learn a growing collection of AI-powered experiences.

The result is often exhaustion, or AI fatigue.

One solution from the past

Ironically, the solution to this problem dates back decades.

Before Google launched in 1998, information was already available across the internet. The challenge was finding it. Users spent hours navigating websites and directories to find answers.

Then Google gave us something beautifully simple in a search box that let users describe what they wanted and get an answer. It hid immense complexity behind a natural-language interface.

We seem to have forgotten that lesson. While consumer technology’s user experience has improved dramatically, many organizations still rely on a maze of dashboards, menus, forms, permissions, and disconnected systems. Employees are expected to know where information resides and how to retrieve it.

AI should end that, but much of today’s AI innovation is recreating the problem by building individual AI tools for individual business applications. Despite the short-term value, it is not what people are looking for.

The real breakthrough

AI can deliver the same experience to the enterprise that search engines delivered to the internet. Employees shouldn’t need to know which system contains an answer, which report to run, or which application owns a process. They should simply be able to ask for what they need.

Behind the scenes, hundreds of specialized agents can perform the work. One accesses CRM data. Another reviews policies. Another updates an expense system. Another drafts an email, and another validates compliance requirements. The user doesn’t need to orchestrate that complexity. Instead, they simply ask.

We experienced this firsthand when building an internal AI solution for sales teams. Representatives often had to navigate multiple systems during customer conversations to find the right information at the right moment. The information existed, but finding it was the challenge.

Rather than introducing another application, we asked a different question: What if the answers were immediate?

The result was a solution that surfaced guidance in real time. New hires using the tool achieved their first sale in less than half the time. It was a productivity gain and redesign of the experience itself.

Our mindset is the challenge

Before the internal combustion engine was first introduced, people spent centuries relying on horses. So when they encountered a powerful new engine, their first instinct was to attach it to a carriage. For years, people treated automobiles as horseless carriages. Eventually, people realized the engine required an entirely different design.

AI represents a similar moment. We have an extraordinarily powerful new engine, yet many organizations simply bolt it onto existing workflows, dashboards, forms, and systems. That approach will deliver efficiency gains, but is unlikely to create the level of transformation many expect.

In some cases, it may simply create more AI fatigue. The better question is “What should work look like if AI becomes the interface?”

That question forces us back to first principles. What does the user actually need? What information matters? What decisions need to be made? Why should employees have to know where data lives, which system owns it, or which workflow completes a task?

The next step

The next major leap in AI will come from applications fading into the background. Systems of record will still matter, because organizations will always need various technological platforms as infrastructure, while AI should become the interface.

This thinking informed the development of WISE, Paychex’s workforce intelligence platform. WISE acts as an intelligence layer that surfaces insights, recommends actions, and helps execute work within customer-defined parameters, bringing intelligence to the user instead of sending the user searching for it.

That distinction reflects a broader truth. The AI platforms that endure will be built on trusted data, domain expertise, and real-world context because that foundation lets AI move beyond answering questions and toward anticipating them.

Enterprise technology today is powerful but often chaotic. AI’s greatest promise may be making the tools we already have finally feel usable.

If we simply attach AI to every existing platform, we will end up with a faster horse carriage. But AI is closer to a rocket, and rockets need rocket ships.

The organizations that rethink work around that reality will be the ones that unlock its full potential.

Beaumont Vance is the senior vice president of data, analytics, and AI at Paychex.