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.

Last year’s AI hype is beginning to cool. Here’s what we’ve actually learned

Experimenting with a new technology is as competitive as it is risky. Here's what I've learned about leveraging AI without jeopardizing my company's reputation or IP.

[Images: Adobe Stock / VRVIRUS]

Profile photo of Alex Atzberger

More than a year out from the initial release of ChatGPT, generative AI has found a permanent home in every corner of the business world. The immediate pressure to have a fresh response to those two letters cropping up in every pitch deck and LinkedIn post has, for the most part, subsided—but more than 12 months after the surge of generative AI, what have we really learned?

The frank truth is that, for a while now, hot takes have vastly outpaced practical innovation, and businesses are still not sure how to maximize the value of these new GenAI tools. 

This new reality presents a challenge for business leaders. The ChatGPTs and GitHub CoPilots of the world seem to be here to stay, but many companies have yet to determine how to effectively apply these tools to their business functions—or even how to think about applying them. It’s clear we need to think strategically but completely avoid overthinking.

As the CEO of a digital experience platform provider, I advocate for the use of structured experimentation when it comes to marketing decisions. In the last year, I’ve found a similar approach critical to determining what generative AI means for both my internal business and the solutions we provide to marketing and product leaders.

Here’s the experimentation-driven playbook we’ve followed for how to think about generative AI. While most companies probably start by thinking of ways to apply AI to their products and services, we found that the best way forward was to start with internal processes and then look outward.

DEVELOP A FRAMEWORK

Experimenting with a new technology is as competitive as it is risky. You’ll need to establish parameters that encourage usage throughout your organization, but that don’t jeopardize your reputation or intellectual property.

To set my team up for success, I find it helpful to ask the following questions, in order:

  1. How can generative AI help us improve our existing internal processes?
  2. How can generative AI help us improve our existing products and services?
  3. How can generative AI help us create new internal processes?
  4. How can generative AI help us offer new value to our customers?

These questions lay out a pathway from lowest to highest risk, along with the groundwork for an organizational AI policy outlining examples of good and bad usage and practical use cases for each function of our business.

IDENTIFY QUICK WINS

Once your framework is in place, you’re free to begin experimenting with generative AI and enjoying the tangible results.

To start, we sought out opportunities for AI-driven quick wins in our internal processes. While this lowest-risk-first approach is broadly applicable across organization types, it’s especially appropriate for B2B vendors, consultants, and anyone with customer relationships to prioritize and protect.

Your customers will be eager to enjoy the benefits of generative AI advances, but may not have the time or resources to upskill or hire in order to experiment fully and quickly. That’s where you come in; it should be your mission to bear the burden of risk for them. While it may take weeks, or even months, for a client to start an AI initiative of their own, they can enjoy the benefits of your experimentation, even with internal processes, immediately.

We found that our in-house improvements created efficiencies that served our customers. For example, we experimented with ChatGPT and Excel to build out more robust customer analytics that would have taken us months to develop beforehand, and have tested an AI-driven chatbot to automate our documentation processes.

Next, we looked to create additional value within our product offering—an effort that realizes ROI much more quickly than developing net-new solutions. We considered where embedded AI features might alleviate customer pain points, such as speeding up content development at scale, and began to incorporate these capabilities after thorough testing.

BUILD SOMETHING NEW

When everyone feels pressure to rush new AI-driven products to market, odds are many of those products won’t be up to market standards. Starting with quick wins improves those odds, and frees up the time and brainpower teams need to develop something truly innovative and viable.

At my organization, we’ve found that embedding generative AI into our existing roadmap is ultimately more beneficial to our team and our customers than investing resources in building expensive foundational models. But even if building from the ground up does make sense for your team, that step should come last, after you’ve had the chance to explore the risks and opportunities in depth. Your customers may not see the rigorous testing and learning that preceded your new AI offering, but they’ll benefit from it in spades.

Pretty soon—within a few weeks—it will stop making sense to measure the progress of generative AI with time. Instead, we’ll see real-world applications and adoption scale with blinding speed. Even more reason to get your proverbial ducks in a row now.


Alex Atzberger is the CEO of Optimizely, the leading digital experience platform provider