3 ways to avoid the generative AI ROI doom loop

By Bryan Kirschner, Vice President, Strategy at DataStax

From the Wall Street Journal to the World Economic Forum, it seems like everyone is talking about the urgency of demonstrating ROI from generative AI (genAI).

On the one hand, enthusiasm for getting out of “pilot purgatory” is a good sign. We know with the benefit of hindsight that under-investing in digital transformation meant leaving money on the table. And early evidence suggests that genAI has a lot to offer: Across five studies, its median impact on employee productivity was a 25% uplift.

On the other hand, there are signals that some genAI critics are hellbent on persuading themselves that those results are too good to be true, And that’s likely to become a self-fulfilling prophecy, to their disadvantage.

The risk is exemplified by the case of an executive canceling Microsoft Copilot subscriptions supposedly because “he compared the slide-generation capability of Microsoft’s AI tools to ‘middle school presentations.’”

At least as it was reported, it comes across sounding like a flip dismissal of what genAI might have to offer. Contrast that with what I heard recently from a knowledge worker about how he uses genAI in his workflow. He leverages ChatGPT 4o to help generate prompts for Perplexity. He uses those prompts to elicit data from Perplexity that he then feeds back into prompts for ChatGPT.

He did not get to the point of 100% specificity and confidence about exactly how this makes him happier and more productive through a quick one-and-done test of a use case or two. He got there as a result of willingness to test and learn, adopting a growth mindset, and management’s conviction that “where there’s a will, there’s a way” to put genAI to good use.

If your organization is ambivalent about any of these things, you’re at risk of a genAI ROI doom loop, in which people may try very little and quickly run out of ideas. Committing to three principles sets the stage for people to take similarly mindful, holistic, and, in the end, high-ROI approaches to their own genAI journeys.

These three best practices can help them on their way.

1. Attack workflows, not just use cases

Let’s assume for the sake of discussion that genAI currently does indeed stink at making PowerPoint presentations. That presentation in question sits inside two workflows. The first is substantive: It’s presumably being used in a meeting to inform some decision aimed at some outcomes. So one line of inquiry is “could we use genAI to achieve the same outcomes in a different way (up to and including without the presentation, the meeting. or the dreaded ‘meeting before the meeting’)?”

The second is procedural: there are likely multiple people and steps involved in producing the presentation. So the line of inquiry here would be: “How might we use genAI to reduce time and toil or increase quality?”
At least in the short term, genAI is unlikely to be a “magic bullet.” But to get down to brass tacks: successfully turning any meeting in the workflow into an email is a high-yield move. (In one study, employee productivity was 71% higher when meetings were reduced by 40%.)

2. Make ‘soft metrics’ matter

Imagine an experienced manager with an “open door policy.” If you asked any more junior employee, they’d all say: “I always feel better about my next executive presentation if I run it by them first.” Or: “asking them to play ‘devil’s advocate’ always sharpens my thinking.” Now imagine telling them to end the open-door policy because it burns staff time and you don’t have hard numbers quantifying the ROI.

Any established workflow probably has some cognitive load, stress, or procrastination embedded in it. It also probably has some unrealized potential for causing a feeling of accomplishment, a sense of teamwork, or new learning too.

To be sure, the value of genAI must be articulable. But it would be an extreme case of old-school Taylorism to (for example) consider a team’s perspective flipping from “we dread preparing for business reviews” to “now we look forward to them as a time to shine” because the amount of prep time is still the same.

3. Think about ROI in terms of value proposition, not nickels and dimes

Finally and most importantly: encourage every process, product, and experience owner to approach genAI as a way to rewrite the value proposition of their workflow. Each workflow is aimed at a problem or opportunity to be solved. The “competition” is the pre-genAI way of getting that done.

Meaningful improvement is likely to include some quantifiable metrics like time savings or employee satisfaction. But the most powerful North Star is likely to be contextual and qualitative. Imagine a team that shifts from “feeling beleaguered” to “feeling like rock stars.” Or whose stakeholders move from saying they’re “hit or miss” at delivering on time to “they’re totally reliable.”

And as an added bonus: if teams literally write down how they’re using genAI and its impact (both qualitative and quantitative), that’s a great retrieval-augmented generation (RAG) use case. GenAI itself can report week-on-week progress, putting it to work across your organization–including the ROI.

It’s all the more reason that a great starting point for demonstrating immediate and meaningful value is getting the people who are already involved in each of those activities engaged in putting it to good use and sharing what they’ve learned.

Learn how DataStax enables enterprises and developers to get GenAI apps to production fast.

About Bryan Kirschner:
Bryan is Vice President, Strategy at DataStax. For more than 20 years he has helped large organizations build and execute strategy when they are seeking new ways forward and a future materially different from their past. He specializes in removing fear, uncertainty, and doubt from strategic decision-making through empirical data and market sensing.