How To Solve AI’s ROI Problem

We’re 18 months into the generative AI (GenAI) revolution and we still haven’t begun the inexorable slide down from peak expectations. Every time the GenAI movement appears to plateau, a new model with staggering capabilities such as OpenAI’s Sora is released, reinvigorating the hype.

Yet, enterprises are beginning to question exactly how and when they’ll start to see the ROI in AI business tools. This question is coming up more often in my conversations with executives. The cost of both publicly available and privately maintained AI solutions is proving to be higher than anticipated. Token, GPU and energy costs are high as demand exceeds supply.

As executives, we want the time, cash and human capital sunk into these technologies to produce measurable results. The problems is that with AI in the enterprise still being an immature technology, it isn’t clear what the correct use cases are or what the results should be. Even though we’re at or close to the peak of GenAI hype, it’s still too early to measure AI initiatives in terms of the hard ROI benefits they’re providing to the enterprise.

This isn’t a subversive attempt on my part to extend the runway; my company isn’t an AI business tools provider. But we’re deeply entrenched in the business of data, so we’re always engaged in conversations about how our customers can use their datasets to see AI results.

When considering an AI project, the first question that should be asked is: Is this really a use case that needs or can leverage AI? Not all AI use cases are necessarily going to generate hard ROI (i.e., cost saving or revenue increases). The key is to focus on soft ROI for initial AI projects, as this will steer your organization toward a sustainable, hard-ROI-oriented deployment of GenAI in the years ahead. Why? Let’s discuss.

You’ll get more out of AI.

People often make the mistake of looking at AI as a like-for-like replacement for an existing role. I’d advise against this. The introduction of large language models (LLMs) doesn’t signal the end of the technical writing industry. Sora doesn’t mean filmmakers will never work again. Code-generating models haven’t decimated the programming world, either. The like-for-like approach suggests that it’s possible to trade an LLM for a certain number of employees performing a given job or task, but AI is far more efficient when used to augment our capabilities as people. Instead of looking at which roles within your organization you might replace with AI, consider where and how these tools might be able to help your people work more productively.

AI has hard-to-quantify benefits.

The 2024 Artificial Intelligence Index from Stanford University reports that AI helps workers accomplish tasks faster and produce higher-quality work. This isn’t as easy to measure, but it’s the right way to initially think about putting AI technologies to work within your organization.

There are additional advantages to GenAI that you simply can’t quantify. Creatives can use LLMs as a sounding board for innovative ideas or digital brainstorming partners. Salespeople or technical engineers who need to generate an email or slide deck can enlist the help of an AI tool to produce a first draft. An LLM integrated with a retrieval-augmented generation (RAG) tool can help your employees search and access internal company knowledge faster. Translating these types of benefits into hard results is almost impossible. Yet, the tools are directly impacting the productivity of your employees.

You’ll ease the cultural shift.

According to the results of a 2023 Pew Research Center study, “52% of Americans are more concerned than excited about AI,” up from 38% the prior year. Yet, a 2023 PwC survey of about 54,000 global workers revealed that nearly one-third of employees believe AI will help them increase their productivity. More than one-fifth reported believing that AI will create new job opportunities. There’s indubitably a real divide in the world.

By emphasizing the augmentative aspects of AI, you can counter the fear and uncertainty. You’ll be equipping your employees with the tools to make them better at their jobs, not attempting to replace them with AI solutions. Educating and advertising this shift internally will be critical. You’ll need to demonstrate how AI is helping people within your organization do their job at a higher level.

You can improve the customer experience.

Any effective executive understands the value of listening to your customers. Deeply engaged and passionate customers can help you steer the evolution of your product or service. A better customer experience can’t necessarily be measured in a hard ROI sense, but as you integrate AI tools to gather more intelligence regarding customer sentiment and preferences, you can really grow and sharpen your understanding of what your customers want and need. This kind of information will not only help you improve the customer experience but also shape the development and quality of your products and services.

You’ll learn more about AI.

Finally, by focusing on soft ROI benefits, you can gain valuable AI experience and start to understand how and where AI solutions could be best deployed within your organization. Working with your employees, you can pinpoint the tasks within your workflow that might be better suited to AI-led automation. Then, you can identify viable hard ROI use cases. In manufacturing, automation can drive a more efficient production cycle, increase yields and reduce time to market when a single repetitive task is turned over to a machine. We should expect similar patterns in the office to what we’ve witnessed in the factory as specific tasks are replaced or augmented—not an entire department or role—and the organization becomes more efficient and productive.

This is the sort of ROI we should all be aiming for as enterprise executives and technology providers: more productive and efficient users contributing to a stronger organization by leveraging data and AI business tools to automate routine tasks so individuals can focus on higher-value contributions.