As leadership teams around the world begin planning for 2025, the topic on everyone’s mind is when to expect their investments in AI and/or generative AI (GenAI) to pay off. New research from Google Cloud has revealed that more than 6 in 10 large (more than 100 employees) companies are using GenAI, and 74% are already seeing some sizable return on investment (ROI). But maximizing ROI from AI/GenAI requires a strategic approach that goes beyond justifying costs, encompassing both direct/indirect returns, a clear understanding of lead times and hidden expenses, and the integration of human-centric features to ensure reliable, scalable processes.
Reframing ROI
Given all the attention that AI/GenAI have gotten this past year in the media, it can be easy to forget that these investments are still relatively new, which means that most companies haven’t even started to see the sort of ROI that is possible. That makes it even more important to manage expectations in the boardroom from the beginning since any early evaluation will create critical impressions that will influence how leadership views future investments. If they have high hopes for immediate, transformative change, their opinion might sour if those changes are still taking root in the early stages. Put another way, new innovations demand new measurement perspectives, and leaders should reframe how they think about short and long-term ROI.
In terms of what constitutes a successful transformation, progress is often best measured in the eye of the beholder, but even “small” wins can lead to greater potential outcomes down the road. Here are three ways to help contextualize your AI/GenAI investments, as well as some examples from those on a similar journey.
1. Distinguish between direct & indirect ROI
In some industries, a direct ROI is easier to spot. For example, if a retail or CPG company begins offering new GenAI functionality, they will likely get an immediate sense from customers of how the features are being received. Whereas in other industries like manufacturing, there is more of an indirect ROI that is dependent on longer-term investments. With those sorts of soft returns, it is usually the “trickle-down impact” that can create new opportunities or unlock new value. Imagine that you’re implementing a new AI solution to improve team productivity. While your initial goal might have been output, that increase in activity could also lead to uncovering entirely new paths of growth that hadn’t even been considered. That’s the most exciting and exhilarating part about AI/GenAI – the unknown potential. And though the potential is tough to measure, it should always be included as a factor in calculating return.
A good illustration of both direct and indirect ROI can be found at the e-commerce company Mercari, which last year added a ChatGPT-powered shopping assistant to its marketplace platform for secondhand items. Their new “Merchant AI” would allow customers to “log onto the site, engage the shopping assistant in natural conversation, answer questions about their needs, and then receive a series of recommendations” for the next steps. The direct ROI of this was a 74% reduction in ticket volume at Mercari, while the indirect ROI was that the resulting time savings allowed the company to gradually reduce technical debt and scale its operations.
2. Factor in the lead time for AI/GenAI investments and the accompanying hidden costs
Considering the constant pressure on the C-Suite to grow profits, there is little chance of them suddenly adopting a “good things come to those who wait” mentality. But the reality is that any foray into AI/GenAI takes time and money, even before you reach the starting line. From investment in infrastructure and training to acquiring different APIs and relevant data, it can be months of prep work that won’t show any “return” other than being ready to begin. Another hidden cost (that a lot of people don’t talk about) is the reality that you’re going to get hallucinations and errors created by AI that can cost companies truckloads of money by sending them in the wrong direction, opening a loophole, or potentially triggering a costly PR problem. The whole experience is very new, which makes everything a bit riskier and more expensive, so it’s important for leaders to take this into consideration when evaluating ROI.
McKinsey offered insight into this decision-making process and its associated costs, riffing on the classic “rent, buy, or build” scenario. In their archetype, CIOs or CTOs should consider if they are a “Taker” (using publicly available LLMs with little customization), a “Shaper” (integrating models with owned data to get more customized results), or a “Maker” (building a bespoke model to address a discrete business case). Each archetype has its own costs that tech leaders will have to assess, from “Taker” costing upwards of $2 million, to “Maker” which can sometimes stretch to 100x that amount.
Endeavor to make investment in AI/GenAI more human-centric
There is still a lot of fear out there (especially among workers) that AI will replace humans. Rather than dismissing those concerns, companies should position any transformation as an enhancement instead of a replacement and try to look for ways to make their investment more human-centric. With GenAI, it’s not a transaction; it’s a partnership, and there is still a real need for humans to evaluate the efficacy of any generated insights or materials to ensure they are free of bias, hallucinations, or other misinterpretations. That’s why it’s critical that companies continuously challenge AI to provide rationale behind each decision to ensure accuracy. It will give the content more validation, your workers will see a defined role in the process, and it will ultimately help ROI because you’re learning at each stage.
It’s also a good idea to set firm guardrails to provide strict limits on what sort of information AI can gather. Ask yourself, “Should we allow the AI to have access to the internet?” Maybe not. The point is, to consider the need first, and if you have other proven methodologies, use those. Sometimes, AI is just useful for summarizing, not “thinking.” It’s all about creating the right balance, and humans still have a critical part to play. According to research from Accenture, 94% of executives feel that human interface technologies will let us better understand behaviors and intentions, transforming human-machine interaction.
Closing the Gap Between Promise and Reality
Experts agree that, while GenAI’s low barrier to entry is a great feature, its “long-term potential depends on evidencing its short-term value.” That means any AI/GenAI pilots should have a series of clearly defined (yet flexible) success criteria before they launch, and companies should constantly monitor processes to ensure they are continually providing value. When it comes to this new era of digital innovation, there might never be a traditional “finish line” we’re all racing towards. Instead, by changing how we think about the short and long-term ROI of AI/GenAI, companies can be savvier with their investment dollars and focus on developing capabilities that can scale alongside the business.
