The universal hype over what generative AI can do has stimulated a widespread interest in their adoption in business organizations.
Generative AI encompasses tools rooted in generative models that use statistical probability to produce likely data sequences. Chatbots, for example, can provide probable answers by determining what string of words would statistically be most likely taken to be the correct response to the prompt.
4 Ways to Improve Your Generative AI Tool’s ROI
- Establish a way to measure the ROI for AI tools.
- Have your stakeholders onboard with AI adoption.
- Factor in human oversight for AI tools.
- Stay current on the latest AI lawsuits and regulations.
AI’s ability to interact with human agents and generate content that they accept as valid has many business applications. From customer support to marketing and design to being the selling point of many software products, generative AI seems to have taken industries by storm.
However, winning the race to be the first to release one sort of AI application or another might no longer be enough to justify this adoption. Companies are bound to ask increasingly more questions about AI’s return on investment (ROI) before jumping the gun to pour more money into its adoption.
Generative AI vs. ROI
Quickly implementing new AI solutions to capture the public’s attention and imagination works in itself as a marketing tool. These implementations, however, cannot come without considerable investments. As the novelty and hype settle, the economic conditions will force firms to think more and more about the return on these investments. Thus, the question is, can generative AI tools demonstrate substantial ROI fast enough to justify their adoption as viable?
In the five years before ChatGPT sprung onto the scene, the hype over AI seems to have plateaued along with the proportion of companies adopting it, which settled between 50 and 60 percent. The almost overwhelming wave of generative AI tool releases has driven a surge of investments in AI startups and adoption last year.
Goldman Sachs economists predict that although AI investments could approach around $200 billion globally, this will be before AI adoption can significantly boost productivity. Thus, companies that depend on immediate results might not be so eager to adopt AI solutions before scrutinizing how exactly they will affect the bottom line.
So far, the effects are limited, and these limitations come from the very nature of generative models. Guessing the apt responses based on statistical probability will inevitably lead to unpredictable, incorrect or nonsensical results at times.
In some cases, these figments of artificial imagination might be manageable; in others, this may not be cost-effective at all. For precision-dependent industries like medical and legal, generative AI raises as many risks today as it promises benefits for the future. These risks are also a matter of consideration when firms approximate AI’s financial potential.
How to Improve Generative AI’s ROI
In general, AI is widely portrayed as a powerful productivity booster for the future. Naturally, the firms that can take up financial and other risks and wait for that future have a clear advantage over those who can’t. Unfortunately, most can’t. Managers’ fiduciary duties force them to think about short-term ROI as much as the future.
Improving the ROI of AI adoption requires having a comprehensive strategy for implementing AI solutions rather than just going along with the trend.
First, this means establishing a way to actually measure the ROI of adopting AI solutions. Using common measurement frameworks for ROI has already been suggested for measuring AI adoption’s ROI in cybersecurity. One or a combination of these frameworks can be applied to grasp AI’s true effect in other fields, too. The optimal approach to measurement might differ from industry to industry, but having some settled methodology is a must to avoid the pitfall of biased assumptions and wishful thinking.
Another important piece of aligning AI deployment with ROI objectives is having your stakeholders on board. No matter how smart your solution is, it will be a loss of time and resources if people do not use it. This may require careful change management during the transition.
But not everything in AI adoption depends on how well firms manage it. In the case of generative AI, due to the aforementioned limitations of generative models, human oversight is still necessary for many of its applications. This necessity sets boundaries to how much this solution can really boost productivity and to what extent returns can reasonably be expected in a short period of time.
Another issue is reliability and cost feasibility of generative AI service providers. Their uptime has not been stellar during this past year, and many independent projects, despite being technically successful, had to close down since using them in production turned out to be too costly.
At the time of this writing, there are also multiple private and class-action suits against OpenAI, Microsoft and MidJourney by copyright owners whose materials have been used to train AI without permission or consent. This may result in stringent regulations regarding what can be and can’t be used for training, license fees and a significant cost increase overall.
Generative AI’s Potential Remains Untapped
Generative AI’s economic potential is still to be decided, but initial results are not promising for most companies that are not tech giants or have serious research and development budgets. The technology itself suffers from inherent problems that can’t be fully solved, only mitigated in certain cases. On top of this, it needs human supervision and there are serious questions about its future once the dust settles over the current legal battles.
Luckily, the field of AI is broad, and generative AI is just the most recent of its many subfields. Others include, for example, biologically inspired computation, federated learning, and causal AI. To uncover the full business potential of AI, companies might need to go beyond generative AI solutions and look increasingly more into the developments in these other subfields. These have been used successfully for decades for things like supply chain optimization, diagnostic systems and consumer products like the ABS on your car or the auto-focus on your digital camera.
While giant corporations investing in AI research is one thing, smaller firms deciding whether to invest in adopting solutions available today is another. While the former can afford to think about innovation and long-term ROI, the latter needs substantial results to hold on to today. As the hype over generative AI settles, it will be interesting to see how firms approach measuring AI’s ROI and making decisions about further adoption in light of these measurements.
Maybe when the next tech hype comes around, investors and CEOs will ignore the media and self-interested authority figures. Maybe, before making any big decision, they will speak instead to those who actually understand the field and don’t have skin in the game.
