The initial hope for Gen AI in businesses was that companies could apply Gen AI to most business functions and get a significant impact. But that is not turning out to be the case.
Many companies have experimented with Gen AI over the past two years, and we now have a better understanding of what the technology will allow companies to achieve.
It is now clear that there is a continuum of impact. At one end is a disruptive impact, which changes the business function to which Gen AI is applied. And the change is significant. At the other end of the continuum, there are incremental impacts; the technology makes only a modest change for the business function or people in it.
Now, with two years of experience of companies experimenting with Gen AI, it is clear that achieving a disruptive impact requires a very significant investment in both technology and operations.
Disruptive Returns
To understand the investments required to have a disruptive impact, consider the audit function as an example. This is an area where Gen AI is definitely having disruptive impacts. Large audit firms that deployed Gen AI find that their ability to conduct fraud detection – which is a key part of the audit function – improves dramatically by using Gen AI. In addition, it enables companies to automate or conduct audits with significantly fewer people than before using Gen AI.
In order to get there and realize those benefits, the audit firms had to make very significant investments both in technology and operations as well as in data. The audit function is an area that is well suited for the application of Gen AI because it has deep data. Audit firms have more than 30 years of really robust data that they can use to train their Gen AI models. Based on that, they are able to understand patterns they can use to quickly uncover fraud.
But, data work alone is not sufficient; they also needed to integrate the Gen AI technology into their existing technology. In addition, achieving the disruptive impact and improving the business function required a significant investment in operations. The audit firms had to retool the way they conduct audits, restructure the processes, and train the audit teams to use these new processes and new techniques.
The same situation of significant investments in technology and operations applies to other business areas that consider applying Gen AI. Consequently, the initial exuberance around Gen AI when it burst on the scene is now being replaced by realization that that to get the desired outcomes requires significant investments in technology and operations. There are relatively few instances where firms get to a disruptive or significant impact because of the high investment required.
It has not yet been determined whether or not all business function areas are, in fact, well suited for Gen AI, given the depth of investment required. We do not know yet whether the number of business functions that can be transformed through Gen AI is a large number or a small number. Currently, the number is small. For others to emerge, significant investments are necessary.
Disappointing Returns
What often happens is companies is that just put Gen AI over an existing set of data without the deep investments in data systems, the tech stack and operations. The result is disappointing. For the most part the result is only a very modest incremental impact.
Another problem is that Gen AI, even at a modest impact, is an expensive technology. Currently, for example, Copilot is $34 per month per person. Other Gen AI tools are priced the same, and there are also usage fees associated with them.
Many companies have disappointing results. They provide Gen AI tools broadly to their employees but do not see a significant uptick in productivity, effectiveness, or quality.
I recently spoke with an executive who said, “We provided Gen AI to our entire team. But we’ve not been able to reduce the headcount at all, and we don’t feel that the quality or quantity of work has materially changed.”
It is quite possible that, over time, the price of Gen AI will drop to the point where it makes sense for these tools to be widely available. But, the current pricing structure makes it unlikely that everyone will broadly adopt Gen AI because the return is not there, at least at this point.
Over the next 10 or more years, as companies incrementally work on the data and integration, it is possible we will see more instances of moving the value from incremental impacts towards a more disruptive and higher impact.
As prices come down, we will see broader adoption. But adoption currently, for the most part, is slow because many companies do not achieve the desired benefits either because the business function to which they apply AI is not well suited for it or because they have not been able to make the very substantial investments in data systems integration and operations necessary to release the value.