Change is hard

While substantial gains might come in the future from adopting new technologies such as machine intelligence, the reality is that in the short run change is substantially more expensive than stasis. Furthermore the majority of companies have zero incentives for leadership to undertake costly and risky change programs vs. delivering continuity and stability. Understanding the road to enterprise adoption of generative AI requires understanding these two challenges.

First a bit of an exploration of the cost/benefit analysis. Let's take a hypothetical finance department with 100 employees who do all of the usual finance tasks -- AR, AP, reporting, planning, payroll, taxes... Let's say the average fully loaded cost of an employee in this group is $100,000 (benefits, office space, computers...) so the total cost of the finance function for this hypothetical company is $10 million a year. What could a company afford to spend to transform this function using generative AI technology?

The first part of the business analysis would of course be the possible benefit case as this bounds the investment window. And for a new technology like generative AI there are no case studies to draw from which provide guidance (and reassurance) for a specific outcome. In circumstances like this companies may turn to corollary examples to infer potential benefits. RPA (robotic process automation) is one such example where finance departments have achieved cost reduction benefits between 20%-40%. Since many of the companies considering generative AI have already had some experience with deploying automation, likely they would evaluate this new technology as being additive -- perhaps getting them to a 40% savings from the 20% savings they have already achieved from automation. Thus the total annual benefit case might be just $2 million in such a calculation.

The next part of the business analysis is to evaluate the investment required -- the benefit case having provided guidance on possible investment levels. For my most successful automation client (they achieved hundreds of millions in automation savings) the filter they would apply to any automation project was whether the business would sign up to a 2:1 return on investment. So if the annual benefit as in the example above was $2 million, the cost to implement the automation could not exceed $1 million.

This leads to the first challenge of technology driven organizational change - the total cost of putting in place a comprehensive technology adoption program will far exceed what is justified by any single benefit case. Thus the only way that an organization can undertake programs which modestly improve productivity is by looking at the investment as a portfolio of changes across multiple functions in the organization. This is the path that my client took with automation -- start with 20 or 30 projects and with $25-30 million you can actually achieve the necessary scale in the competency and technology investment to deliver the target individual project benefits.

But committing this larger scale investment requires executive sponsorship, likely derailing other planned investments. How for example did American Airlines CEO C.R. Smith make the commitment of $40 million (in 1960s dollars!) to the creation of the Sabre reservations system? By electing not to purchase new aircraft which risked continued stable growth for the chance that a costly change program would deliver a very different future result. And by the mid-1960s, Sabre was handling 7,500 reservations an hour, cutting the time to process a reservation from an average of 90 minutes to several seconds. This turned out to be a very good bet by Smith.

But most executives are not willing to make this choice. I recently had the opportunity to have dinner with a group of top HR executives in order to discuss the future impact of AI on the workforce. When the question was asked, "in what time frame will AI make a meaningful change in our work environment," the common answer was a decade from now. And 100% of the individuals around the table did not anticipate working for their current employer by then, with some even expecting to be retired. Many executives simply do not expect AI to change the areas they are responsible for during their tenure, so why take a risk?

Even in cases where change may be imminent, we make decisions because we anticipate some benefit and we avoid decisions where there is risk. Advocating for a large financial investment in a risky technology will not necessarily lead to a fat bonus or big promotion. And worse, a failure of such a program might lead to the opposite, even termination. So why would Smith have made the bet he made on Sabre and what can we learn from that experience to inform today's decisions about generative AI?

In the 1960s airlines were experiencing a rapid increase in passenger numbers and flight stops due to the increasing affluence of the middle class in the United States. American was a beneficiary of the increased business but also one of the first to struggle with the clerical errors from manual reservations which led to empty seats on some planes and overbooking on others. It was the business risk from these inefficiencies which led Smith to seek a solution which required a technical innovation in the way that airlines booked reservations. Smith perhaps saw an existential risk to the future of American and as the CEO was the one executive who could make the decision to bet the company on this change program.

We are right at this "Crossing the Chasm" moment and it is worthwhile to recall In the context of Geoffrey Moore's theory, the analogy of "oil gushers" and "shale oil trapped value" representing the difference between a large, immediate market opportunity (like a gushing oil well) that is readily accessible to early adopters, and a potentially vast, untapped market (like shale oil) that requires significant technological innovation and market development to unlock its full value. Businesses are going to have a hard time adopting expensive incremental innovations like Microsoft Copilot which unlock those shale oil opportunities. Instead what I expect to see in 2025 is a number of "oil gushers" where a few businesses find, like Smith did, the large, immediate market opportunities which generative AI can help unlock. It is only after these larger opportunities prove the potential that we will see broad scale adoption for the "shale oil."