June 18, 2024

My Observations About Artificial Intelligence

On Monday, June 17, 2024, I spoke at the AI/COMM Conference co-hosted by Jeff Pulver and Joe Mazzeo. The conference was focused on the implications of artificial intelligence for the communications and cable industry, but its speakers covered a broader range of topics. It was a very well-designed and well-executed conference. Jeff Pulver is doing another one on October 29, 2024, in Westchester County, the details of which will be released on his www.Pulver.com web site. Anyone who wants to get a quick, deep primer on AI should attend the next conference.

I did a fireside chat about the implications of AI for businesses and other larger firms, although my remarks could apply more broadly. My main points were about how AI will change the people we hire and promote, and how we manage the change that AI will bring. These were my four main points:

1.`Making the Business Case Solely or Primarily on Cost Savings from Reducing Staff is an Invitation to Deep-Seated Resistance.

One of my MoveFlux business' first clients was an advertising agency whose President was very enthusiastic about how we could deliver generative AI to create marketing copy faster, of better quality and a great deal less expensive than the humans who were then currently creating it. After initial success, the marketing leaders stubbornly refused to continue using our platform. They were rebelling, just as the Hollywood writers did in striking against the studios.

The mistake the senior marketing leadership made in presenting the business case to the President was to position our work as a way of reducing the copy writer population. There were many other benefits, mostly relating to the ability to learn faster, improve response rates, and help clients secure more profitable revenues, but these were harder to quantify, so they were not used in the business case presentation. As challenging as it might be, those trying to promote AI need to identify and attempt to quantify benefits beyond staff reductions. As I will discuss below, they also need to describe the roadmap as to how employees can thrive in an environment in which AI becomes integrated into workflow.

  1. AI Demands and Rewards Different Skills From Those Traditionally Demanded and Rewarded.

Historically, when we source knowledge worker talent, we looked for people with broad and deep experience in a particular discipline. We also sought out people with particular academic credentials, such as multiple degrees from Ivy League schools or Stanford. We looked for so-called "experts" who had more "answers" and "knew" more than the rest of us relative to their knowledge domain.

AI flips this on its head. What AI demands and rewards is the ability to frame great questions and to do more out-of-the-box thinking about a topic. Yesterday, I gave an example from the pre-AI and pre-Internet era. When I became the head of Human Resources at Pitney Bowes in 1990, the CEO gave me these challenges relative to our employer-sponsored health program: Reduce healthcare cost increases, improve healthcare quality and access, make employees healthier, and improve employee satisfaction, even while we asked them to pay more.

These challenges seemed incompatible and would seem to have required trade-offs, according to the domain experts. But I applied what former Rotman Business School Dean Roger Martin referred to as "integrative thinking" in his book with that title. I framed the question in terms of how we could achieve these seemingly incompatible goals.

In the pre-AI era, I caught a lucky break: I received an audio tape of a lecture by Dr. John Wennberg, the lead researcher of the Dartmouth Atlas Project. Weinberg's research team had gathered data on over 300 Medicare Physician Service Areas. It analyzed spending and results. Its conclusion: there was no correlation between healthcare spending and health outcomes. In fact, in a few cases, low-spend areas produced better outcomes. That insight caused me to ask the question: "What were the low-spend, high performance regions doing differently that enabled them to achieve better results?"

Today, we have the ability with AI to ask these kinds of questions and get much quicker answers. We can frame problems more broadly and get answers that "experts" in a discipline might not identify. I did that recently in asking Chat GPT about how we might solve the problem of electric vehicle recharging in the environment with an insufficient number of vehicle charging stations. The response included two options related to "wireless magnetic inductive charging," a process by which cars are recharged as they travel over or remain on a road bed. These AI-generated answers are not definitive or determinative, but they point us to opportunities to think more expansively about problems and solutions.

People who get to high level positions because of mastery of a body of knowledge need to retool themselves to ask better questions and leave their comfort zones. Many are frightened to do so, especially if they have spent years getting to those higher-level positions. Most are too young to retire comfortably and support families, so they are often the most resistant to admitting AI into their toolkit. But it is a matter of time before their prior skills and experience will be progressively less valuable. They need to lead the way in adapting, not to stand in the barricades blocking change.

  1. AI Enables Us to Mine the Value of "Outlier" Data.

In my example of my inquiry about healthcare spending and health outcomes, we sought out the drivers of better health in the lower spending communities. As a self-insured employer, we were freer to explore different options for health improvement and healthcare cost reductions than those who delivered health insurance through the state-regulated insurance systems.

We learned that five initiatives take in combination could deliver all the benefits my CEO demanded of me:

  • Invest in prevention and wellness to keep healthy people healthy;
  • Provide onsite or near-site urgent and primary care;
  • Help employees and their families navigate the healthcare system in complex acute cases;
  • Redesign our insurance program to reward better healthcare and health-promoting employee behaviors; and
  • Create a health-promoting environment for employees with reduced stress, increased sense of purpose, better food, more access to physical activity and fitness, and more nurturing physical environments.

AI would have enabled us to get to these answers far sooner because it is particularly good at analyzing "outlier" data quickly.

We found a new business in a small piece of data as well. Quite a few of our customers needed emergency postage deposits at the end of months to get mailings completed. We charged a $25 fee. We made $6 million a year from these fees.

One of our executives, Lisa DeBois, suggested that this customers were signaling that we had a potentially much bigger business opportunity, to offer lines of credit to mailroom customers. We acquired an industrial loan bank charter, deposited our customers' advance postage deposits into their bank accounts, and lent out the money at credit card rates. It became a very successful $200 million revenue business.

AI is great at seeking out these small, weak signals that, if expanded, might be the next big business. It can identify, size, and analyze these opportunities while humans are asleep or otherwise engaged.

  1. AI Enables Those Who Use It to Migrate More Quickly to Higher-Value Tasks.

The marketers whose copy writer tasks are either eliminated or severely reduced have the ability to migrate to higher-value tasks. Ultimately, the goal of marketing is to help their internal or external clients achieve whatever goals they set for marketing campaigns. These goals usually include increasing the revenue-producing responses from outbound communications, improving customer loyalty and retention, and building the client's brand.

To build the most successful marketing campaigns, particularly for a new product or service, marketers often need to do multiple controlled tests, called "A/B testing" in marketing parlance, through which they do side-by-side testing of different messages, offerings, prices, channels, and trusted sources, and different timing of their outbound marketing activity. They analyze the results, and, over time, learn what works.

In the AI era, marketers can get this learning far faster. It took Pitney Bowes 2-3 years to develop the optimal marketing messages, prices, channels, and timing to market the Personal Post Office, a low-end product launched the year before I became Pitney Bowes' CEO. Had AI been available, the marketing copy production would have shortened, we could have done the A/B testing far faster, and we would have learned what worked best in months, possibly even in a few weeks.

The marketing executive, manager, and professional, as well as any other functional or line operational leader, has to think of AI as an accelerator that progressively eliminates lower-value tasks and makes the output from higher value tasks happen faster and more fruitfully. It challenges every functional and operational to migrate as quickly as possible to much higher value tasks, a progression that the medical profession describes as enabling professionals to "practice at the top of their license." The "top" of the license consists of the highest cognitive skills and tasks.

AI can do other tasks of potentially high value that are beyond what humans have the time to tackle today. One speaker talked about the ability of AI to gather, analyze and recap customer phone conversations, emails, text messages, customer reviews and other interactions, as well as what frontline field sales, service and administrative employees learn about customer preferences. While there are multiple regulatory hurdles to recording and using phone recordings, the potential to understand customer feedback is limitless. This is an opportunity that probably did not exist before AI. It will require employees to orchestrate the mining of this data and translate it into actionable responses.

Clearly, not as many people will be needed at any level, but every employee in a function will become far more marketable at the end of a process than at the beginning. This is the promise of AI.


I led many change management initiatives at Pitney Bowes. The most challenging task of a leader is to describe a future in which everyone can "win," even those not likely to have work to do inside particular organizations. This takes a lot of forethought and a great deal of empathy.

Regrettably, we have caused people around the world, particularly Americans, to define their self-worth by the credentials they have accumulated and the experience they have attained. AI is indifferent to the past. It rewards those who can continually reinvent themselves. Those who strive to learn and improve and work to get comfortable with AI will succeed.

This will be a daunting task because what we know or think we know about AI today will be obsolete tomorrow. What it can do will change and increase, as will the challenges it presents. Businesses will be in a metaphorical "arms race" of increasing speed. Governments will not be able to stop the "race," although they can slow it down and redirect it.

I am excited about the future because of what AI can do. It has frightening downsides, but, on balance, I believe it can be very positive. By how we frame questions and do analyses, we can solve seemingly intractable problems like the one my CEO gave 34 years ago relative to our employer-sponsored health program. The Dartmouth Atlas Project changed my life; AI can do that for many others.

Let's embrace its potential, even while we protect ourself against its threats!