What do foot-worn paths on grass have in common with AI?

Have you ever seen an expanse of grass with paths worn down to dirt where people walk to short cut the provided sidewalks? I was once told the Walt Disney Company looks for these patterns to understand how people move and where they need to re-think access, often replacing the grass with a new walkway. People find their own solution when one isn't provided, and that reality is a threat to the adoption of AI for customer services in the world of retail.

I've had the same issue crop up multiple times over the past year with Amazon, and all thanks to their use of AI. As happens on occaision, a package out for delivery from Amazon never arrived. Yesterday I went online to follow up and was given the option to cancel the order. I clicked the link and was dumped into the customer support Bot. I explained I wanted to cancel the item selected because it never arrived. Amazon's AI had other plans. Instead of cancelling the order, it informed me the item was en route and asked if I'd like help with anything else. Uh, yes. I want to cancel the item that never arrived. Instead of helping me, I was presented a list of recent purchases, excluding the one in question, and asked which one I wanted help with. Irritated, I started over again. This time the AI informed me the item that didn't arrive on the 19th would arrive on the 20th despite it being the 27th. 

Even more irritated at having my time wasted, I sought out a customer service agent via chat and was finally able to get cancel and get a refund for the item that never arrived. That the AI decided my probelm was solved initially with no input from me was irritating. That I had to fight the system only to give up and go to a humn was very irritating. But the ultimate irritation is this repeated "experience" when trying to use Amazon's return process as implemented. What I've learned is not to waste my time, their AI isn't ready for prime time and I'm not a beta tester.

But it's not only Amazon who's failing. Enter Target.

Our family loves the game Catana, so much so we needed another copy to keep here at the house instead of always chasing down our daughter to bring hers home on visits. My wife was notified that Target had it on sale for 20% off, an AI generated message. Awesome! She ordered it for pick up at our local store, but when the pick-up ready notification arrived, it was out of stock and she was asked to choose another store. A second option, about 15min further away, showed 6 in stock so we drove over. Nope, not a single one in stock. The explanation? Their AI can only see what was in stock recently, not real time. Wonderful. At that point she gave up on the garbage AI and got her refund, and we stopped at a third target on the way to dinner and picked up the game.

My third immutable law of AI is that AI always gives an answer, no matter how nonsensical. It's not enough to implement AI and pray; the customer experience has to matter too! Any organization who wants to make effective use of AI needs to recognize it's limitations, and until it's proven infalible, provide alternatives. 

Retailers would we wise not to expect people to continue using a broken process. Customers are too smart not to see the dirt paths criss-crossing the landscape.

My Three Immutable Laws of AI (from the 1990's)

I started in AI R&D in 1991 at Allen-Bradley followed by three years in the AI Development Group at Eaton-Cutler/Hammer. I learned three lessons that have proven to be immutable truths over the past 30 years:

1. AI is Biased - thankfully this is no longer an issue hidden on the underbelly of the beast, and the truth is that likely all algorithms are biased. There are many biases built into training data, including the decision of what training data to include and where to source it. Be aware that finding uniqueness, that diamond in the rough, is highly unlikely if the data set includes no prior examples.

2. AI Cannot Predict the Future - There's a specific bias many overlook: time. AI is biased entirely in favor of the past; it cannot predict the future. Instead, what we call "predictions" are really just the recognition of patterns from the past, extrapolated into a guess about the future. Keep in mind the hot hand fallacy because AI often doesn't. Probability itself is based on a biased assumption that we can predict the likelihood of future events, and being close, even often, is not the same as being right.

3. AI Always Gives an Answer - today we call it "hallucinations", but the problem with AI giving non-sensical answers is not new. AI can be trusted, until it can't, and there's no way to know in which realm the answer you're provided sits. Given my background in plant floor automation and embedded systems, you can understand how this was, and continues to be, a showstopper in many ways. 

AI matters. It’s helping us solve problems and gain efficiencies never realizable before. It's a great tool, but it's not a replacement for independent, critical thinking. And always be skeptical of any AI deployment where safety and security are paramount. Back in 1994, some of my AI research in fuzzy logic and neural networks was focused on self-driving vehicles. And we're still waiting.

Gold Mining or Equipment Rentals?

If you owned a gold mine, would your focus be on extracting the gold, or helping others extract the gold by renting them the mining equipment? Telcos would jump at the chance to provide equipment rentals.

After a decade and a half in the Telco world, in May of 2025 I escaped. I had joined the telco world with the belief that they would own the cloud. Who could possibly be better positioned in the burgeoning world of cloud computing than a company entrusted to transport data? I knew in my heart as a strategist that leadership was ready and understood the fatal flaw in the cloud strategies of Amazon, Microsoft and Google: large, regional data centers. In contrast, telcos had available infrastructure all the way down to the highly coveted last mile. Given the adage that compared to the cost of moving data, everything else is free; the opportunity was obvious. Able to put the compute and storage at the last mile, as I blogged about a decade ago, and telco's would rule the cloud.

Was I right about the opportunity? Yes. Edge computing as it's now called, proved it. Was I right about telco leadership seizing the opportunity? Not even close.

After stints at AT&T and Verizon I have learned one immutable truth: there is no innovation in telecommunications. That era ended with the forced divestiture of AT&T Labs during the famous 1980's DOJ breakup of Ma Bell. Executives at telcos have one idea they live and die by: maximize return on assets. It makes sense to a degree given the massive investment they've made in copper, fiber and spectrum. In the telecommunications world, innovation is outsourced. Ericson, Nokia, Cisco, Juniper, etc. What is sold as innovation is acutally integration. Telco's own the assets, everyone else owns the innovation. 

If you were involved in cloud between 2010 and 2015 you know that enterprise customers didn't trust public cloud and often railed against it. Customers wanted cloud, just not public cloud. They needed something more secure that hung off their backbone and enabled easier and faster integration with their partners (innovation that was outsourced again, this time to Equinix). And the truly visionary CIO's wanted a step further, to push process down into the network.  

While at AT&T, I navigated my way to the cloud computing product team and worked to influence their thinking by bringing an outsider point of view and the voice of our customers. I had financial services customers who envisioned pushing settlement down into the network, healthcare customers interested in performing image analysis at the edge, and one bank that wanted to buy up to 600,000 compute nodes. AT&T's NetBond product served an important tactical need, connecting customers to the public cloud providers they were starting to use. But the big opportunity was providing the secure, reliable, trusted compute and storage that Fortune 500 CIO's felt wasn't available in the cloud market. Whomever satisfied that need was poised to talk about the SDN and Edge Computing, capabilities that would threaten the hyperscalers. 

NetBond was the begining, and the end, of cloud at AT&T.

Despite providing a strong business case and making introductiosn to the founders of starups able to bring the needed capabilities into the network, there was no interest. On the surface, it's hard to find money for anything other than spectrum at a telco. But hidden under that argument there was an even better reason. While speaking at InterOp in Vegas I received a phone call from an AT&T SVP explaining my deal to sell hundreds of thousands of compute nodes was dead. I pushed for an explanation and was told the internal cost of compute was an order of magnitude greater than the hyperscalers. Gobsmacked, I asked how we made money on the customers we already had. We weren't, but luckily we'd sold less than 2k nodes. AT&T shut down it's cloud operations shortely thereafter. 

After a short stint at Tangoe, I arrived at Verizon, knowing their cloud story a bit better from previous experience. In a former life I'd been asked to partner on a plan to help Terremark avoid a second bankruptcy. I asked to be read into their strategy. Cloud computing. Yes, but what's the strategy? VMWare. No, that's a product, what's the strategy? No, that's the strategy. Well, I said in my not so humble opinion, they're destined to fail because VMWare is crazy expensive and will lock them into innovations a generation and build cycle behind the public cloud providers. Nobody listened. Verizon bought Terremark, then wrote it all off soon after I joined months before Verizon divested the Terremark assets and sold the business to IBM. What remained was Verizon's NetBond competitor, Secure Cloud Interconnect, with the same value prop: lock in the innovations of everyone else in exchange for locking in a portion of the transport; transport that will be devalued every year as customers demand lower and lower prices.

How could I have been so wrong? After significant soul searching, I realized what I saw was invisible to the generations of telco executives focused exclusively on transport. They talked in assets, whereas I talked in architectures. What I inherently understood escapated their notice: the company who owns the architecture owns the future. Telcos by nature don't believe in owning architecture, and asset owners don't innovate, they manage. 

I learned a valuable lesson painfully. Now I see the fingerprints of architecture vs assets everywhere. Why are car companies developing their own infotainment systems? Architecture. Why are streaming companies steaming to the precipice of their own demise? Assets. Why are the naysayers warning of an AI bubble? Assets.

If only I'd re-read my own articles on the evil of assets in 2012, or that economics always wins in 2013 about leveraging the assets of others. 

Now the telcos are in a race to the bottom of the cost curve, the eventual reality of every company who cannot differentiate the value of their product. Firings are the norm. AT&T fired 9.5k people in 2024, Verizon fired 13k people just before the holidays this year. Why? Their executives were sitting on a gold mine but were focused on the mining euipment.

Ugh