AI's big impact won't be at the bleeding edge, the majority of it will be at the bungling edge.

Three Companies, AI, a Wedding and a Funeral

At least for the next two to four years, the majority of AI impact will come from fixing and improving broken and bungled parts of the enterprise. Meanwhile, the firms that don't improve get mortally wounded, and many of them have their funerals.

Firms will begin to bifurcate into intelligent and not so much.

Intelligent in business is not an abstract concept it is these 10Ps. The Toby 10Ps that have evolved as AI has traveled on its rocket ship trajectory

A lot of bad process, product, and experience design, awful service chains, and antiquated tech approaches will get fixed with smart human and AI work. That is the bungling.

The impact of that is not new.

Back to the Future

In the 80s and 90s at FedEx we were obsessed with process and smart controls backed up by brilliant tech inside of the best service-positive, intentional cultures. The real reason was that by eliminating rework and dumb errors, you delivered absolutely, positively game-changing levels of service, and the best cost structure.

In business terms, far superior cost structure, amazing customer experience, and an intentional, positive culture that delivered discretionary effort, agile, empowered people at every level, and customer-first behaviors. I wasn’t the smartest person at B school but am pretty sure that is a winning combo.

I was there for 17 years. We won the Malcolm Baldrige Award. I won the first Golden Falcon award in International as a front-line staffer (fourth in the company). That's a brag; FedEx-ers will know. It gives me service cred.

We had very little bungling.

Excellent service wasn’t an accident it was leadership, intentional culture and maniacal process control discipline married to brilliant IT.

Bungling, Bad Process, Lack-of-Intelligence Opportunities are Everywhere.

Still today, there is a mountain of bungling that, with the right leadership and the new tech (AI), can be harvested to win your competitive battle. Said another way, that is to simply reduce cost structure, out-agile the enemy, and improve outcomes.

You will also see responses to that bungling from outside your company, with new entrants that will show up with no tech debt, no legacy blinders, better cost structure, superior agility, and better outcomes.

Simply, they will be 10Ps companies. Chunka Mui and John Sviokla helped in the early formation of the alliteration intelligence thing, back when it was 5Ps.

The 10P’s: Proactive, Predictive, Pattern Matched, Preventative, Permissioned, Process Performant, Peer Connected, Personalized, Precise, and Pragmatic

Think about it. These are the goals/targets you use to eliminate waste and rework. These are the things you do to optimize cost structure and customer experience/outcomes. You have to ‘be’ these to win. These have to be part of leadership vision, Product Development, and customer/experience thinking.

Thought experiment: You have a choice of which hospital you want to go to, a 10Ps one or the old one? Which would you prefer to work at, invest in, etc?

If you have any doubts about the attacks from new entrants, look at the track record and trajectory of Revolut and think about all they accomplished even with pre-Agentic AI. Imagine that with AI afterburners. Exponential customer acquisition, high compounding revenue engine, record-breaking profitability/efficiency, and deepened share of wallet with existing customer base up 66% in a year to $67B. Yes, 66%!

The Bungling This Week

Think about the application of the new AI tools attached to 10Ps leadership as you follow through this goat rodeo of a story.

So, I am on a Delta Air Lines flight south, heading to a very sad funeral. The Delta experience, app, priority desk, and lounge were all great. 08 12 2026 DL2397 into Atlanta in bumpy air, the pilot landed it with the impact of stepping off a small sidewalk (had to compliment that).

I am a million-mile, long-term United Airlines customer. I am on the Delta flight because when I called United to speak with a human about a multi-factor filter for what I needed on my flights, the machine said, ‘Due to being really busy, your wait time is 28 minutes’! I hung up.

It took me 1/3 of the wait time to book my flights with Delta. Turns out it was also cheaper. I wasn’t even shopping other airlines. United made me do that, re ‘bungling’.

Meanwhile, my wife was supposed to fly out yesterday to pick up her parents in South Carolina and then head into Atlanta. United  canceled her flight two hours ahead of flight time. The only alternative was to drive, which added 18 hours on the road to what is an arduous trip at a difficult time. That’s not the real bungling I want to share, but I think general experience with airlines tells us there is an intelligence upside to be harvested.

Has anyone ever landed early and there’s no gate crew so you just sit on the plane? One other tell is the habit of pushing away from the gate and then just sitting there so they can log an on-time departure, which in the real world is only on-time for the person who runs that report, not real customers. United is a serious firm, not some fly-by-night rickety outfit. If they bungle stuff…who else does?.

Another serious firm with excess bungling is IHG Hotels & Resorts , which has over one million rooms in 100+ countries with a smart multi-brand approach. I think they have 400k+ employees, which doesn’t seem to make sense, as that would be about two to three rooms per employee. Bungle?

I am a Platinum Elite IHG customer. I just needed to find a hotel at a good stop on the highway for my wife. That took 30 seconds with my AI helpers. So I went online with IHG, logged in, and filled in all the data, but it wouldn’t let me put the reservation in my spouse’s name. So, I called the hotel number, where they told me, ‘Yes, that’s a bug on the website.’ After 18 minutes on the phone, on and off hold, me giving them my credit card info 4 times- yep, four times, I am told, ‘The credit card machine seems not to be working again'. The call ends with, ‘We need you to fill out a credit card authorization form for the hotel (Lexington, Va.).’ What the …

‘Where is the form’ ? Oh someone will email it to you. So eventually, the form shows up, and it’s an MS Word doc asking for all of the credit card details and security codes, basically requiring me to fill it out, print it, and fax it. Yes, fax it, with my security code on it. In addition to all of that I spoke to someone in the Philippines three times.

I spent 20 minutes on the phone with someone at the next hotel (also IHG) where I met my spouse today just to ensure that if she arrived before me she could get into both rooms for us and her parents. It took 20 minutes and just ended with a personal promise and confirmation my spouse would show up with a credit card.

So those aren’t catastrophes. I will not need psychiatric help. But they are examples of very common bungling, of poorly designed process and controls married to siloed, subpar IT married to a leadership structure incentivized for the wrong things. Poor experience and negative cost structure impact are a bad combo. (Thanks Northwestern University - Kellogg School of Management )

My best estimate is a third of all businesses work like that top to bottom and and rest have some level of compartmentalized bungling. That is where AI, especially Agentic AI will make the difference. It will improve the outcomes and lower the cost structure. Unfortunately, a lot of that will happen as new entrants eat into the revenue/market share of slow, tech-legacy-intensive, non-agile incumbents.

The Economic Side Effect

All of that eliminates jobs. Happening broadly at a low level adds up to a serious economic/societal impact. As that shrinks the tax base via job elimination and wage compression , entitlements/government costs rise, deficits increase, bond risk goes up, servicing the national debt becomes more expensive, and then AI continues to accelerate, and that’s a flywheel.

We are not prepared. We are not preparing.

A Few More Labor Impact Warning Flares

I recently did an analysis on the shocking progress and activity at Huawei. I had my team do a similar piece of work when I was at Verizon, where I had a Bell Labs PhD genius and a gaggle of MBA smarties work on it. The one I did with three competing AI’s is better. My best estimate is it was 120x less hours to produce.

Again, with my little AI bot pals, I dug into the WSJ article on how 1/3 of last quarter's GDP growth was all AI, including finding the source data from the Fed and others. I then created a really comprehensive and easy-to-use dashboard that will get updated and build trend charts every quarter automatically. That took 8 minutes, literally 8 minutes. While tech-savvy, I would starve to death as a coder. Tehre was no coding. Thanks Anthropic.

If you don’t know what Harvey has done in extracting costly labor from the legal profession, you should. While less than 4 years old, it is used in 60+ countries, by a majority of the top 100 global law firms, and has a valuation of $15.5B for a reason. It only has 1300 employees. It is definitely a tech-plus-lawyers firm (deep context matters). 300 of those folks are lawyers. It looks like it cuts labor content by 60% – 90% across critical and complex areas. What are the analogs across the economy? How hard would it be for well-funded Harvey to replicate this to new domains?

In the next 3 years, robotics will follow the same amazing progress curve AI has, especially as we leave behind the ego-centric stupidity that a human body is a great design for a robot. Humans are intrinsically great at two things: making mistakes and generating complexity in groups. Robots need to be purpose-built. What does that do to labor? China already has 4 million miles of autonomous truck driving data in a database, improving autonomous truck driving. Truck driver is the top labor category in 28 states.

Ironically, today EVs require more labor to build than Internal Combustion Engine (ICE) vehicles, mostly because of battery construction. In China only, where 46% of vehicle sales are now EVs, they are slightly less labor-intensive due to the robots and un-bungled process they use at battery plants. All plants in automotive will continue to be more robotic and labor content will shrink. Battery tech will move along a tech-shaped curve while ICE tech will continue to trundle along on its industrial curve. Meanwhile, in three years the price in China of an EV will be 50% better than an ICE vehicle, with a lifetime operating cost advantage of around 40%. This will go global. It will vary around the efficacy and cost structure of power grids. Because of that, the economic advantage of a lower cost structure in transportation as EVs grow may actually be minimal or comparatively negative in the US, which is bad for global competition and jobs. If you compare the penetration and share of EVs in Addis Ababa and Phoenix, it is shocking. All on the back of cost (used BYDs from China) and operating cost. That is a peak over the horizon.

Is there really a lot of bungling?

Just in case you don’t think there’s a lot to improve in the bottom half of the performers in the economy think of ‘revenue flows’ and efficacy in our healthcare ‘system’, or education, or the pharma industry and its bonkers media spend, or weapons systems (Javelin costs $178k and an anti-tank, one way drone is $2.2 k) or big banking where Citi , Bank of America and JP Morgan Chase & Co spend $40B between them on IT a year. I am pretty sure 65%+ of that is just keeping the lights on. Revolut’s entire cost structure, including IT, is about a tenth of that. I have banked with Citi on four continents and am certain of two things. One, lots of bungling (some amazing keystone cop like stories), and that my self-interest is not a driving force in their approach to me; it's their self-interest. Ditto Chase; less experience there.

(P.S. After I posted this, I needed my Citibank physical address, bank code, and SWIFT number to put on a paper form for a big UK bank to wire me money out of my deceased father's account, having cleared local probate. It took a while to escape the truly awful AI speech bot to get a person (in the Philippines) who said she didn't know the answer but would get a supervisor. Twenty-seven minutes in, she told me she just gets cut off trying to get a supervisor, so she was going to transfer me to the wire transfer department. She transferred me, and I got a supervisor who got me the info 32 minutes in. Bungling. And yes, a paper form I had to mail to them.)

Best estimate is Revolut's Net Promoter Score is 72, and teh big banks are around 20 or less, with I think JPM up at 31.

On top of that, 44% of the economy consists of companies with 500 or fewer employees, receiving less attention from the software/tech industry and lacking large IT departments. Agentic AI will have massive positive impacts in that area, especially with new entrants. A great friend of mine just took 30% of the work out of his tiny firm with stuff he ‘coded’ using Anthropic. He cannot code.

I am all for fixing the bungling bits in business, but the labor and macroeconomic impacts need attention.

AI is moving faster than you think.

Kimi 3 (from China) is open-source and was just released. It has 2.8 trillion parameters, a Mixture-of-Experts architecture with 896 experts, a 1.25 TB download size, and is at the top of the leaderboard on coding.

Just two years ago, ‘massive’ Llama 3.1 was this huge open-weight model that astounded everyone. It was 7x smaller than Kimi 3. It had 8 experts, not 896. It sort of sucked at coding.

All of this- the speed of adoption and ease of productive use, especially at the small business end of the game- tells me there’s a 300-foot tsunami just over the horizon.

I asked a couple of AI’s to debate and come up with an analogy to help you feel the unprecedented speed of change in AI from Llama 3.1 to Kimi 3 in just two years. It gave me a truly awful one about the Pentagon; clearly, it has never visited there. A complex one about human biology and this one that surprised me:

The Car Analogy: Model T to a Mobile Smart-City Grid

·       Ford Model T or maybe an early 1990s Honda Civic to an autonomous, electric high-speed bullet train grid. It doesn’t just drive down the road; it dynamically shifts its track layout mid-journey depending on the weather, reading the entire terrain ahead simultaneously.

The Wedding

So to complete that movie reference and just one more reason to love FedEx in the 90s. Both my wife and I were there, me 17 and her 18 years, Dave Broncek offered her (then my girlfriend) the head of Customer Service in Canada as Tom Oliver talked to me about running our Spain operation.

Not being extremely dumb, I realized that geographical spread could ruin a relationship. I asked her to marry me, and in a moment of weakness, she said yes. We turned down both jobs. Dave later became president of FedEx, and Tom chairman of IHG. You have to love a firm that then offers us both expat positions, based in Hong Kong, to help lead a turnaround of our Asia , Pacific, and Middle East business. Even at Malcolm Baldrige Award-winning FedEx, there were years of work there, including fixing some bungling, a lot of bungling. My other job was running all the international customer automation program around the world, which was using tech to move people away from manual shipment work, lower their costs, and improve customer experience and outcomes. That whole sector was full of bungling. Pretty much a law in physics that manual stuff leads to bungling. Once done (six years), we were sent to the Latin America region to do the same, where, while ‘chapucear’ isn’t a word used much, there was a new level of bungling.

Again, if FedEx had a bunch of bungling, who didn’t?

The simple 10Ps approach, born out of great mentors/experience and leadership from the top at FedEx, will shine a light on bungling. Agentic AI, especially as it becomes process and controls savvy, will supercharge the fixing of it. About a decade ago I saw Fred Smith (legendary FedEx founder) at wedding, he really didn't remember me until I mentioned my pals Hazarika, Bullion, Stapleton and Carrasco. He asked me if the stuff we did in the 80s/90s was still new news to most companies. I said 'yes, very much so'.

To end on a really positive note today, as I upload this, it is the anniversary of our first date way back when in Memphis, TN.