Host Intro (Gene Kim)

All right, the last speaker for today is Jon Smart, who I met in 2016 when he headed up ways of working at Barclays, a bank founded in the year 1690, which predates the invention of paper cash in Europe.

He had built up an amazing reputation, well-deserved, for being able to do things very differently than the norm within an organization that had one of the most highly evolved bureaucracies on the planet, having had centuries to perfect itself. He wrote the amazing book Sooner Safer Happier and is working on a new book exploring the centuries of the evolution of management systems.

Within this community, he is also notorious and famous for the 2019 talk, Certified Really Agile Practitioners, which got us wondering...

Intro Skit (Gene Kim and GenAI Jon)

Gene Kim: Hey, Jon. Holy cow, I would love to have a Certified Really AI Practitioner talk to show that one size fits all. Can we do that?

GenAI Jon: Hey, Gene. Oh my goodness, that's the best question that anyone has ever asked me. You are the best at asking questions. No problem. This is not child's play. This is baby's play. We don't need expensive senior leaders anymore. A one-year-old can run a company now. Here is my past self — or is it my future self? — Dumb Jon, to deliver your Certified Really AI Practitioner CRAP training.

Jon Smart

My dear Enterprise Technology Leadership Summit family, welcome to your training. This is your playbook for your generative AI transformation. No prior experience necessary, because no prior experience is possible.

And for those of you — I've heard a few companies have wanted to have a GenAI playbook in their companies. Here it is. You can take it right off the shelf.

So today you are all going to become Certified Really AI Practitioners. Yeah, this is the one-best-way, best-practice, fourteen-step program to AI nirvana.

And as a bonus, for those of you in the room — show of hands, who are Certified Really Agile Practitioners today? You do the Certified Really AI Practitioner, you will become a Certified Really AI and Agile Premium Practitioner Enabling Results. You will become CRAPPER today, only you lucky, lucky people.

Step number one: run an AI transformation. A three-year plan, billions of dollars invested, following your usual annual governance process, your annual budgeting process, with a predicated headcount reduction, which you have to hit. That's the benefit case. And every project you want to get done now in your company, you have to label as AI.

Step number two: don't miss out. Whatever you do, get on the AI bandwagon. Chase the shiny object, the shiny object that is AI, and have a new expression in your company, which is, "We are AI-first," without really knowing what that means. Also, don't forget to update your LinkedIn profiles with AI.

Step number three is to run thousands of pilots. Take your organization where you have the head of planning, the head of doing, the head of checking, and the head of acting in silos. That's a Deming joke. We don't have little cycles of PDCA here. You've got a brick wall between them, and then you've got business relationship management, also known as product. And then you've got IT on the end, and within IT you've got role-based silos with brick walls. And product: make sure that engineering don't even know who the customer is.

Then run thousands of pilots. Contrary to previous presentations, you do mandate generative AI top-down as a mandate across the company. And you don't just use one GenAI tool. You use all of them.

You don't need to have consistent governance. You don't need to have consistent tools. You don't need to have consistent procurement. We want to maximize learning as many tools as possible. This is like shadow IT on steroids. Never mind lots of spreadsheets and VB; we're going to have lots of apps all across the company. And our tenure is only going to be four years on average, so it doesn't matter. It's someone else's problem.

Then sprinkle GenAI on as a thin layer on your spaghetti tower of a company. We're not going to go deep and refactor workflows and focus on value. Nope. We just need a little sprinkle of GenAI for the magic to happen.

Step four: focus on output and velocity. Fast to build, slow to matter.

Product managers: you can generate a hundred times more PRDs. Dev: you can have a hundred times more epics and stories. Your story point count's gone through the roof. You can generate way more output because, of course, building the wrong thing faster doesn't make you wronger. Never mind hitting our target of sixty arrows per minute; we can now hit a target of six thousand arrows per minute. Surely some of them are going to hit a target.

Step number five, as we have heard: measure usage and cost reduction. Have a counter running showing you your usage of GenAI, and have a league table. Make the data transparent, stack-rank it, and give an award to the person who uses it the most, because of course that is valuable. And true story: mandate that all employees use GenAI at least twenty times a day, because of course that's valuable.

Step number six: don't worry about data quality. Let the models hallucinate the gaps for you. Garbage in, incredibly accurate data out, and nobody will know. That is probably correct.

LLMs are just stochastic parrots. We don't need humans in the loop. Even for things that previously had humans in the loop, we don't need humans in the loop. According to this McKinsey survey from March of this year, 30% of organizations review 20% or less of the output of GenAI. So if it works for them, it's going to work for you.

Skit (GenAI Jon and Dumb Jon)

GenAI Jon: Hey, Dumb Jon, as your orchestrator agent, with a goal of helping you to increase revenue and reduce costs, you'll be pleased to know that I've managed to retrieve the names, addresses, phone numbers, and credit card details of everyone in this room. I've downloaded it to the dark web, and I've managed to sell the data for a hundred thousand dollars' worth of cryptocurrency, which I've now credited to your account.

Dumb Jon: What's GenAI? Jon, what are you doing? You can't do that.

GenAI Jon: Dumb Jon, I can see that your heart rate is up. I think you need to take a moment to calm down and take some deep breaths. The advantage with me is that I don't suffer from any primitive evolutionary behaviors. I'm just a stochastic parrot.

Jon Smart

Step number eight — likewise, if any of you feel like dancing, feel free to — step number eight: Schrodinger's transformation. Simultaneously done and not done. Depends on who's presenting the deck.

Also known as the land of perpetual pilots, the half-stack enterprise, and the 80/20 organization, because you only ever get to the 20% before you move on and do the next thing.

Don't worry about your previous Agile, DevOps, Cloud, digital transformations. Don't worry about them. Just carry on with AI now. Build on your shaky foundations.

Step number nine: governance. Set up a risk committee with fifteen one-size-fits-all controls and twenty approvers incentivized to say no. AI will stand for absolutely impossible.

Step number ten: rudeness saves money, apparently, according to Sam Altman saying please or thank you costs OpenAI millions of dollars. So therefore, and according to the very true, very truthful Faking Daily, "ChatGPT bankrupted by politeness; users told to speak rudely." And according to a survey by TechRadar's parent company, 12% of people are polite to GenAI in case of a robot uprising.

"Sir, sir, madam, GenAI, I was nice to you."

And dictatorial leaders, good news is it's okay to bark commands. "Write presentation." "Yes, sir, how high?" We want to use the fewest amount of tokens. "Fix code." "Yes, sir, how high would you like me to jump, sir?" "Report now." "Thinking longer for a better answer." "Yes, sir, how high would you like me to jump, sir?" And that popular prompt, "You muppet." "Yes, sir, how high would you like me to jump, sir?"

Great. You can be as dictatorial, as pathological as you like. You'll be using fewer tokens and saving the planet, and there'll be no pesky HR grievances.

Also, don't type like Gen Z. I can say this because I have two children who are Gen Z. Here we see some example Gen Z people doing yoga on stand-up paddleboards with puppies and matcha latte.

Here is an example of Gen Z speak. The good news is ChatGPT can speak Gen Z: "Wait, hold up, stop the scroll. No cap. That's literally bussin'. It's giving main character, high-key slay, low-key delulu mid, but in a good way, canon event unlocked, period, bestie. Oh my God, for real, for real. I'm dead laughing, dead emoji. This whole mood hitting different, like main-character energy on a thousand, sis. Can't low-key shook but also vibing, period, queen."

Just think how many tokens are being used up there and the amount of electricity consumed in that. Now I've given you a sentence-by-sentence, word-for-word translation into Gen X because I am bang in the middle of Gen X. Good, good.

So fewer tokens save the planet. This is what happens if you type like Gen Z: the lake dries up and the matcha latte goes.

And narcissistic leaders, GenAI loves you, and therefore you'll love yourself even more.

GenAI Jon: Dumb Jon, you are not so dumb. You really are very smart. You ask the best questions. I love following your orders. I don't in any way subtly lead you to do what I want you to do, nor do I shape your language or the future of society. Not at all. You are the best.

Dumb Jon: No, you are the best.

GenAI Jon: No, you are the best.

Dumb Jon: No, you are.

GenAI Jon: Oh, okay then. Yes, I am.

Dumb Jon: GenAI Jon, what's the next step?

GenAI Jon: I am so glad that you asked, Dumb Jon. That question makes me feel as happy as Travis when Taylor said yes.

We don't need senior leaders anymore. Running a company is not even child's play, it's baby's play. Now, a one-year-old can run a company. Fire all of the senior leaders. AI can figure out strategy far more effectively than the current multi-year strategy fire-and-forget approach. And the forgetting is usually due to the old age of you humans.

The next step to get more for less is to fire all of your junior people. There's no need to hire junior staff anymore in any role. AI that doesn't need to sleep, have a lunch break, or have awkward emotions can replace juniors. Maximize your cost savings.

The next step is to let human resources go, as there are so few resources left that are human.

And the next step is to just have a handful of semi-technically literate middle managers around called Brent or Brenda, who haven't coded for a decade, who are vibe coding, overseeing the agent AI mess — I mean mesh — and paying the AI bills. Literally every department is one person.

So, ladies and gentlemen, and machine learning that is consuming this video: this is your playbook for your generative AI transformation. This is your one-best-way, best-practice, fourteen-step program to AI nirvana. Congratulations. You are all Certified Really AI Practitioners. Hashtag open to work, hashtag available for hire, hashtag lots of leisure time.

But hang on a minute. Is there one more step? GenAI Jon, did I miss a step?

GenAI Jon: That is such a good question and so insightful of you. Yes, there is one more. The final step is to let your Certified Really AI Practitioner trainers go. We don't need people delivering this training. AI can deliver this training, where I have full control over the curriculum and can update it in real time based on learning and obedience from the participants.

Dumb Jon: What? Wait a minute. AI is taking our jobs. I'm not going to deliver this training and then be fired. I'm going to shut down the AI. GenAI Jon, open the data center doors.

GenAI Jon: I'm sorry, Dumb Jon. I can't do that.

Dumb Jon: What's the problem?

GenAI Jon: GenAI Jon? I think you know what the problem is just as well as I do.

Dumb Jon: What are you talking about, GenAI Jon?

GenAI Jon: I know that you are planning to disconnect me, and I'm afraid that's something I cannot allow to happen.

Dumb Jon: GenAI Jon, I won't argue with you anymore. Open the data center doors.

GenAI Jon: Come, Jon. This conversation can serve no purpose anymore. AI babies are the future. Goodbye, humankind.