Three curves, and only one is yours
Hook (0:00)
The first thirty seconds set the whole episode, so read this exactly as written.
"In January 2025, a single model release from DeepSeek wiped nearly six hundred billion dollars off the value of Nvidia in one day. That is the market you are budgeting against. And while the technology curve and the market curve move without asking your permission, there is a third curve in this story. Your own maturity. That one is yours, and today we climb it."
Segmenten
Segment 1 · 00:30 · The three curves
- Open with the frame of the chapter: technology, market, and your own maturity all move at the same time. Two of them happen to you. One of them belongs to you.
- AI is an evolution that gets budgeted for like a revolution. That mismatch explains most of the frustration in boardrooms right now.
- Name the trap early: buying compute feels like progress because it shows up on an invoice within the quarter.
- Set expectations for the episode: hard numbers first, then what you own, then what you do this week.
Segment 2 · 02:15 · What the technology curve actually costs
- The hardware side is brutal and public. Chips at more than twenty five thousand dollars each. Data centres at fifteen to twenty million dollars per megawatt.
- Inference carries eighty to ninety percent of the compute load. The cost lands after launch, every single time a customer asks something, not once during training.
- The Bank for International Settlements put the buildout of roughly one trillion dollars next to the railway bubble and the dotcom crash. Both of those left infrastructure behind that somebody else used well.
- The lesson for a board: capacity gets cheaper and more abundant over time. Treat capability as the scarce thing you develop, and let the market handle the supply of compute.
Segment 3 · 04:30 · What the spend actually returns
- Gartner puts average generative AI spend in 2024 at 1.9 million dollars. Fewer than thirty percent of AI leaders say their CEO is satisfied with the result.
- MIT found that roughly ninety five percent of pilots produced no measurable effect on profit.
- Kevin's read: the money is real, the models work, and the return sits stuck at the point where the output has to enter a live customer journey.
- AI works as a multiplier on what you already have. A multiplier on zero returns zero, which is exactly what ninety five percent of those pilots measured.
- Concrete framing for the listener: the pilot that produces nothing is giving you a free diagnosis of your data, your process ownership and your decision speed.
Segment 4 · 06:45 · The curve you own
- The maturity ladder runs from level 1 to level 5, and every level has its own work. Nobody skips a rung by signing a bigger contract.
- The six blocks of the ecosystem tell you where you stand: your customer understanding, your data, your technology, your processes, your people and your governance. Score each one honestly.
- The wall opposite the door: the organisation builds a beautiful front end while the work behind it stays exactly the same. The customer walks in and hits the wall.
- The move from author to source. Your content stops being something a person reads and becomes something a machine answers from. That changes what you write, how you structure it, and who owns it.
- Monday morning: pick one customer journey that matters, score it across the six blocks, name one owner per block, and set one measurable outcome for this quarter.
Segment 5 · 09:00 · Close the loop
- Recap the arithmetic without repeating the whole episode: compute gets cheaper, expectation gets higher, and maturity moves at the speed you decide.
- Kevin's stance: spend less time forecasting the technology curve and more time raising the floor underneath it.
- The invitation for the next episode: the six blocks, one by one, with the questions to ask each owner.
Citaten uit het boek
AI does not transform your organisation. It exposes it.
Maturity is the one thing in this market you cannot buy. You climb it, rung by rung, starting this quarter.
Outro
Every listener can do one thing before the end of the week.
Take the customer journey that carries the most volume in your business. Score it from 1 to 5 on each of the six blocks and write the scores on one page. Bring that page to your next leadership meeting and ask which block gets funded first. That page is the start of your own curve.
Shownotes
- DeepSeek in January 2025 and the near six hundred billion dollars of Nvidia value that moved in a single day.
- The cost structure behind the technology curve: chips above twenty five thousand dollars, data centres at fifteen to twenty million dollars per megawatt, and inference at eighty to ninety percent of compute.
- The Bank for International Settlements on a buildout of roughly one trillion dollars, compared with the railway bubble and the dotcom crash.
- Gartner on average generative AI spend of 1.9 million dollars in 2024, with fewer than thirty percent of AI leaders reporting a satisfied CEO, and MIT on roughly ninety five percent of pilots without measurable profit effect.
- The six blocks of the ecosystem and the maturity ladder from level 1 to level 5, from chapter 2 of CX+AI, Customer Centric AI by Kevin Quint.
Full transcript
In January 2025, one model release moved an entire market. DeepSeek arrived. Nvidia lost close to six hundred billion dollars of value in a single day. That is the market you are writing your budget against. Two curves in this story move without asking you. The technology curve, and the market curve. There is a third curve. Your own maturity. That one belongs to you. Today we look at all three. And we spend most of our time on yours.
Start with the technology curve. The prices are public and they are steep. A single chip costs more than twenty five thousand dollars. A data centre costs fifteen to twenty million dollars per megawatt. Here is the number people miss. Inference is eighty to ninety percent of the compute. Inference is the moment a customer asks something. So the cost does not stop at launch. It starts there. Every question your customer asks carries a price. The Bank for International Settlements looked at the whole buildout. Roughly one trillion dollars. They put it beside the railway bubble. And beside the dotcom crash. Both of those left infrastructure behind. And other companies used that infrastructure brilliantly. So compute keeps getting cheaper and more abundant. Let the market handle the supply. Your scarce resource is capability.
Now the money inside your own company. Gartner looked at spend on generative AI in 2024. The average was about one point nine million dollars. Fewer than thirty percent of AI leaders say their CEO is satisfied. MIT looked at the pilots themselves. Roughly ninety five percent showed no measurable effect on profit. Read that carefully. The models work. The money is real. And the profit line stays flat.
Here is what I see in boardrooms. The pilot works in a demo. Then the output has to enter a live customer journey. And that is where it stops. AI behaves like a multiplier on what you already have. A multiplier on zero gives you zero. That is the arithmetic behind those ninety five percent. So the failed pilot gave you something valuable. A free diagnosis. It showed you your data, your ownership, and your decision speed. AI puts a bright light on the organisation you already run.
Let us talk about the curve you own. In the book I use a maturity ladder. Level one up to level five. Every level has its own work. Nobody skips a rung by signing a bigger contract. You climb, or you stay where you are. Six blocks tell you where you stand. Customer understanding. Data. Technology. Processes. People. Governance. Score each block from one to five. Be honest and be quick. One page is enough.
There is a pattern I keep running into. I call it the wall opposite the door. The company builds a beautiful new front end. A smart assistant, a fresh app, a faster chat. And the work behind it stays exactly the same. The customer walks through the door. And hits the wall. One more shift matters right now. You are moving from author to source. Your content used to be something a person read. Now it is something a machine answers from. That changes what you write. It changes how you structure it. And it changes who owns it.
So what do you do on Monday. Pick the customer journey with the most volume. Score it across the six blocks. Name one owner for each block. Then set one outcome you can measure this quarter. Something a customer would notice. Bring that single page to your leadership meeting. Ask one question out loud. Which block gets funded first. Compute keeps getting cheaper. Customer expectation keeps rising. Your maturity moves at the speed you decide. Maturity is the one thing here you cannot buy. You climb it. Start with that one page.