Are we behind on AI?
The question a board really asks about AI is short: are we behind? I have heard it in insurance, in energy, in retail. Someone opens a slide with competitor logos, someone else mentions a pilot, and then the chief executive asks the only question that matters to a board. Not what the technology can do. Whether the company is late.
The honest answer is that nobody can tell you, because being behind assumes one race with one finish line. There are three races running at the same time, and your budget is only the entry fee for one of them.
The day the market changed its mind
In January 2025, DeepSeek wiped nearly six hundred billion dollars off the value of Nvidia in a single day. One model release, one trading session. Boards that had spent six months building a business case watched the ground under that case move before the case was approved.
That is the technology curve. It moves on research, on chips that cost more than twenty-five thousand dollars each, on data centers priced at fifteen to twenty million dollars per megawatt. You do not steer it. You read it.
Spending like it is a revolution
The money going in is being budgeted as if everything changes next quarter. The Bank for International Settlements compared the buildup of roughly a trillion dollars in AI infrastructure with the railway bubble and the dotcom crash. Railways were real. The track stayed. The companies that laid it mostly went under.
Meanwhile the running cost sits somewhere most plans skip: inference, the actual answering of questions, takes eighty to ninety percent of the compute. The expensive part is the day after launch, every day.
A revolution has a date. An evolution has a slope, and you climb it with the people you already have.
The curve nobody else owns
Three curves move at once: technology, market and your own maturity. Only the third one belongs to you.
Technology moves whether you watch it or not. The market moves because customers change what they expect from you. Your maturity moves when your people, your data and your decision rights get better at using what exists. That is the curve where a budget turns into something.
Gartner found that organizations spent on average 1.9 million dollars on generative AI in 2024, and fewer than 30 percent of AI leaders said their chief executive was satisfied with the result. The spending happened. The climb did not.
Why the pilots stall
MIT looked at corporate AI pilots and found roughly 95 percent with no measurable effect on profit. I have sat in the review meetings behind numbers like that. The demo works. The model is fine. What is missing is the plumbing around it: who owns the decision the model makes, which system holds the customer data, what happens when the answer is wrong on a Friday afternoon.
Three things separate the pilots that survive:
- a named owner who carries the business outcome, not the experiment
- one customer journey end to end, instead of six departments with six tools
- a measure agreed before the build, in euros or in handling time
Answer the board question out loud
Tell your board you are behind on exactly one curve, and say which. That turns a vague anxiety into a plan with a slope and a date. The technology will keep surprising everyone in the room. Your own readiness is the part you can show progress on every quarter, and it is the part that decides what any future model is worth to you.
What you do on Monday
- Write down which of the three curves your current budget is actually buying, in one sentence
- Pick one customer journey and name the person who owns its outcome
- Agree the measure before the next build starts, and put a number on it
- Ask what your running cost looks like twelve months after launch, given that inference is eighty to ninety percent of compute
- Score your own maturity honestly, then set the next level as the target
The question your board asked deserves an answer with a slope on it. Chapter 2 of CX+AI works through the three curves and shows how to budget for the one you own, and the CX+AI Maturity Check gives you the score to bring to the next meeting. Bekijk het boek