Spend twenty-five years in financial services and the discipline comes through in how you read technology strategy. It works the same way it does in investing. Diversification. You don’t put all your eggs in one basket, and you don’t put all your eggs on AI.
It’s the most useful reframe I’ve found for the current rush, because it turns the whole decision from a technology question into a risk question. I first heard it put that way in a conversation with a strategy executive on seeing the whole chessboard.
Think of your technology like an investment portfolio. A financial advisor who told you to put every dollar into one stock would be fired, no matter how good the stock looked. Yet that’s exactly what companies are doing with AI right now, pouring the entire budget into one bet as though the bet can’t go wrong. It can. And the way it goes wrong is the same way a concentrated portfolio always goes wrong.
Why is betting everything on AI a bad strategy?
Start with this: the asset is unstable. Today’s version of AI won’t be tomorrow’s version. That’s a bigger deal than it sounds. If you rebuild your entire operation around the current models, you’ve tied yourself to a technology that’s changing under your feet. Something else could evolve, a different approach could win, and now you’re holding a system built for a moment that’s passed. In finance you’d call that concentration risk. In technology it has no fancy name, but it’s the same trap.
Then add the failure modes. A single-vendor, single-technology stack has a single point of failure. One cyber attack, one security gap, one breach from our friends outside the borders, and all of it can collapse at once. Diversification is how you survive the day one piece fails, and in technology, on a long enough timeline, that day always comes.
What does diversification look like in practice?
I learned this in the least glamorous corner of technology, backups, and it saved me.
My backup strategy has three layers, three separate backups from three different vendors. None of them are expensive. If any one of them fails, I still have two. That’s diversification at the smallest scale, and it’s the reason I still have my life’s work.
I’ve taken more than 980,000 photographs over my life, along with thousands of videos and 113+ books’ worth of manuscripts. I nearly lost all of it once, years ago, when a disk drive died and I had to rebuild it by hand, a nightmare I never wanted to repeat. Recently a disk failed again. This time it was a shrug instead of a disaster, because the data lived in three places and I recovered it.
It took a few days, because it’s terabytes, but I got everything back. One vendor, one copy, and that story ends very differently. The principle scales all the way up. What protects a photo library protects an enterprise.
Does diversification mean avoiding AI?
No, and people get it wrong in the other direction too. Diversification is a reason to hold AI as one position in a broader portfolio instead of the whole portfolio.
Use it where it helps, which is real and worth using, as I’ve written about in Six Places AI Will Break Your Work. Just don’t tear out systems that work to bet everything on a technology that will look different in a year.
The organizations that come out ahead are the ones that think in five and ten year horizons, holding a mix, keeping what works, adding the new thing as a measured position instead of an all-in shove. The ones that lose are the ones chasing the crowd, ripping out functioning systems to slap AI on everything because everyone else is. That crowd behavior has a cost, and I traced why these bets fail in Why Most AI Rollouts Fail.
What is the vendor lock-in trap with AI?
There’s a second layer to this, and it’s about vendors as much as technology. Sticking to one vendor is its own concentration risk, and it’s always bad, whatever the vendor promises. When you build everything on one provider’s stack, you inherit their outages, their price hikes, their security holes, and their business decisions. If they change their terms or their direction, you have no move. My three-vendor backup rule is a vendor-diversification rule as much as a technology one, and the logic is identical at enterprise scale.
Vendors know this, and plenty of them design for it. Proprietary formats, exit fees and integrations that only talk to each other exist to make leaving painful. I’d treat any AI provider that makes it hard to take your data elsewhere as a risk to manage, and I’d keep a working alternative ready before I needed it.
This is what a transformation done right looks like.
It spreads the positions so no single failure can take down the whole operation, and nobody’s betting the company on one shiny thing. I’ve written about the order of operations that makes transformation work, starting with people before technology, in People, Process, Technology. The through-line is the same as the finance rule. Don’t concentrate. Spread the risk, keep what works, and treat every new technology, including AI, as one position in the portfolio.
I think the executives ripping out working systems to go all in on AI are gambling with money and jobs that belong to other people. When that bet goes bad, they won’t be the ones up at night rebuilding the data by hand.
There’s a fuller version of this argument, including a three-horizon read on where the AI infrastructure race is heading, in this conversation on seeing the whole chessboard. If you’re leading technology decisions and want to think about them as risk instead of hype, start with the Digital Transformation Hub, and if the lessons you’ve learned running real systems belong in a book, my writing services are built for exactly that.
