When I talk to different people about AI right now, I don’t see one reaction, I see three.
Some are quietly panicking.
Some are a little too in love with it.
And some don’t see it clearly enough yet, so they push it to the side and wait for it to “stabilize”.
It’s that third camp that really worries me — especially when they’re the ones running the business.
Models outpace each other monthly. New tools launch daily, making it tough to keep pace. Prices shift in ways that make no sense. Vendors consolidate before we’ve even figured out the last round of changes — and in the middle of all this, your customers (and you, if we’re honest) are quietly teaching themselves to go to AI first and everything else second.
In this kind of environment, I don’t think the differentiator comes from “predicting the future”. It’s much less glamorous than that: keep moving, learn fast, and still make decisions while the ground refuses to stop shifting.
For me, the simplest way to think about this is: design for fear (because volatility isn’t going anywhere), measure excitement (because you need proof this is actually benefiting your customers), and keep moving.
What I don’t buy is the idea that you can sit this one out until it “matures”.
THE SCENE RIGHT NOW
The infrastructure spend is staggering — data centers, power, chips. There aren’t many idle GPUs sitting around waiting. Prices feel subsidized, which means we should brace ourselves for shocks. We’re basically building on discounted compute and pretending the bill will never come due.
On the capability side, there is real momentum around these agent-style systems for software and research work. They’re clearly better at chaining steps and calling tools than they were even a year ago. But they still need guardrails and human supervision. We’re not at reliable, fully autonomous agents yet — and that gap matters.
The early wins are also real — case studies are showing 40–50% improvements on the right, well-scoped tasks with people willing to adapt how they work. Recent work from MIT and Wharton points in the same direction: most genAI pilots still fail to create real ROI, while a smaller group of teams does well when they integrate tools into everyday workflows instead of adding on one big “AI feature.”
But let’s be honest about what’s still hard: accuracy issues, memory limitations, hallucinations, compliance headaches. These aren’t small problems. Adoption is happening at two very different speeds, and many first pilots stumble badly inside large organizations.
There’s a real gap now between products that are AI-native and ones that just have “AI glued on”. You can feel it the second you start using them. If the only place AI shows up is a shiny button on an unchanged journey, you haven’t transformed anything, you’ve just decorated it.
And perhaps most critically for those of us who’ve built businesses on content and search, user behavior is fundamentally shifting toward AI-powered answers. Traditional SEO traffic? It’s at risk. Welcome to the era of GEO (generative engine optimization).
WHAT ACTUALLY MATTERS
The teams I see winning are the ones who manage to turn all this volatility into something that actually builds up over time, instead of just scattering their efforts everywhere.
At Spring, we’ve spent almost two decades working with businesses on customer-focused digital experiences. And in the last few years, as AI has moved from buzzword to real work, I’ve learned to look for five things in teams that are serious about it. They:
- treat their data like capital — because it is
- own their distribution channels instead of renting them
- embed utility where it genuinely changes the customer’s day, not just because you “need an AI feature” somewhere on the roadmap
- steadily raise AI fluency — one individual, one team, one department at a time
- and, crucially, keep trust — and their brand promise — intact every step of the way
Notice what’s not on that list: internal AI labs with no ship dates, endless pilots that never touch a real customer, or roadmaps that live and die in slide decks.
Everything else comes down to the choices we make about what to do first, what to delay, and how much risk we’re willing to carry at each step.
Which customer journeys do we start with?
When do we invite agents into the flow instead of adding yet another “assistant” no one uses?
How fast can we learn without locking ourselves into decisions we’ll deeply regret six months from now?
The teams I’d bet on are the ones who figure out this rhythm — design for fear, measure excitement, and keep moving anyway. They’re not calmer because they’re sure of what’s coming next; they’re calmer because they’re clearer on how they’ll respond when it hits.
In the end, fear and excitement both make sense. The real risk is planning as if this wave has extra time to wait for you.
I also dig into this in more detail (in Arabic) on our podcast حكي بزنس — recent episode: بناء استراتيجية الذكاء الاصطناعي بين الخوف والتهور.
