If you’re an IT leader, there’s a good chance your company’s board, your CEO or even the market itself has told you: “We need an AI strategy.” For some, that translates into scrambling to bolt AI onto existing products, sometimes without a clear sense of why or how it will create value.
As someone who has spent years bringing products to market and leading teams through multiple technology waves, let me offer a dose of hard-earned reality: AI is not a strategy, and it’s certainly not a silver bullet.
The problem with ‘Just add AI’
There’s no denying the pressure to “do something with AI,” but the rush has set unrealistic expectations. For example, people assume that AI will instantly make development teams twice as productive, replace entire departments of employees or supercharge marketing overnight.
Here’s what is actually happening on the ground: We’re seeing a lot of experimentation, but as a recent MIT study showed, most projects don’t make it to production, and fewer still see use outside of a company’s own employee base. Success at most companies is typically concentrated among the use of a few productivity tools by those teams that can get the most leverage from them.
That’s not failure, and it’s certainly not surprising. From the adoption of electricity in the industrial revolution to the advent of personal computers and the World Wide Web, improvements at the margins has been the natural process when first working with new technologies. What’s less natural is the lack of patience that some leaders tell me they feel from leadership and the market.
It’s tempting to believe that you’ll flip a switch and suddenly have a next-generation, AI-powered business. The truth? AI is finding real wins for businesses as a powerful assistant: improving search, surfacing insights and automating repetitive tasks. But as a number of companies have already found out, it hasn’t (and likely won’t) serve as a plug-and-play human replacement across the board.
There’s also the matter of cost. AI isn’t free to experiment with, and it certainly isn’t cheap to run at scale. We’ve seen organizations implement AI in customer support only to discover that the workload was pricier than using human agents. The lesson: Unless you can draw a clear line to ROI, you’re not solving a problem, you’re adding one.
Where’s the ‘One AI to rule them all’?
Early on, many IT leaders assumed that one platform or model would pull ahead as the undisputed AI winner. Think of it as the “one ring to rule them all” fantasy. The reality check: Different AI tools and models work best in different contexts.
What we’re seeing is a market shifting to a more pragmatic stance: Embracing model-agnostic infrastructures that let companies mix, match and swap out models as needed. At Twilio, we adopted this approach from the start because it prioritizes flexibility. That’s what real builders, those who solve concrete customer problems, need.
Bespoke vs. bought: It’s all about focus
Early in the cloud era, many large companies raced to build everything in-house, convinced their needs were so unique that no vendor could serve them. Then reality set in.
We’re seeing the same learning curve with AI. Businesses are realizing that the true value is in customizing the experiences, workflows and data that are unique to them, not reinventing their entire tech stack. My advice to technical leaders: build things that truly differentiate you, and buy (or partner with) the infrastructure and platform providers that support it.
What IT leaders must do now
So, how can IT leaders ensure they’re steering their organizations in the right direction? It starts with clarity. Develop a solid strategy rooted in your customers’ real pain points. Ask, “Is AI actually the best tool for this job?” When it is, be transparent about the expected impact, risks and costs with your teams and company leadership.
Businesses that may be navigating significant tech debt should prioritize modernizing their tech stack to become more AI-ready. That includes structuring data and workflows so both humans and, ultimately, AI agents can interact seamlessly with your products and services. Finally, most importantly, expect speedbumps along the way and know that scaling globally will take time.
It’s only a race if you make it one
The AI transformation for most businesses will take longer than many expect. Upskilling takes time. Projects will fail. That’s not a sign to back out, but a call to learn, iterate and focus on where AI can truly drive impact. Let’s face it, sometimes the best move is knowing when not to use AI at all.
The most effective builders don’t just chase the latest trend. They use the best tools at the right time, for the right reasons. As leaders, it’s our job to set that example. Let’s move beyond buzzwords and get back to solving real customer problems and needs.
