And I say that as an AI enthusiast using it daily.
Financial Times recently released a segment that absolutely nails what I’ve been feeling.
Everywhere I look, companies are touting “AI-powered” this and “smart” that. Yet on the ground, the reality doesn’t match the marketing hype. As someone in the architecture, engineering, and construction (AEC) industry, I’ve seen firsthand that the AI rollout is not a smooth revolution, it’s a messy, frustrating grind.
The Hype vs. My Reality
Let’s be honest: AI is portrayed as a magic bullet. Vendors promise it will revolutionize everything, from design automation to project management.
CEOs drop the term “AI” in every earnings call to excite investors. But in day-to-day operations, the results often fall flat.
The FT segment backs this up with hard data, an MIT study found “95% of corporate Gen AI pilots fail”, meaning most AI pilot projects never get beyond experiments . Think about that: nearly all enterprise AI projects flop or stall out. Why? The reasons are decidedly un-magical: technical hurdles, unclear goals, and poor data infrastructure trip up most pilots. In short, we’re witnessing a huge gap between AI hype and reality.
My experience mirrors these findings. We pour time and money into pilot after pilot, only to hit walls. Systems don’t integrate with our legacy tools. Employees are wary and don’t trust the outputs. Teams aren’t sure how to fit AI into workflows. As the FT video put it, we’re “stuck in a chaos of overwhelming hype, genuine fear, and endless ‘AI for everything’ pitches,” all while still figuring out how to actually use these tools in real work. It’s not for lack of trying it’s that true adoption is a lot harder than the sales pitch makes it sound.
AI Projects Are Failing for a Reason
Why do so many AI initiatives stumble? From what I’ve seen (and what research confirms), it’s often a cultural and organizational challenge as much as a technical one. That MIT analysis revealed that companies are pouring billions into AI, yet most pilots stall out due to a combination of technical hurdles, unclear objectives, and inadequate data infrastructure. In fact, internal in-house AI efforts succeed less than 5% of the time, whereas partnering with outside AI vendors yields higher (2030%) success rates.
This rings true in the AEC world. Many construction and engineering firms experiment with AI (for example, using AI for predictive maintenance or design optioneering), but then abandon those pilots when they prove too cumbersome to integrate with old systems or unclear in ROI . Common roadblocks include:
- Data woes: Our industry data is often siloed, unstructured, or just too sparse. AI can’t learn from bad or scant data.
- Skills gap: We have brilliant engineers and architects, but few have AI/ML expertise. A recent global survey found 46% of construction firms cite lack of skilled personnel as a key barrier to AI adoption .
- Process integration: Implementing AI means changing how people work. That’s scary and disruptive. Employee resistance (even fear of job loss) is real . If folks don’t trust the AI or know how to use it, the tool gathers dust.
- Undefined goals: Too often, companies adopt AI because “everyone’s doing it”, not because of a clear business need. With no clear problem to solve or KPI to hit, the project flounders.
When 95% of pilots are failing to scale, it’s a flashing warning sign. Throwing AI at problems without a plan and foundation is a recipe for disappointment . I’ve learned that the hard way, and I’m evidently not alone.
What Companies Say vs What They Disclose
One particularly eye-opening tidbit from the video was how companies brag about AI in public, but sound far more cautious in official filings. It’s almost comical: the same execs who gush about “AI transforming our business” will quietly admit the risks in the fine print.
For instance, tech giants like Meta and Microsoft arguably the biggest AI cheerleaders strike a very different tone in their SEC 10-K reports. Meta’s latest annual filing bluntly warns: “There are significant risks involved in developing and deploying AI and there can be no assurance that the usage of AI will enhance our products or services or be beneficial to our business…” . Microsoft’s 10-K echoes this, noting “significant costs” being sunk into AI/cloud infrastructure and conceding it’s “uncertain” if users or revenue will materialize to justify it; if they fail to get enough usage, “we may not grow revenue in line with the […] investments.”
In other words, even the biggest players privately recognize that the AI payoff is far from guaranteed.
This contrast speaks volumes. Publicly, it’s all “AI is the future we’re all in!”; privately, it’s “Well, AI might not actually pan out as we hope.” And it’s not just a few companies. Fully 72% of S&P 500 companies now include at least one AI-related risk in their annual reports up from just 12% two years ago. Boards have rapidly shifted from treating AI as a cool experiment to treating it as a core business risk . They worry about things like regulatory uncertainties, huge expenditures that may never pay off, intellectual property pitfalls, and even reputational damage if AI goes awry . No wonder regulators are chirping in too: the SEC has cautioned companies to avoid “AI washing” i.e. don’t overhype or lie about your AI capabilities . (In fact, the SEC recently penalized some firms for falsely claiming they used advanced AI when they really didn’t .)
For those of us tired of the hype, it’s almost satisfying to see these admissions in print. It validates what we suspect: a lot of the AI emperor’s new clothes are still being stitched together.
The View from AEC: Low Adoption, High Hopes
In the architecture, engineering, and construction sector, the hype/reality gap is especially glaring. On one hand, optimism about AI’s potential is sky-high, a global survey found nearly 70% of project managers and quantity surveyors believe AI will help them deliver greater value in coming years. We hear constant talk about AI-driven design, AI in BIM (Building Information Modeling), autonomous equipment, you name it.
And yet, actual adoption remains very low in our field. That same survey revealed 45% of construction organizations report no current AI use at all, and only 1% have truly scaled AI across their projects. The rest are mostly dabbling in small pilots or still in “wait and see” mode. I can count on one hand the number of genuinely successful, AI-at-scale implementations I’ve witnessed in AEC. It’s telling that many contractors plan to invest in AI, but their actual usage hasn’t caught up with their aspirations . We’re essentially stuck in pilot purgatory, while the marketing brochures show a future that hasn’t arrived yet.
The reasons are familiar: lack of skilled people, poor data quality, and integration challenges are cited as the top hurdles in construction AI adoption . Sounds an awful lot like the enterprise story at large. We have tons of legacy processes and fragmented systems in construction, so plugging a fancy AI in is rarely plug-and-play. And our workforce spans all skill levels not everyone is ready to embrace a complex new AI tool without serious training and change management. In short, the AEC industry is a microcosm of the AI rollout mess: huge promise, plenty of hype, but very uneven traction on the ground.
Beyond the Hype: A Call for Pragmatism
So, where do we go from here? Despite my frustration with the hype, I’m not “running for the hills” or giving up on AI altogether.
In fact, I see this messy phase as a crucial first step. Every transformative tech in history from the internet to smartphones started out chaotic, unproven, and a little overhyped. The dot-com era had absurd hype too, yet out of that chaos emerged Amazon, Google, and the internet as we know it. Chaos and uncertainty are where real innovation eventually takes shape .
But the key word is “eventually.” In the meantime, I believe we need a more pragmatic, results-focused approach to AI in business. Less grandiose futurism, more concrete value. Here’s my take:
- Focus on actual problems: Instead of adopting AI because it’s trendy, let’s identify specific pain points where AI can truly help (e.g. automating a tedious task, improving bid accuracy by analyzing past project data, etc.). Start small and prove value.
- Get your data and process house in order: AI without good data is a non-starter. We need to invest in data quality, integration, and core systems before layering AI on top. Same with processes if a workflow is broken, throwing AI at it won’t fix it.
- Invest in people, not just tech: Perhaps the best point from the FT segment was that “The winners will not be the companies that spend the most, but the ones that build the most AI-enabled workforce.” . Amen to that. We need to train our teams, hire the right talent, and create a culture that truly understands how to use AI as a tool. An AI-savvy team will beat a giant AI budget, every time.
- Stay honest and realistic: As leaders, we should set realistic expectations. No more sugar-coating to the board that “AI will add $100M revenue next year” if we’re not sure. A healthy skepticism isn’t negativity; it’s good business. Acknowledge the risks and unknowns (like those 10-K risk sections do) and have a plan to address them.
Yes, the current AI frenzy is exhausting. I’m done chasing shiny AI hype for hype’s sake. But I remain cautiously optimistic.
In this messy rollout, I see a parallel to the early internet days, a lot of hype, a lot of flops, yet underlying progress inching forward. AI isn’t a magic wand, it’s a tool, one that requires the right foundations to actually work. My commitment, as a weary-but-hopeful executive, is to cut through the noise and focus on meaningful AI adoption: the kind that empowers our workforce and adds real value over time, not just headlines.