From Steam Engine to Electric Motor: Why 95% of AI Strategies Are Stuck in 1881
Created on 2026-04-01 02:42
Published on 2026-04-01 02:49
In 1881, a factory owner in Manchester did something that seemed perfectly logical.
He had been running his entire operation off a single massive steam engine. Enormous rotating shafts ran through the building. Heavy belts connected every machine back to that one central power source. The floors had to be reinforced to carry the weight. Workers lost limbs to machinery that ran constantly, whether it was needed or not.
Then someone showed him an electric motor.
He did the obvious thing. He ripped out the steam engine and installed an electric motor in the exact same spot.
The result was almost nothing. Trivial efficiency gains. The factory still needed the massive shafts. The machines still had to be in the same positions. The floors still had to carry the same weight. He had spent a fortune to do the same thing with a different power source.
This is what 95% of organizations are doing with AI right now.
The Three Phases Nobody Talks About
Rory Sutherland, the behavioral economist and Vice Chairman of Ogilvy UK, laid out this analogy recently at The Drum’s Predictions event, and it stopped me cold. Because it perfectly describes the three phases of technology adoption I’ve been watching play out across APAC for the past two decades.
Phase 1: Same But Worse (and Cheaper)
This is where most organizations are today.
They take an existing process, a human call center, a quarterly report, a manual assessment, and replace the human component with AI. The stated goal is efficiency. The actual goal is headcount reduction.
The gains are trivial. Often, the quality gets worse. But the cost line looks good on the quarterly report, so everyone declares victory.
I watched this happen at a logistics company in Southeast Asia. They deployed an AI-powered routing system that cut delivery times by 30%. The COO was celebrated. The board was thrilled.
Within six months, driver turnover had doubled.
The AI’s “optimized” routes ignored everything that experienced drivers knew: roads that were dangerous in rainy season, customers who needed five extra minutes of conversation or they’d complain to headquarters, timing patterns around school zones and market days. The drivers didn’t leave because the math was wrong. They left because they felt erased.
AI accelerated whatever was already there. In this case, it accelerated a management approach that valued spreadsheet metrics over human judgment. The 30% efficiency gain was real. The 100% increase in turnover was also real. Nobody connected the two numbers because they lived on different budget lines.
This is Phase 1. Same but worse. And cheaper. Until it isn’t.
Klarna is the global poster child for this phase. They replaced 700 customer service agents with an AI chatbot, boasted about the savings, then quietly started rehiring humans twelve months later after customer satisfaction collapsed. Their CEO’s admission was remarkably honest: cost had been too predominant a factor. The result was lower quality.
55% of companies that executed AI-driven layoffs now regret them, according to Orgvue and Forrester research. That’s not a minority. That’s the majority. And the regret comes not from moral squeamishness but from hard business consequences: customer churn, reputation damage, and the expensive scramble to rebuild capabilities that were destroyed in the name of efficiency.
Phase 2: Same But Better
This is where a smaller group of organizations is heading. It’s better than Phase 1, but it still misses the real opportunity.
In Phase 2, you stop using AI purely to cut costs and start using it to improve the experience. The AI doesn’t replace the human. It augments the human. The customer service representative gets AI tools that surface relevant information during calls. The logistics driver gets AI-assisted routing that includes their local knowledge as an input. The quarterly report gets generated faster, freeing the analyst to spend time on interpretation rather than data assembly.
This is genuinely better. And for most organizations, getting to Phase 2 would be a significant achievement.
But there’s a problem. Phase 2 still operates within the same process architecture. You’re still running the factory with massive shafts and heavy belts. You’ve just replaced the steam engine with a more efficient motor.
The real gains don’t come from a more efficient motor. They come from redesigning the factory entirely.
Phase 3: Something Genuinely New
Here’s where the electrification analogy becomes powerful.
The breakthrough in manufacturing didn’t happen when factories replaced big steam engines with big electric motors. It happened when someone realized that, unlike steam engines, electric motors could be made small.
Instead of one giant motor running the whole factory through a system of shafts and belts, they put a small electric motor on every individual machine. Each machine could run independently. It could be turned off when not needed. It didn’t have to be in the same location as every other machine.
The factory floor could be completely redesigned. The entire production process could be rethought from the ground up.
It was only when they reinvented the process around the technology’s actual capabilities that the technology delivered its real benefits.
This is the phase that separates the 5% from the 95%.
I saw this play out in Singapore’s tourism industry. The Singapore Tourism Board didn’t use AI to generate their quarterly visitor reports faster. That would have been Phase 1. They didn’t use AI to make the reports prettier or more insightful. That would have been Phase 2.
They redesigned the entire process.
They built an AI platform called Stan that operates in real-time, providing hotels and attractions with visitor insights they can act on while the tourists are still in the country. The old process was: collect survey data quarterly, publish a report, hope hotels read it, wait for next quarter. By the time insights reached operators, the visitors were home.
The new process couldn’t have existed before AI. It’s not the old process done faster. It’s a fundamentally different way of managing tourism that was only possible because someone asked: “If we were designing this from scratch, knowing what AI can do, what would we build?”
Singapore earned S$23.9 billion in tourism revenue. Not by optimizing the old process. By inventing a new one.
Why Most Organizations Get Stuck in Phase 1
MIT’s Project NANDA research, published in July 2025, found that 95% of organizations are getting zero return from their AI investments. When they dug into why, the answer wasn’t technology quality or budget. It was approach.
The majority focused on the AI itself: the tools, the algorithms, the platforms. They layered AI on top of broken processes and wondered why nothing changed.
The minority, the 5%, focused on everything around the AI: leadership alignment, cultural readiness, process redesign, governance. They redesigned processes around AI capabilities rather than layering AI on top of workflows that were already broken.
There are structural reasons why most organizations get stuck in Phase 1.
The cost-reduction trap. Publicly traded companies face relentless pressure on quarterly earnings. The easiest way to justify AI spending is headcount reduction. It shows up immediately on the P&L. Process reinvention takes longer to show returns, and the returns are harder to attribute directly to the investment.
The consultant incentive problem. Most consulting firms work on gain-share agreements where they take a percentage of identified cost savings. Nobody gets paid a percentage of “we invented a fundamentally better process.” The incentive structure pushes every engagement toward Phase 1.
The imagination gap. Phase 1 requires no imagination. You look at what humans are doing, ask whether AI can do it cheaper, and implement. Phase 3 requires you to ask a question most organizations have never practiced asking: “If this process didn’t exist and we were designing it today, knowing what AI can do, what would we build?”
The J-Curve penalty. Research on the productivity J-Curve shows that organizations implementing AI into unprepared environments see productivity decline by up to 60 percentage points before any improvement appears. Most companies quit at month 3 when the numbers look terrible. Success comes at month 9. Phase 3 has a deeper, longer J-Curve than Phase 1 because process reinvention is inherently more disruptive than simple replacement. Most leadership teams don’t have the patience or the mandate to wait.
The APAC Advantage (and Constraint)
Here’s what makes this particularly relevant for APAC mid-market leaders.
MIT’s research found that mid-market organizations implement AI in 90 days while enterprises take nine months or longer. You don’t have the bureaucratic constraints that force large enterprises into Phase 1 by default. You can actually get to Phase 3.
But there’s a cultural constraint that works against you.
In much of APAC, risk-taking is career-threatening. The social cost of a failed experiment is higher than in Silicon Valley. This creates a powerful gravitational pull toward Phase 1, because Phase 1 is safe. It’s what the consultants recommend. It’s what the board expects. It’s what your competitor’s LinkedIn post announced.
Phase 3 requires experimentation. It requires leaders who understand AI well enough to envision new processes, not just approve cost reduction proposals. It requires a culture where people can try things that might not work.
Family businesses in APAC have a structural advantage here that they rarely recognize. They operate on longer time horizons than publicly traded companies. They can tolerate the J-Curve. They can invest in Phase 3 process reinvention because they’re not hostage to quarterly reporting.
It’s interesting that four out of five winners of the 2024 IPA advertising effectiveness awards were family-owned businesses. As Sutherland points out, you have to ask whether it’s only family-owned businesses that have the freedom to operate around different time scales. Short, medium, and long-term objectives all at once.
If your organization has that freedom, use it. It’s your single greatest advantage in the AI transition.
How to Know Which Phase You’re In
Be honest. Ask yourself these questions.
You’re in Phase 1 if: Your AI business case is primarily built on headcount reduction. You’re measuring success by cost savings and adoption rates. The underlying process hasn’t changed. You’re doing the same thing, just with fewer people (or cheaper people).
You’re in Phase 2 if: AI is augmenting your people rather than replacing them. You’re measuring experience improvements alongside efficiency. But you’re still operating within the same fundamental process architecture. The factory has a better motor, but the shafts are still there.
You’re in Phase 3 if: You’ve asked “What would we build if this process didn’t already exist?” and the answer looks fundamentally different from what you have today. You’re measuring outcomes that weren’t possible before. New workflows have emerged that couldn’t have existed without AI. Like Singapore’s real-time tourism platform, you’ve built something that isn’t the old process done faster. It’s a new process entirely.
Most organizations will tell you they’re in Phase 2. Most are in Phase 1 with better marketing.
The Path from 1 to 3
You don’t jump from Phase 1 to Phase 3. The path runs through preparation, and that preparation is what I call the Human Layer.
Leadership must go deep. Leaders can’t envision Phase 3 processes if they don’t understand what AI can actually do. Not “AI will transform everything,” but a working understanding of what’s possible, what’s not, and where the boundaries are shifting. I tell this story in detail in The Human Layer, but the short version is: I was the CMO at an enterprise AI company who avoided actually learning AI because I was too busy with marketing. The CEO called me the laziest person he’d ever met. He was right. I was trying to lead a transformation I didn’t understand. Phase 3 is impossible when leadership operates at that level.
Processes must be mapped before they can be reinvented. You cannot redesign what you don’t understand. Most organizations don’t have documented processes with sufficient detail for AI analysis. The Six Dimensions framework we use at AIR APAC weighs Process Maturity at 15% of overall readiness. Not because it’s less important, but because you have to get leadership, data, and skills right before process redesign becomes meaningful.
The Context Graph must exist. Phase 3 reinvention requires deep understanding of why processes work the way they do, not just what they do. The drivers at the logistics company knew things the AI didn’t. The experienced hotel concierge knows things no chatbot understands. That institutional knowledge, what I call the Context Graph, is the accumulated record of why decisions were made. Lose it in Phase 1 headcount cuts, and you’ve destroyed the raw material you need for Phase 3 reinvention.
This is the cruelest irony of the cost-reduction approach: Phase 1 destroys the very people and knowledge you need to reach Phase 3. Every experienced worker you let go takes a piece of the Context Graph with them. By the time you realize you need it, it’s gone.
The Ultimate Takeaway
Don’t use new technology to optimize an old process. Use it to invent a new one.
The factory owners who replaced a big steam engine with a big electric motor got trivial gains. The ones who realized small electric motors could power individual machines, in any configuration, turned off when not needed, placed wherever they made sense, reinvented manufacturing.
We are in the big electric motor phase of AI.
Most APAC organizations are using AI to do the same things, just cheaper. The real opportunity, and it’s closing, is to reinvent processes around what AI actually makes possible.
MIT found that the strategic positioning window closes between mid-2026 and early 2027. Organizations that reach Phase 3 in that window will build advantages that compound over time. Organizations still stuck in Phase 1 will find themselves competing against companies that aren’t just more efficient. They’re fundamentally different.
AI is the engine. You are the steering wheel. But the question isn’t how fast the engine runs. The question is whether you’re driving on the same road as everyone else, or building a new one.
The first step is knowing where your organization actually stands across the six dimensions that determine AI success. The AIR APAC Scorecard takes 15 minutes and gives you clarity on what the real gaps are, not the ones the vendors talk about.
Take it here: airapac.org/scorecard