Tony Christodoulou, Founder and CEO of Cyber Dexterity and Adjunct Faculty for GIBS Business School.
Business leaders are not seeing the expected results from adopting AI. Should they do something different?
To find answers, let's turn the clock back over a hundred years. Today, electricity is an indisputable revolutionary force that has reshaped the world. But around the start of the 20th century, several years after electricity emerged as a commodity, little had changed. Instead, many early electricity adopters ran into what we now call the "productivity paradox".
Despite using a technology superior to steam, they lagged. They didn't see the results. In an age when factories were designed around steam-powered driveshafts, merely introducing electric motors didn't produce enormous leaps in productivity. They needed to do more than just swap out some components.
Tony Christodoulou, Founder and CEO of Cyber Dexterity, compares this to today's AI adoption strategies: "The motor was the cheap part, and getting the AI licence is the easy part. But if your people are still working a certain way, thinking a certain way, you're not going to make the pivot. You will not have the transformation."
Giving everyone an AI licence won't make a massive impact when businesses narrowly focus on efficiency. While AI can make operations more efficient, that is an outcome, not an intent. Instead, cyber psychology emphasises that genuine change starts with human behaviour, and in the AI realm, that hinges on human amplification.
"Customers are coming to us to help them with AI; they're under pressure to show efficiencies," says Christodoulou. "Well, your first problem is that if you're telling me you've got to show efficiency, you've already decided the strategic intent. That's not a good thing, because then you might use AI to replace people when actually you're diminishing your tacit knowledge capability. You probably spend more on tokens and you create a facade of efficiency. That doesn't mean you can't get efficiency from AI. What's the actual strategy?"
Foisting AI onto people has limited benefits and more often creates issues, a pattern numerous studies have now revealed. MIT first broached the topic when it found that 95% of enterprise generative AI projects failed to generate measurable profit and loss results. McKinsey agreed, adding that the average large company loses $7.2 million per failed AI initiative. Fewer than a third of companies that adopted AI saw any measurable value, Deloitte reported.
Christodoulou says the problem is aimless execution. To fix this, rather than focusing on the tool, go back to the core pillars of any major business transformation: people, processes and data.
"People are the adopters. People ask the right questions. People bring value. Information and action are knowledge management. This is no different. Most organisations approach execution incorrectly. They don't take a step back and ask the right questions."
Reflecting on his customers' struggles and research into human/AI psychology, Christodoulou designed an AI adoption matrix, published in the upcoming book Digital Frontiers, a collection of insights from various technology transformation experts.
He calls this the Cyber-Dexterous Leadership Operating System, framed by four decision modes for an activity:
- Does the task require moral judgment, empathy, accountability or strategic context (such as people decisions, sensitive customer outcomes or crisis response)? Make the task human-led.
- Can AI improve analysis, options, speed, foresight or challenge assumptions (such as strategy, risk analysis, customer insight or policy planning)? Then the task is suited for AI-augmentation.
- Is the task repeatable, lower-risk, rules-based and monitored (such as workflow routing, alerts, service, triage or routine classification)? Such tasks can be AI-delegated.
- What if AI use would undermine trust, fairness, privacy or human dignity (such as sensitive personal data, high-risk ethical areas or opaque profiling)? Consider no AI or AI-restricted.
Much like electricity, AI adoption is a given for the absolute majority of industries. But it doesn't mean adoption is absolute. Business leaders should look at what will work best for their strategies by taking a methodical approach.
Christodoulou suggests using the above modes to start auditing decisions for a week (what did you choose versus what you could have chosen), actively naming the mode you select for new choices ("language shapes culture," he explains), defining boundaries before delegating (escalation triggers, sampling cadence and signals to pull back delegation), and creating an inventory of human knowledge before an AI-driven restructures ("before letting people go, ask what walks out of the door with them").
"Do it properly. Think of these things. Audit your own decisions and say, well, what's making sense here? You can't avoid the topic, because AI is here to stay. But don't just chase efficiency. This is a transformation that augments and amplifies humans. Start with that in mind. Efficiency is achieved as an outcome of doing the right things."