Originally published at Forbes
Alphabet CEO Sundar Pichai previously described AI as “more profound” than fire or electricity, so you’d be forgiven for feeling a touch underwhelmed by the day-to-day reality of the technology.
As the hype gets louder and louder, CEOs everywhere are issuing increasingly heroic statements about their AI investment plans. Oracle spending $5 billion and Salesforce spending $15 billion on AI investment are typical trailblazing headlines, dwarfed only by the even bigger visionaries, such as pharmaceutical giant Eli Lilly building its own AI supercomputers.
It’s impressive. It’s intimidating. And yet, the reality on the ground is that many companies, including global conglomerates, are watching from the sidelines while doing very little internally. I know. I’ve had that chat. And as a fintech entrepreneur with a portfolio of cutting-edge developments, you might expect me to shake my fist at “the dinosaurs” and demand revolution.
I’m not. I completely understand.
Risk Versus Reward
Among its many capabilities, the brilliance of AI, especially with agentic systems, is that you can now build a working software prototype almost as quickly as you can write the specification. Accountants everywhere have eagerly run the numbers on cutting development teams, streamlining processes and offloading administrative tasks to platoons of bots.
But most CEOs instinctively know the truth: Implementation can wreak havoc faster than you can generate an image of a dinosaur riding a dot-com bubble.
The issue begins with how corporate restructuring worked before AI, then understanding what has and hasn’t changed. Take a classic outsourced software team. Requirements gathering, client liaison, functional design and acceptance testing: These aren’t just steps on a checklist; they rely on conversation, nuance, trade-offs and human judgment. Yes, you can speed up admin and even accelerate coding, but proportionately, these are small slices of the total effort.
And large-scale change management? That’s still a people skill, with technology playing sidekick rather than lead actor. If you don’t bring staff with you, they will fight the implementation until the system is quietly retired to a metaphorical cupboard, wearing a smug “told you so” expression.
Ignore this at your peril.
The first wave of AI saw customer service divisions everywhere gleefully preparing to replace human agents with chatbots. Then one company’s bot went rogue, began swearing at customers and turned into a global cautionary tale. Euphemistically labeled by some as “teething problems,” it sent CEOs a very clear message: Delay big changes while simultaneously assuring investors you’re doing the opposite.
Startups, of course, enjoy a head start. They can design AI-led processes from scratch without legacy baggage. But that doesn’t mean big companies should sit back. There’s a practical middle ground every CEO can follow.
Practical Steps To Implement AI In Big Companies
1. Step the gatekeepers down.
Robust IT systems rely on central control, so the idea of letting Harry from accounts build his own customer invoice portal understandably makes IT teams nervous. But shutting down curiosity is the worst possible move.
Noncritical business processes are perfect for experimentation. Encourage teams to prototype, break things, fix them and occasionally create something brilliant. Host competitions. Run showcase events. Build a culture where small-scale AI wins bubble up. Your IT department becomes less the “department of no” and more the guardian of guardrails, ensuring no one strays beyond the boundaries of data protection, security or reputation.
2. Reimagine from the ground up.
The detail can smother even the boldest ideas, so encourage teams to reimagine their function as if they were a startup. Not a full-scale reorganization, just a change of perspective away from the daily grind.
These exercises often reveal the bones of bigger, more ambitious AI projects. More importantly, they involve staff in the process rather than making them feel the technology is being imposed on them.
3. Know that sensitivity matters.
A COO at a FTSE 100 firm described his board-level AI meetings to me as feeling like “turkeys voting for Christmas.” AI affects jobs at every level, and pretending otherwise is pure disrespect. Acknowledge it. Plan for it. Prepare for it. Reskilling, promotion pathways and humane exits were essential before AI and remain essential now.
4. Keep abreast of developments.
You need a team of users with both time and authority to experiment. What’s impossible today may be trivial next quarter. Progress in this field moves at a dizzying pace, and keeping up isn’t indulgence so much as survival.
5. Build your internal AI supply chain.
Most CEOs still treat AI like a single product rather than an ecosystem. In reality, successful AI adoption needs a supply chain every bit as structured as your physical or financial ones. Models depend on data. Data depends on access. Access depends on governance. And governance depends on people who understand what happens when you let an enthusiastic middle manager wire a large language model directly into a production database.
You need to create a clear internal route for AI ideas to move from spark to prototype to approval to production. Define who owns the data, who owns the risk, who signs off on deployment and, crucially, who monitors the thing when it steps into rogue chatbot territory.
This doesn’t slow innovation so much as prevent an AI free-for-all. It’s the scaffolding that stops small experiments from collapsing under their own success.
Your Next Steps
Some companies have already executed companywide AI shifts—some because their sector is easier to change; others because they invested early in this groundwork. Doing nothing is clearly not an option, while changing everything is rarely wise.
Taking these intermediary steps keeps your toes in the water without becoming an embarrassing headline in a year’s time.
Comments are closed.