A CIO’s 5-step roadmap for scaling AI initiatives

For the past two years, CIOs have been told that AI is the latest and most disruptive innovation ever since the invention of the internet. And if they fail to immediately transform their work processes with AI, they run the risk of falling behind the competition and the current technology structure will fall like a shaky house of cards.

The FOMO is strong and many executives I speak to admit they have been or are currently being pushed into a race well before they’ve even been given a map. “We are afraid that competitors will use AI to their competitive advantage and we won’t,” the main idea lingers.

And so the budgets get allocated.

The tools are being purchased.

The tension is increasing and mounting. But nothing changes.

In fact, a report from Boston Consulting Group shows that over 74% of companies are struggling to scale the value of AI. Internal POCs quietly stall, budgets burn and the promised transformation is just not there.

So how do CIOs move from experimentation to meaningful business value? And how do they do it without drowning in frameworks, buzzwords and pressure?

The fear-driven trap

Technology is not — and should never be — the bottleneck. And neither is motivation. The actual bottleneck is the fear:

• The fear of leadership falling behind competitors (“We are racing to not look outdated”)
• The fear of employees that they will get fired because of AI (“I want to be enthusiastic, but I fear being replaced”)
• The fear of IT teams that their functioning workflows will get broken (“We just got our systems stable … introducing AI feels like opening Pandora’s Box”)
• The fear of everyone that investments will get used incorrectly (“Too many tools and not enough clarity. We’d better not spend millions on the wrong AI stack”)

AI has created an environment of fear: FOMO, fear of not being able to keep up and the fear of being fired because of AI. According to the recent research that explores findings from 700 CIOs, “79% of CIOs say businesses today have to take risks on emerging technologies or they will go the way of the dinosaurs.”

Fear leads to paralysis disguised as progress. The organization looks active — but it is not moving forward.

However, implementing AI is a relatively straightforward 5-step process:

Step 1: Don’t order, guide

Most people fear AI just like little kids fear monsters that supposedly live under their beds. But usually, fear disappears when discovery begins: just like, instead of shushing the kids and telling them to not to be afraid (which doesn’t help much, does it?), it is much better to spark their curiosity. Instead of ordering, it’s better to let them explore — and the same stands for employees, too.

Instead of saying “Everyone must learn AI” and “New mandatory workshops are introduced,” I’d invite everyone to embrace AI through habit and familiarity. For example, when I train teams, I start with aspects that could improve their personal life. I encourage employees to use AI:

• To plan what they will cook for dinner
• To organize their family’s schedules
• To learn or improve a foreign language

According to Microsoft, 75% of global knowledge workers are using AI and employees, struggling under the pace and volume of work, are bringing their own AI to work. Why? Because comfort builds confidence and once AI enters personal habits, people naturally bring it into work.

Step 2: Let employees discover use cases bottom-up

Many CIOs aim to identify every AI use case on their own. But it’s impossible. The people who actually know which workflows are repetitive, stressful or inefficient are not even execs; they are on the ground, doing the work.

From my experience, only a bottom-up engine drives real adoption. This way, employees feel in charge and the company surfaces the right use cases. Some improvements are incremental and may often even seem insignificant, such as 20 minutes saved on a teeny tiny task and 15 minutes on another. But multiplied across hundreds of employees, the gains compound. Culture builds. Momentum forms.

At Dyninno, for example, we’ve conducted an internal competition of AI ideas across the entire company to simply brainstorm and pick the brains of employees on what other processes we could employ that might bring us value. Employees who shared the best ideas were rewarded with a substantial monetary reward.

Step 3: Define the destination; allow the team to choose the path

Bottom-up energy is powerful. Boston Consulting Group, in fact, made a fantastic playbook that describes the top-down framework in detail. But without direction, it becomes chaotic — this is the exact stage where leadership should come in to provide focus.

The key here is to not overdo it. Choose very few, but the right few:

• Workflows that have high impact
• Repeatable processes
• Direct links to revenue or cost savings
• Clear owners
• Realistic timelines
• Success metrics that are defined early

This is where most organizations panic since they demand guarantees that it is going to work out perfectly. But AI doesn’t scale through certainty — it scales through disciplined experimentation. Think of AI as your personal student assistant, which is very brilliant, but also very inexperienced. Would you give your 19-year-old, fresh-out-of-college pal the company strategy to deal with? No. But would you ask the freshman to handle your schedules, read through numerous customer reviews, summarize research and draft first versions of presentations? Absolutely yes.

Remember — if AI doesn’t bring value, it was applied to the wrong process (workflow). To avoid this at Trevolution, we use quarterly planning cycles where we carefully select AI initiatives for development. We use a custom scoring framework that takes into account ease of implementation, monetary impact and strategic alignment. This allows us to select ideas to work on — something that employees prefer and identify as a potential bottleneck.

Step 4: Build the right technical foundations

Carried away by the popularity of the AI wave, many CIOs jump immediately into massive AI infrastructure spending. This is often premature. A smart AI infrastructure strategy follows this path:

1. Start with commercial & managed AI platforms
2. Check if your selected process is well defined and you have all the necessary data for it
3. Optimize and automate data pipelines only when value exists
4. Invest incrementally and only for proven use cases

You don’t build a factory before proving the product. AI is no different.

Step 5: Measure what matters

What do you call a person who counts the times he or she goes for a run but doesn’t care or measure any other inputs? An amateur. The same goes for those who work with AI without measuring the outcomes. The volume of your activities rarely has any significant impact, but the output does.

Many companies celebrate the number of pilots they launched or (even worse) hours of AI training delivered. These are nothing but vanity metrics, which in reality mean close to nothing.

The real questions one should ask:

• Did productivity increase? Can it be measured?
• Did operational cost drop?
• Did we shorten the output time?
• Are teams working smarter (not longer)?

Even a small AI initiative with actual impact is worth infinitely more than a bunch of experiments with results no one can assess.

AI won’t replace companies (but misused AI might)

While many consider AI to be a revolutionary development, I see it as a gradual advancement. Those with the largest workshops or eye-catching headlines won’t be the winners in the future for sure. The future winners will be those who employ AI on a daily basis to work more efficiently, where leadership scales what works and technology investment grows only once AI’s impact is proven.

First — the tool is introduced. Then people adopt it. Then workflows evolve and the business accelerates. Those who chase the hype will burn out. But capability builders will outlive everyone else.

This article is published as part of the Foundry Expert Contributor Network.
Want to join?