Building an AI Portfolio
AI has risen to the top of every executive's agenda. Yet many leaders are asking the wrong question. Don't ask whether AI is worth investing in - ask instead how to get both quick results and long-term competitive advantage from it.
This guide will help you build an AI portfolio that works: you'll gain practical improvements to everyday operations, expand your business into new areas, and prepare for the disruptions and breakthroughs of the future.
Why You Need Both Safe Wins and Bold Bets
The executive's dilemma is familiar: quarterly results have to be delivered, yet at the same time you must invest in things that may only pay off years from now. With AI, this tension is heightened, because the opportunities are enormous while the resources and budget to act on them are often limited. So where should you place your AI bets?
Ultimately, the question isn't whether to adopt AI, but which problems it's right for in your specific company. That's why isolated experiments aren't enough. You need a comprehensive approach in which the risks are managed and the direction is clear.
Three Kinds of Initiatives for a Balanced Portfolio
A smart AI strategy doesn't put all its eggs in one basket. It splits investments across three categories, each with its own role.
The 70-20-10 Split
A well-balanced AI portfolio breaks down like this:
- 70% to Core improvements – safe bets that make existing operations more efficient
- 20% to Growth initiatives – new revenue models and added value for customers
- 10% to Breakthroughs – bold experiments where even a single success could transform an entire industry
These percentages are only a guideline. A growth company might split its portfolio 20-50-30, while a well-capitalized bank may settle for an 80-15-5 allocation. What matters is that the split is deliberate and aligned with your strategy.
Every initiative needs a clear business rationale: does it save money, drive growth, reduce risk, or improve the customer experience?
If It Does None of These, Why Are You Doing It?
Core Improvements: Everyday Efficiency Boosters (70%)
These initiatives focus on integrating ready-made AI solutions into existing ways of working. The answer might be off-the-shelf software, a fully custom-built solution, or anything in between.
Results show up in weeks and months, not years. These are safe bets that pay for themselves and, when successful, fund the other experiments.
“A practical example: JP Morgan's AI reviews loan agreements in seconds, where it used to take 360,000 working hours a year. Nothing revolutionary - the same work is simply done faster and more cheaply. This is a typical core project: proven technology, a clear benefit, and lessons learned along the way.
Core improvements also serve as learning platforms. The first one reveals data quality issues, the second tests change management, and the third builds on what came before.
It's worth starting with processes that involve a lot of repetition or costly steps. This is how you secure quick wins and strong examples for the business that build confidence to continue.
It's also worth prioritizing projects and initiatives that open several doors at once. When you build a proper customer data system, it doesn't just improve marketing - it enables personalization, prediction, and risk management across the whole company. Initiatives like these are worth doing first, even if the immediate return is modest.
Growth Initiatives: Builders of New Business (20%)
Growth initiatives create new business or add value to the existing one. The risk is moderate, but when they succeed they open up new markets and revenue models.
“A practical example: Netflix's recommendation engine didn't just make streaming more efficient (that would be a core improvement). It got people to watch more, switch to competitors less often, and pay more for their subscription. The AI made the service more addictive - this is a growth initiative that expands the business.
Success requires shared infrastructure behind your solutions, not isolated silos. Build a proper data platform that all your initiatives can use. Even though the upfront investment is larger, over the long run you save money, speed up development, and ensure consistency.
The technology and business sides also need to work closely together. It isn't enough for something to be technically possible - customers have to be willing to pay for it. Test your concepts first with a few key customers. Ideally, customers are involved as early as the design phase.
Breakthroughs: The Makers of the Future (10%)
Breakthrough initiatives mean both major risks and potentially enormous rewards. They might include autonomous systems, transformative AI applications, or entirely new business models. Most will fail, but a single hit can change the whole game.
“A practical example: Lovable, Europe's fastest-growing startup, built a service on top of existing language models that creates ready-to-use software from nothing more than a verbal description. This doesn't improve coding - it eliminates it entirely and brings software development within reach of far more people.
Act like an investor: many small bets, clear milestones, and the readiness to stop or change direction. The goal isn't a quick return, but learning and preparing for disruption in your field.
Consider an innovation lab or collaboration with startups and research institutions. This is how you gain access to top expertise without disrupting day-to-day business.
Why Traditional Metrics Don't Work for Breakthroughs
Different initiatives call for different kinds of evaluation and measurement. If you measure breakthrough initiatives by return on investment alone, they die at the outset. But if you don't measure core improvements properly, money goes to waste.
- Core improvements: Classic metrics such as payback period, net present value, and rate of return. Clear calculations and comparison against other investments.
- Growth initiatives: A combination of financial figures and strategic metrics, such as market share or customer satisfaction.
- Breakthroughs: Technical feasibility, learning value, and transformative potential. Let the engineers assess these, not just the business side.
Remember this: everywhere, prioritize initiatives that build a foundation for what's to come. Proper data pipelines, reusable AI models, and scalable systems pay for themselves many times over when they enable dozens of future initiatives.
How Do You Organize Your AI Initiatives?
Business units are good at core improvements, but they rarely give rise to breakthroughs. R&D, on the other hand, may be detached from business realities, yet that's where the most exciting ideas are born.
For breakthrough initiatives, it's worth setting up a dedicated unit with a direct line to leadership and the freedom to experiment. But keep the connection to the business, so an ivory tower doesn't form.
Core and growth initiatives belong with the business functions that benefit from them. This is how you ensure commitment, adoption, and funding. Without business ownership, all you get is "successful pilots" that never spread.
A model that works is a centralized center of excellence that supports the units in their own initiatives. You bring technical expertise and business understanding together.
From Strategy to Practice
An AI portfolio is like an investment portfolio: diversified, regularly rebalanced, and adjusted according to market conditions. With a deliberate structure, you get results today and secure your position tomorrow.
Start by mapping your current initiatives. Which are core improvements, which are growth initiatives, which are breakthroughs? How do they connect to your strategy? If an initiative doesn't fit any category or doesn't serve a clear objective, stop it.
Remember: every initiative has to teach you something. Along the way, you're building infrastructure, expertise, and processes for the future.
The path demands the courage to invest in the uncertain while still running everyday operations. But when the portfolio is balanced, governance is in order, and the metrics are right, you'll get both quick results and secure long-term change.
Is your AI strategy in order? Change starts with understanding where you are and knowing where you're going.
