Article by Nadine Soyez

 

A few weeks ago, I joined a leadership workshop at a large European organisation that had been “experimenting with AI for over a year.” I expected to see a clear overview of their progress: maybe a dashboard, a structured report, or at least a central document. Instead, they handed me a stack of documents in their SharePoint: spreadsheets, screenshots, meeting notes, and half-finished pilot descriptions collected from various teams.

“Honestly,” he said, “we don’t even know how many AI initiatives we have right now.”

As we went through the documents together, the situation became clearer. There were far more AI experiments and pilots than anyone expected, ranging from small reporting automations to chatbot ideas, data exploration projects, tool trials, and analytics experiments. Many were duplicates. Several had stalled without anyone noticing. Others looked promising but had no owner, no sponsor, and no defined next step. Everyone was doing something with AI. But no one could say what mattered, what should stop, or what was ready to scale.

Experiences like this have shown me how common the experiments and pilot trap truly is and why so many organisations struggle to turn experimentation into real business results. The patterns repeat everywhere: regardless of industry, size, or maturity. That is why I wrote this newsletter. I want to show you why experiments with pilots don’t scale, how to shift toward an AI portfolio, and how to use a practical AI Portfolio & Scaling Tracker Template to bring structure, alignment, and real business impact.

 

1. Why experiments and pilots don’t scale

Most organisations begin their AI journey with enthusiasm. A team tries out a new tool. Someone builds a quick automation. A colleague runs a small experiment. At first, this feels like progress because activity is happening across the organisation. But pilots and experiments rarely translate into measurable results because they are not designed to scale.

Pilots are temporary experiments, not end-to-end solutions. They often lack a strategic purpose, a sponsor, and a defined path for what should happen if the idea works. Teams continue exploring, but no one decides whether the pilot should move forward, be redesigned, or be stopped.

Another challenge is fragmentation. Pilots and experiments run in isolation, often without visibility across business units. Different teams explore similar ideas without realising they are duplicating work. Knowledge stays trapped in silos, and there is no central system where leaders can compare initiatives or allocate resources strategically. Pilots also tend to ignore practical constraints. A proof of concept might work under idealised conditions — using a simplified dataset, a manual workaround, or a temporary integration — but scaling requires robust data pipelines, governance, user adoption, and cross-functional collaboration. Without these foundations, even promising pilots stall before reaching production.

Finally, most organisations lack a mechanism for making decisions. They don’t have a standard process for evaluating pilots, prioritising use cases, or determining which initiatives should be scaled. This creates uncertainty, slows down progress, and leads to “pilot fatigue,” where teams feel they are always experimenting but never advancing.

 

2. What an AI Portfolio is and how to evaluate it

An AI portfolio is much more than a list of use cases. It is a strategic management system that brings every AI initiative — from first ideas to production-ready solutions — into one coherent framework. Most organisations underestimate how powerful this shift is. When you move from isolated experiments to a portfolio approach, AI stops being something “individual teams are doing” and becomes something the organisation can actually steer, measure, and scale.

The purpose of an AI portfolio is simple: to replace fragmentation with clarity, and to replace experimentation with direction. When everything is visible in one place, leaders can finally understand where value is emerging, where risks are rising, and where the organisation should invest its time, budget, and attention.

 

2.1 Visibility: seeing what you actually have

In almost every organisation I work with, the first breakthrough comes when we gather all initiatives into one view. Leaders are often surprised — not only by how many isolated pilots exist, but also by how many of them are duplicates or disconnected from strategy.

A portfolio gives you real visibility. You see:

• where different teams are solving the same problem multiple ways,
• which initiatives are stuck or abandoned,
• which ideas have potential,
• which pilots consume time without progress, and
• where the organisation unintentionally created “shadow AI.”

This visibility alone creates a level of clarity most organisations have never had.

 

2.2 Strategic alignment: connecting AI to business outcomes

An AI portfolio also forces a fundamental question:

Why are we doing this?

Every use case must be directly linked to a business outcome. Not a trend. Not curiosity. Not “we should try this tool.” When organisations use a portfolio, use cases become strategic investments rather than random experiments. If a use case does not deliver measurable business value — cost reduction, improved quality, reduced risk, or increased revenue — it should not move forward. This alignment ensures that AI supports the business, not the other way around.

 

2.3 A clear scaling path: turning ideas into outcomes

One of the biggest advantages of an AI portfolio is that it defines how initiatives progress. Instead of asking “What should we do next?” the organisation has a standard pipeline:

Idea → Pilot → MVP → Production → Scaling

Each stage has clear requirements:

• data readiness
• technical integration
• user testing
• governance steps
• training and onboarding
• assigned owners and sponsors

This structure removes ambiguity and prevents endless pilots.

 

2.4 Practical governance: enabling speed, not slowing it down

Governance often has a reputation for slowing everything down. But with a portfolio approach, governance becomes a built-in support system. Each use case receives a risk level and predefined guardrails. Teams know exactly which controls apply, which approvals are needed, and which risks matter. This clarity accelerates execution rather than blocking it.

 

 

 

3. Consistent evaluation: the five-dimension model

To prioritise effectively, organisations need a consistent evaluation model. When every use case uses different criteria, decisions become slow, political, and subjective. That is why I use a five-dimension evaluation framework with clients across Europe. It allows organisations to compare very different AI initiatives objectively and decide what moves forward.

1. Business Value: Does it solve a real problem or create measurable benefit? Time savings, accuracy, cost savings, reduced workload, or financial outcomes.
2. Data Readiness: Great ideas fail without data. Many pilots work only because they rely on small, manually prepared datasets. Data readiness evaluates whether the organisation can actually support the use case in real operations.
3. Feasibility: Some use cases sound exciting but require integrations or infrastructure the organisation does not have. Feasibility ensures focus on what is realistic rather than hypothetical.
4. Governance & Risks: Every use case has risks: privacy, compliance, security, transparency, oversight. Governance should be practical and predictable — built into the process, not added at the end.
5. 5. People & Adoption: Almost every failed AI project shares a common root cause: people never started using it. Evaluation must include: How workflows change? Who is affected? What training is needed? How ready the organisation is for adoption?

This is where AI value is either unlocked or lost.

 

 

 

4. What the AI Scaling Tracker should include

A portfolio becomes actionable when it is paired with a structured tracking system. I use an AI Scaling Tracker template that turns the portfolio into a practical operational tool that guides the organisation from experimentation to production.

It captures every use case in a consistent format and shows exactly where each initiative stands, what it needs next, and how it can progress toward production and scale. The tracker ensures that each use case has a clear business purpose, defined ownership, and direct alignment with strategic goals. It highlights whether the required data exists, whether the solution is technically feasible, and which governance or compliance steps must be completed. This prevents pilots from succeeding under ideal conditions only to fail later in real operations.

It also makes adoption requirements visible by clarifying how work will change, which teams are affected, and what training or communication is needed. This helps organisations anticipate challenges early and design solutions people will actually use.

Most importantly, the tracker defines a structured scaling path. It outlines the requirements for moving from idea to pilot, from pilot to MVP, from MVP to production, and eventually to organisation-wide rollout. This removes ambiguity and turns AI work into a predictable pipeline rather than a collection of isolated experiments.

The tracker template includes:

• a clear use case description and owner,
• alignment with specific business goals,
• a quantified potential impact,
• data requirements and technical feasibility,
• governance and risk classification,
• user groups and adoption needs,
• and a defined scaling path with milestones.

This creates transparency and reduces bottlenecks. The result: AI becomes manageable, measurable, and scalable.

 

 

 

5. Why leaders need a portfolio now

The organisations that are making the fastest progress with AI are not those running the most pilots. They are the ones building systems: systems for prioritisation, governance, implementation, and scaling. A portfolio gives leaders visibility, alignment, and the confidence to invest. It ensures that the organisation focuses on the use cases that matter, not the ones that simply sound interesting. It reduces duplication, speeds up decisions, and helps teams move from experimentation to real execution.

When I spoke with the organisation in the story above, his biggest frustration wasn’t the number of pilots, it was the lack of a system to connect them. And this is exactly what an AI portfolio solves.

 

 

 

6. What companies should do now

Here is a practical plan for organisations that want to move from scattered pilots to a structured AI portfolio.

1. Create a complete inventory of all AI activities: Gather every pilot, idea, experiment, automation, or trial — even the small ones.
2. Evaluate each initiative using the five dimensions: This gives you an objective view of what’s valuable and what’s not.
3. Stop or deprioritise low-value experiments: Reducing noise is essential to gaining speed.
4. Prioritise three to five high-value use cases: Focus helps the organisation deliver real results quickly.
5. Define a clear scaling path: Specify what is required at each stage: data, integrations, governance, training, and ownership.
6. Assign business sponsors and technical owners: Clear responsibility accelerates execution.
7. Establish regular portfolio reviews: Weekly or bi-weekly cycles keep progress visible and unblock issues early.

These steps create clarity, direction, and measurable momentum.

 

 

 

Final thoughts

2025 was the year of experimenting with AI. 2026 will be the year of scaling it. Organisations that make the shift from scattered experiments and pilots to portfolios will be the ones that create real value, build AI-ready teams, and implement solutions that truly transform their work. That is where real AI transformation begins. When organisations adopt a portfolio mindset, they gain a system that reduces chaos, prevents duplication, and makes scaling predictable.

Would you like to learn more and apply this to your organisation? I’m here to help with my AI Use Case and Strategy Sprint.