Overview

Lieselotte Vandenberg runs AI capability for a mid-size logistics company in Rotterdam. Last spring her team scored themselves on a maturity framework and came out feeling pretty good - a 3.4 out of 5. Then a consultant asked one question: "Which of those capabilities actually connects to a business outcome that the board cares about?" Lieselotte went quiet. The score was real, but it meant almost nothing without knowing which gaps were costing the company money and which gaps were simply cosmetic.

That is the trap of gap analysis done wrong. You measure everything, prioritize nothing, and produce a roadmap nobody funds. This lesson is about doing it differently.

What Gap Analysis Actually Is

A gap is the distance between where you are now and where you need to be to achieve a specific objective. That last phrase matters. *To achieve a specific objective.* Without a destination, there is no gap - there is just a list of things you do not have yet.

Gap analysis in the context of AI capability has three steps:

• Define the target state in terms of business outcomes, not technology features.
  - Honestly assess the current state across people, process, data, and technology.
  - Identify and rank the gaps by their impact on reaching that target state.

Step one is the hardest and the most skipped. Most organizations jump straight to step two. They audit what AI tools exist, survey which teams use them, and document skill levels. The result is a detailed picture of the current state with no anchor. It is like mapping a route without knowing the destination.

> The gap that costs you the most is rarely the most obvious one. It is usually the capability your most important process silently depends on.

Defining the Target State

Start with the business objectives that matter to leadership this year. Not the AI strategy - the business strategy. Revenue growth, cost reduction, cycle time, customer satisfaction, regulatory compliance. Pick two or three that AI could plausibly accelerate.

For each objective, describe what the AI-capable version of your organization looks like. Be specific. Not "we use AI across the business" but "our sales team qualifies inbound leads in under four hours using an AI-assisted scoring model, freeing account executives for complex deals."

This target description is your destination. Now you can identify what capabilities are missing.

Four Dimensions of Capability

Capability gaps usually fall across four dimensions. Work through each one when assessing distance from your target state.

People. Who needs to use, oversee, or maintain AI in this scenario? Do they have the skills? This includes both technical skills (prompting, data interpretation) and judgment skills (knowing when AI output is wrong, understanding limitations).

Process. Does your current workflow accommodate AI in this role? Are there handoffs, approvals, or quality checks that assume a human is doing the work? Those assumptions often need to change before AI can be inserted.

Data. Does your organization have the data this use case requires? Is it clean, accessible, and labeled appropriately? Data gaps are often the longest to close - months, not weeks.

Technology. Do you have the tools, integrations, and infrastructure? This dimension is often overemphasized. Organizations spend months selecting technology while people and data gaps go unaddressed.

Prioritizing Gaps by Impact

Lieselotte's team, after the consultant's question, went back and mapped each identified gap to a specific business objective. They found 23 gaps. Of those, 7 had clear links to revenue or cost outcomes. Those 7 became the priority list. The remaining 16 went into a "monitor" category.

This is a useful heuristic: if you cannot explain how closing a gap makes a measurable difference to an outcome leadership cares about, that gap can wait.

For the priority gaps, score each one on two axes:

• Business impact. How much does closing this gap move the needle on the target objective? Try to attach a number: hours saved per week, conversion rate improvement, error rate reduction.
  - Feasibility. How hard is it to close? Consider time, cost, dependencies, and change management complexity.

Plot gaps on a simple two-by-two: high impact / high feasibility gaps are your quick wins. High impact / low feasibility gaps need a longer runway and executive sponsorship. Low impact gaps go to the backlog regardless of how easy they are to close.

Building the Improvement Roadmap

A roadmap is only useful if it reflects reality. Two failure modes are common.

The first is the over-ambitious roadmap. Everything is priority one. Timelines assume perfect execution. Leadership approves it because it looks comprehensive. Six months later, three initiatives are stalled and nobody knows which one to rescue first.

The second is the under-committed roadmap. It is so hedged with "pending further review" language that it communicates nothing. Teams do not know what to work on. Nothing changes.

A good improvement roadmap has four components for each gap:

• Owner. One named person accountable for closing this gap. Not a team - a person.
  - Actions. The two or three concrete things that must happen. For a skills gap: identify training program, enroll target population, set completion date. Not "upskill the team."
  - Resources. What budget, headcount, or tooling is required? Be honest. Underfunded initiatives fail and consume management attention on the way down.
  - Milestone dates. Specific checkpoints, not just a final deadline. A gap that takes six months to close should have a checkpoint at week six and week twelve.

Estimating Resources Honestly

The most politically uncomfortable part of gap analysis is the resource estimate. Closing a data quality gap might require a full-time data engineer for four months. Closing a skills gap across 150 employees might require 20 hours of structured training per person plus manager coaching time.

Underestimating resources is one of the most common reasons AI initiatives stall. They were approved with optimistic numbers, hit real-world constraints, and then quietly failed without anyone formally deciding to stop.

Use three categories when estimating resources: people (hours or FTE), money (tooling, external support, training programs), and time (calendar duration accounting for competing priorities). For each gap, build a low and high estimate. Present both. The range is honest; a single number usually is not.

Tracking Progress and Adjusting

Milestones are not enough on their own. You need a lightweight review cadence - a monthly thirty-minute check-in where owners report status against their milestones. The review should answer three questions: Is this on track? If not, why? What needs to change?

The most important output of the review is the decision to adjust. Plans go stale. A vendor changes their pricing. A key employee leaves. A regulatory update shifts the priority. Organizations that treat the roadmap as fixed miss these signals. Organizations that treat it as a living document adapt faster.

When an improvement stalls, resist the urge to simply extend the deadline. Diagnose first. Is the resource estimate wrong? Is the owner unclear on what they are supposed to do? Is the gap more complex than originally assessed? The answer changes what you do next.

Putting It Together: Lieselotte's Six Months Later

After anchoring gaps to business objectives, Lieselotte's team focused on three: improving AI-assisted exception handling in the logistics platform, reducing time-to-insight for operations managers, and building a data readiness standard for two key data sources.

Each had a named owner, a concrete set of actions, and a monthly checkpoint. Six months later, two of the three were closed or substantially improved. The third was descoped when the business objective it was tied to shifted. That decision took one conversation at a quarterly review. It would have taken three months of committee deliberation under the old approach.

The scorecard now sits in a drawer. The roadmap is on the wall.

Key Takeaways

• Anchor gaps to business objectives. A gap that does not connect to a measurable outcome is not a priority, regardless of how obvious it seems.
  - Assess four dimensions. People, process, data, and technology each contribute to capability gaps. Most failing initiatives have a people or data gap that was invisible on the technology roadmap.
  - Score by impact and feasibility. High-impact, high-feasibility gaps become quick wins. High-impact, low-feasibility gaps need executive sponsorship and a longer timeline.
  - Assign one named owner per gap. Shared ownership reliably produces no ownership. One person is accountable; others are contributors.
  - Estimate resources honestly. Present low and high estimates. Underfunded plans fail slowly, which is more damaging than a plan that was never started.
  - Build a review cadence. Monthly checkpoints with three questions - on track, why not, what changes - keep the roadmap alive and allow fast adaptation.
  - Treat the roadmap as a living document. Business conditions change. A gap plan that cannot be adjusted is a plan that will eventually be ignored.