‍

3X Optionality: AI solutions that create value from day one

Published

December 3, 2025

Author

Mads Hofman

CEO & Advisor

Many projects fail long before a single line of code is written. Too often, companies try to define the final solution and delivery date upfront, locking themselves into long planning cycles, rigid assumptions, and expensive, over-engineered products built before anyone understands what users actually need.

At Abtion, we take a different path led by new AI opportunities.

Drawing on Design Thinking and inspired by Kent Beck’s 3X model of Explore, Expand and Extract, we’ve shaped an AI-ready, optionality-driven approach to product development that embraces uncertainty, learns from real usage, and evolves solutions based on evidence rather than prediction.

3X Optionality

Instead of heavy upfront decisions, the 3X approach offers a simple start: small experiments, low risk, and fast learning.

We rapidly explore possibilities, test multiple paths in small, safe steps, and only scale what we know works. The result is digital projects led by AI, that don’t start with complexity and assumptions, but with fast learning loops, low risk, and tangible results from the very first sprint.

Mads Hofman

CEO, Abtion

3X in practice

From legacy OCR to AI data entry

For DSV Transport we replaced an outdated OCR system that struggled to read delivery slips, slowed drivers down, and cost too much to maintain. Instead, we tested modern AI on real photos and metadata like GPS and task type, and quickly confirmed it could read the slips far more accurately. We then built a prototype that autofilled data in the backend so drivers only had to verify, not type. After testing, the AI fully replaced the old OCR, delivering faster workflows, lower cost, greater accuracy, and a smoother data-entry experience for drivers.

Why 3X Optionality makes sense for AI projects

Kent Beck’s philosophy is simple: You cannot optimize something you do not yet understand. Therefore the work is divided into three stages:

Explore - understand the problem and the opportunities:

In this phase we test hypotheses, validate data sources, work with real examples, and examine where AI can realistically create value.

‍

Expand - grow the right solution:

When we find something that works, we build a functional prototype, test it with users, gather data, adjust and expand only what shows real effect.

Extract – make the solution robust, efficient, and scalable:

Here we streamline the solution, reduce dependencies, optimize costs, and prepare the technology for production-level performance.

What we learned

AI enabled a faster and smoother workflow where the driver maintains control but spends much less time on manual entry.

AI as support, not replacement.
The human decides, AI reduces workload.
Context beats complexity
GPS, task type and history yield better results than simply using a larger model.
Errors are best prevented early.
Smarter warnings and validation make the process easier without feeling like surveillance.
Why 3X Optionality works
- quick results - minimal upfront cost - a clear link between AI and actual business needs - a robust final solution only once we know what works In short: AI without hype, but with real value from the very first steps.