By Rakshit Choudhary, CEO, Deriv

The first weeks of 2026 have made one thing clear. AI is no longer moving in steps, it is moving in leaps. Across the Middle East and globally, organisations are spending hundreds of billions on AI, yet most will fail to see a lasting advantage. This isn’t a technology failure, it’s an architectural one. They are using 2026 intelligence to automate 2016 processes that shouldn’t exist in the first place.

One question separates genuine transformation from expensive automation. If you were building this business from scratch today, how would you design it?

 The asymmetry of the legacy burden

Established companies face a challenge startups do not. Every advantage built over time eventually hardens into a constraint. Processes reflect historical decisions made years ago, and systems are optimised for legacy technology.

A startup building your business today wouldn’t carry your infrastructure or justify changes to existing teams; they would simply build what makes sense now. This creates a painful reality where startups move faster not because they are smarter, but because they don’t have to preserve a museum.

At Deriv, we faced this asymmetry head-on. We had to redesign our entire foundation while maintaining over $650 billion in monthly trading volume for 3 million clients. It is the equivalent of building a new aeroplane while flying at 35,000 feet.

Designing for intelligence, not compensating for its absence

Most organisations approach AI by asking, “What can AI do for us?”. That is the wrong question. It leads to incrementalism, existing workflows executed slightly faster.

When we applied “startup thinking” to Deriv, we stopped treating AI as a tool and started treating it as a design constraint:

• Customer service: The answer wasn’t faster scripts, but an AI agent with direct system access. Our agent, Amy, now handles 79% of customer chats globally with 97% satisfaction.
• Engineering: We didn’t just ask for more “copilots.” We built for AI-generated code with built-in quality controls. Today, over 50% of our code is AI-generated, putting us ahead of most software firms in a regulated environment.

Every time we asked the “startup question,” we discovered that legacy processes were designed around constraints that no longer existed. Technology limitations from a decade ago or organisational structures reflecting a much smaller company.

The investment that actually matters: Readiness

AI capability is no longer the bottleneck. Access to breakthroughs is now commoditised and available across markets as quickly as it emerges. The real constraint is organisational readiness.

The most valuable investment we made in 2025 wasn’t software, it was people. We have hired over 100 AI engineers to build AI-native operations, but we also upskilled our existing global workforce. This wasn’t about teaching them to use a chatbot, it was about changing their AI literacy so they instinctively ask if a process should exist at all.

The widening gap

We are at a critical inflection point. Product lifecycles and release timelines that took months now happen in weeks. Companies that redesign workflows for autonomous systems will benefit automatically as AI improves. New capabilities will integrate without disruption.

Conversely, those automating legacy processes will find themselves trapped in a cycle of continuous, expensive rebuilding. By mid-2026, this gap will become permanent.

The startup question isn’t comfortable. It challenges every inherited assumption. But for businesses operating in sophisticated, highly regulated markets, it is the only question that leads to growth rather than mere efficiency.

The time to ask the startup question is now.