In 2004, I was leading engineering and product development at IG Group, the LSE-listed financial trading platform. We were profitable, we were dominant in desktop trading, and we had built a significant institutional business. We were, by the standards of the time, a successful software company.
We were also about to face an inflection point that we did not fully understand.
Mobile was coming. It was obvious in retrospect, but it was not obvious in 2004. Smartphones did not exist. The iPhone would not launch until 2007. Yet we could see the trend: computing was becoming ubiquitous. People would want to trade on their phones. It was not yet, but it would be.
We made a strategic decision that, in hindsight, was correct but felt risky at the time: we were going to go mobile-first. Not mobile-second. Not mobile-as-a-feature-on-our-existing-platform. Mobile-first.
This meant deliberately rebuilding core aspects of our platform around the constraints and capabilities of mobile devices. It meant designing workflows for small screens, for intermittent connectivity, for touch rather than keyboard-and-mouse. It meant investing in capability we could not yet monetise because the customer base did not yet exist. It meant having our most senior technical people spend time on a platform that, in 2004, was essentially a novelty.
We were not the only financial trading firm. The incumbents — and there were several large ones — saw mobile coming too. But they followed a different path.
The incumbent playbook was: Keep the core desktop platform running. Add a mobile client that wraps the desktop functionality. Maintain the desktop as the primary product, the place where the real power is. Make mobile useful but secondary.
This made intuitive sense. You have a working, profitable business on desktop. You are not going to blow that up to chase a hypothetical mobile future. You will evolve desktop, and you will add mobile alongside.
The problem was that mobile was not just a new channel. It was a different computational model. It had different constraints, different opportunities, and different user expectations.
By 2010, we had launched our first mobile trading app (on J2ME, before smartphones existed). When the iPhone launched, we had a native iOS app. By 2012, we had a fully AI-powered Watch app that was featured at Apple's Watch launch. Our entire platform had been redesigned around mobile-first principles. The mobile experience was not a shrunken version of desktop — it was the primary experience, and the desktop experience was a more powerful variant.
Meanwhile, the incumbents were managing the transition from desktop-primary to mobile-secondary. They had two codebase. Two user experience paradigms. Two sets of feature parity problems.
Mobile did not displace desktop. Rather, organisations that built mobile-first ended up with architectures that were flexible enough to support both better than organisations that bolted mobile onto desktop. The platform that started with mobile constraints in mind was actually more robust and more elegant than the platform that started with desktop and had mobile grafted on.
By 2015, when I left IG Group, the mobile transition was largely complete. The companies that had gone mobile-first in 2005-2007 were now dominant. The companies that had tried to maintain desktop-first while adding mobile had experienced the messy complexity of managing two different product models.
There were other factors, of course. Product quality mattered. Execution mattered. Capital matters. But the strategic choice — mobile-first vs. mobile-bolted-on — was a first-order determinant of competitive position.
The companies that won were not necessarily the ones that were "better at mobile." They were the companies that had reorganised their entire product development, their engineering culture, and their business model around mobile from the start. That organisational commitment was hard to reverse-engineer. Competitors could hire engineers, could copy features. But they struggled to reorganise their entire platform around a different model.
I see the exact same pattern beginning with AI.
There are two AI strategy playbooks circulating right now:
Playbook 1: AI-First
Playbook 2: AI-Bolted-On
If I am right about the AI parallels to mobile, Playbook 2 feels like the right choice right now. It is the incumbent playbook. It is the safe choice. And in the next 2-3 years, it will feel like the right choice.
But in 5-10 years, when AI is truly embedded in every workflow and every decision in the organisation, the companies that went AI-first will have competitive advantages that companies that went AI-bolted-on cannot easily replicate.
The mobile transformation at IG Group was not fundamentally about the technology. We did not have some magical mobile algorithm that competitors could not implement.
The real transformation was organisational and architectural:
First, organisational: The teams that were responsible for desktop product development understood desktop deeply. They had mental models, heuristics, and decision-making processes built around the desktop paradigm. When they tried to add mobile, they kept trying to apply desktop logic to mobile problems. This created friction and suboptimal products.
In contrast, teams that started with mobile-first had to learn a different paradigm. They could not fall back on "the way we have always done it on desktop." They had to think from first principles: What does a good mobile experience look like? How do we handle disconnection? How do we design for small screens? How do we handle the unique capabilities of mobile (location, sensors, notifications)?
That learning mindset — combined with the constraint-driven design that mobile forced — created a culture that was more adaptable to new paradigms later.
Second, architectural: A mobile-first architecture is fundamentally different from a desktop-first architecture. Desktop applications were built around synchronous request-response flows, rich client-side state management, and powerful computational substrate on the client machine.
Mobile forced you to think differently: asynchronous communication, lightweight state management, distributed computation between client and server. These constraints forced you to build more robust, more distributed, more resilient systems.
When it came time to add AI, the companies that had gone mobile-first already had the architectural thinking in place. They were comfortable with distributed systems. They were comfortable with asynchronous workflows. They had learned to build systems that worked under constraints.
Companies that had maintained desktop-first architectures had to learn all of this again. They had to restructure their thinking about how systems work.
Third, cultural: The mobile-first companies had developed a culture of innovation. They were used to building for new paradigms. They had learned to prototype quickly, to test assumptions, to fail small.
Companies that had maintained desktop-first architectures had a culture of stability. They were optimised for executing against a known model, for squeezing efficiency out of existing systems, for avoiding the risks that come with fundamental change.
When AI came along, the mobile-first culture was ready. They could spin up AI pilots quickly. They could fail and iterate. They could reorganise around new opportunities.
The desktop-first culture struggled. They wanted proof that AI would work before investing. They wanted certainty before committing. By the time they had that certainty, the first-movers had already built competitive advantage.
Look at the companies that are winning with AI in 2026. Look at the companies that are effectively using AI to create new products, to automate operations, to gain competitive advantage.
Many of them are technology-forward companies that go AI-first: they assume AI will be foundational and they are reorganising around that. Some are small companies that do not have legacy constraints and so can think AI-first from inception. Some are incumbents from other industries that understand transformation (financial services companies, for instance, are relatively comfortable with technology transformation because they have done it before).
Look at the companies that are struggling with AI. Many of them are trying to maintain their existing business model — whether that is a traditional consulting model, a traditional insurance model, a traditional manufacturing model — and bolt AI onto it. They are asking: How do I add AI to my current business? Rather than: How does AI change my business?
The second question leads to different answers. And over time, the companies asking the second question will build more sustainable competitive advantage than the companies asking the first.
If you are a CEO, a board member, or an investor, here is what I would advise:
| Dimension | AI-First Strategy | AI-Bolted-On Strategy |
|---|---|---|
| Organisational structure | Reorganise teams around AI, not around legacy products | Maintain existing teams and add AI expertise alongside |
| Talent allocation | Best people working on AI and new models | Best people maintaining core business, moderate people on AI |
| Product architecture | Redesign around AI assumptions; treat legacy as constraint | Maintain legacy, add AI features |
| Data strategy | Build comprehensive data infrastructure from scratch; assume all decisions mediated by data | Use existing data, add AI analysis alongside existing reporting |
| Culture | Shift toward experimentation, iteration, failure tolerance | Maintain stability culture, add innovation appetite |
| Short-term results | Investment phase; less visible results in year 1 | Visible near-term improvements, but hits constraints later |
| Long-term position | Platform advantage; higher defensibility in year 3+ | Feature advantage; erodes as competitors copy |
For fast-growing companies and startups: You have a choice. You can go AI-first and assume AI will be embedded in everything you build. Or you can build normally and add AI later. I would recommend AI-first. Do not default to "we will add AI later" just because that feels more conservative. It actually is not more conservative in the long run.
For incumbents in stable industries: You probably need AI-first even more than fast-growth companies, because the inertia of your existing business model is stronger. The mobile companies that went mobile-first were successful partly because they were willing to disrupt themselves. If your industry is facing AI-driven disruption, the way to protect yourself is to become the disruptor.
For PE-backed companies: Here is a question for due diligence: Is this company positioned AI-first or AI-bolted-on? If AI-bolted-on, what is the company's plan to transition to AI-first? What investment would that take? How long would it take? How much competitive advantage would be gained by that transition relative to peers?
The company that can articulate a clear path to becoming AI-first, and has the capital to invest in that transition, is more valuable than the company that is focused on optimising the existing business model.
The pattern repeats across major technology transitions:
History does not guarantee outcomes. But it suggests patterns. And the pattern is clear: the companies that fundamentally reorganise around a new technology paradigm, rather than trying to add it onto existing structures, tend to win.
AI is not a feature. It is not an optimisation. It is a new computational paradigm that will, over the next decade, affect how almost every organisation creates value.
The question for your organisation is not: How do we add AI to our business? The question is: How do we reorganise our entire business around AI?
The answer will be different for every company. For some, it might mean breaking existing systems. For some, it might mean reorganising your data architecture. For some, it might mean redefining what your product is.
But the question matters. And the companies that ask it early, that take it seriously, and that commit resources to answering it, will build advantages that compounds over time.
I made that choice at IG Group in 2005. It felt risky then. In retrospect, it was the most strategically important decision we made.
Ivan Gowan is the founder and CEO of Opagio. He spent 15 years as a senior technology leader at IG Group (LSE: IGG), overseeing engineering growth from 12 to 250 during the company's rise from £300m to £2.7bn market capitalisation. He led the mobile transformation from 2005-2015 and was responsible for IG Group's first Apple Watch trading application. He holds an MSc from Edinburgh with research in neural networks (2001).
AI investment has crossed $500 billion annually, yet measured productivity growth remains flat. This article explores why conventional metrics fail to capture AI-driven value creation and what alternative measurement approaches reveal.
Read more →
The San Francisco Fed's recent research on AI and productivity is among the most rigorous work in the space. But it also reveals critical blindspots in how economists measure the impact of AI on economic growth. Here is what they got right, what they missed, and why the measurement gap itself is the most important finding.
Read more →
Global AI investment has surpassed $500 billion annually, yet total factor productivity growth remains stubbornly flat. This is Solow's paradox reborn: you can see AI everywhere except in the productivity statistics. History tells us why — and what needs to change.
Read more →Get the latest insights on intangible asset growth and productivity delivered to your inbox.
Book a free consultation to see how the Opagio Growth Platform can help your business.