AI eats the world

Benedict Evans is an internationally recognised technology analyst with over 20 years’ experience spanning mobile, digital media, and venture capital. Formerly working in Silicon Valley, he now operates independently, sharing insights into consumer technology and digital ecosystems through his essays, a widely read newsletter, and sought-after presentations.

NashTech was delighted to welcome Benedict as a keynote speaker at the NashTech Connect Conference 2026.

At the conference, his latest fast-paced, entertaining presentation, “AI eats the world”, offered an incisive look at where artificial intelligence (AI) sits in the broader arc of technological change.

The pattern: every technology shift starts with confusion

Evans opened by reminding the audience that every major technology shift, from PCs, the web, and smartphones, began with uncertainty. We don’t know what to measure, what matters, or what the killer applications will be. AI is no different. Benedict said, “It’s hard to know where AI will affect your business in the future. In the 1990’s, no one knew exactly what the internet would do for their business. They might have guessed and got a few things right, but not everything.”

The real question, he argued, is not how big these models are, but what they enable, “AI gives you infinite interns, scrap that, infinite associates”.

Software is being rewritten, not replaced

One of Evans’ most compelling points was that AI is not ‘replacing software’, it is changing what software is.

Traditional software is deterministic: rules, logic, and structured data. AI introduces probabilistic systems that can interpret, generate, and interact with unstructured information. This doesn’t eliminate existing software; it expands the surface area of what software can do.

He framed AI as a new ‘runtime’ for the digital world, something that will sit underneath countless experiences, often invisibly, just as the web and mobile did before it.

The distribution advantage

Evans challenged the assumption that AI startups will automatically dominate. In his recent essays, he notes that OpenAI, despite its prominence, doesn’t have a unique technology advantage and faces incumbents that can match its models and deploy them at a massive scale.

In the talk, he extended this argument: the companies that already own distribution, such as hyperscalers and enterprise SaaS giants, can integrate AI into existing workflows faster than new entrants can build them.

The implication is clear: AI alone is not a moat. Distribution, data, and product integration still matter.

The real value: new experiences we haven’t invented yet

Evans stressed that the most transformative AI applications are not the ones we’re talking about today. Just as early mobile apps were desktop websites squeezed onto a small screen, today’s AI tools are often old workflows with a chatbot bolted on.

The breakthrough moments will come from:

• New interfaces that feel native to AI
• New workflows that eliminate entire categories of friction
• New products that couldn’t exist before probabilistic computing
• Automate the repetitive
• Accelerate the complex
• Amplify the skilled
• The biggest impacts are still ahead
• The winners are not predetermined
• The most important applications haven’t been invented yet
• And the real transformation will come from rethinking industries, not just tools

These will be built not by one company, but by thousands of teams experimenting across every industry.

AI and the labour question: automation vs. acceleration

Evans avoided simplistic narratives about AI replacing jobs. Instead, he framed the shift as a reallocation of tasks, not a wholesale elimination of roles. AI will:

Organisations that redesign processes around these new capabilities, rather than simply inserting AI into old ones, will see the biggest impacts.

Evans quipped, “Introducing Microsoft Excel didn’t mean accountants worked shorter hours; with price elasticity, it just meant they did different things.”

Regulation, risk, and the shape of the market

Evans also addressed the regulatory landscape, noting that governments are struggling to define what AI is, let alone how to regulate it. The risk is premature or misaligned rules that freeze the technology in its current form.

He argued for a focus on outcomes: safety, transparency, accountability, rather than prescriptive controls on model architecture or training methods.

To conclude

Evans closed with a powerful framing: AI is not a single technology but a platform shift on the same scale as the internet or the smartphone.
That means:
In his words, AI is “eating the world” not because it replaces everything, but because it becomes part of everything.

If you want to see what AI could do for your business or chat about ideas, get in touch with NashTech today. We can help you get ready for what’s next.