By Chris Dodds – Co-founder and Managing Director, Digital
Software is getting easier to describe, easier to shape and easier to delegate – for everyone.
For the last few decades, software companies have largely sold access to interfaces that interpret and display data. Tabs, menus, forms, workflows, dashboards. Learn the system, click the buttons, follow the process. It worked because building software was super hard, changing it was expensive, and most people had to live with whatever product teams shipped.
That logic is weakening fast.
The newest wave of agentic AI models is changing the relationship between people and software. OpenAI is pushing hard into agents that can work across tools, files and computer environments. Anthropic is doing the same with Claude and Claude Code. Google is turning natural language into web apps, prototypes and design systems. Microsoft is making it easier to build agents inside workplace software using plain English. Amazon is packaging browser-based automation into production-ready agent workflows. Different wrappers, same direction of travel.
The important shift is not that AI can help engineers write code faster. That story is already old. The bigger shift is that the barriers to imagining, designing and building software are being lowered for everyone else.
A marketer with a sharp idea can now sketch an internal tool in natural language. A strategist can describe a workflow and get a working prototype back. A designer can move from prompt to interface to interactive flow without waiting for a dev sprint to open up. The phrase “vibe coding” can sound a bit silly, and to be fair, it is a bit silly. Still, it points to something real. You no longer need to know how to write every line of code to create useful software. Increasingly, you need to know what good looks like, what problem is worth solving, and how to steer the machine.
For Software as a Service (SaaS), the old moat was complexity. Build the app, host the app, charge per seat, and hope customers tolerate the friction because the alternative is disconnected spreadsheets and pain. Now the alternative is closer to: describe what you need and have an agent assemble it, adapt it, or operate it for you. I can imagine a future where new delivery software is spun up for every project – fully customised for specific requirements, with no bloat or subscription fees. Call an AI agent into your kick-off meeting, give it the tender doc and response, and presto – “I build you an app for that.”
Some SaaS products will hold up well in that world. The ones with deep data (your IP in their hands), embedded workflows, trust, compliance and real domain expertise still have an edge. But plenty of software categories are about to feel pressure from below. If a team can spin up a tailored internal tool or customer-facing app with natural language, the premium on generic one-size-fits-all software starts to wobble. That is especially true in categories where the interface matters more than the underlying IP.
This does not mean SaaS disappears. It means SaaS gets more fluid.
Software will behave less like a fixed product and more like a responsive layer. Interfaces will become more conversational. Workflows will be assembled on demand. Agents will sit between users and systems, doing the fiddly parts, translating intent into action, and cleaning up the grunt work that used to require three tools and six tabs. In plenty of organisations, the next useful app will come from the person closest to the problem, not the person with the deepest engineering background.
That is one reason we think ICON is in a strong position.
We have been leaning into agentic ways of working early, and with intent. New agents are being developed each week to help our team handle repeatable and complex task management across the agency. Some speed up research. Some support drafting, synthesis or workflow execution. Some help with the operational glue work that usually eats time and attention. Others build prototype software. Or convert UI design to front-end. We even have agents building agents. Our drive is to build practical systems that free people up to spend more energy on judgment, creativity and client value.
That said, there is a version of this story that gets overexcited and starts talking as if humans are about to become optional. We are not buying that version.
We are doubling down on a human-in-the-loop approach because it is the only serious way to do this well. Subject matter experts across the agency are being trained to use these tools, guide them, challenge them and validate what comes back. Speed on its own is cheap. Quality is still hard won. Context still matters. Taste still matters. So does accountability.
I feel the winners in this next phase will not be the teams that automate the most. They will be the teams that combine fast machines with capable people who know where the edge cases are, what the client actually needs, and when an answer is technically fine but strategically wrong. Clients want cost efficiency, but they also want human-to-human counsel.
The future probably belongs to businesses that are more agent-ready, more nimble and adaptable and far less precious about who gets to build. Some of that software will still be sold in the classic SaaS model. Some of it will feel more like a service layer for agents. Some of it will be assembled on the fly by people who would never have called themselves builders a year ago. And that’s super exciting.
So what happens when millions of new agents, apps, websites and services hit the market at scale?
For one thing, the internet gets a lot noisier. There will be more software, more automation, more polished-looking products, and plenty more things that are technically functional but not especially useful. If you’re old enough, you may remember a flood of desktop publishing revolution “slop” when anyone could produce a brochure. When the cost of building falls, the volume goes up.
But volume is only half the story. The more interesting shift is that expectations rise with it. If anyone can spin up an app, a site or a service from a decent prompt, then “we built a thing!” stops being impressive on its own. The advantage moves elsewhere. Better judgment. Better service design. Better taste. Better trust. Better understanding of what people actually need once the novelty wears off.
That is the world taking shape now. A market flooded with generated products, lightweight services and specialised agents, all competing for attention. Some will be clever. Some will be genuinely useful. Many will disappear as quickly as they arrived. The winners will be the ones who solve a real problem, fit cleanly into people’s lives, and keep earning their place.
In that environment, human oversight becomes more valuable, not less. The businesses that stand out will be the ones that know how to use AI to build at speed, while still applying the things scale does not automatically create: judgment, creativity and the amazing people who help ICON make what matters.