Someone has to press the “Go” button
An op-ed from Chris Dodds, ICON co-founder and Managing Director, Digital.
For a long time, business software has been built around a simple assumption: people do the work, and software helps them do it. A CRM helps a sales team manage leads. Project software helps an agency run jobs. An accounting platform helps a finance team keep the books. The software matters because the people matter. It gives structure, speed and visibility to human effort.
UX exists because humans need intuitive interfaces to interpret and use content and data.
AI starts to unsettle that model because it doesn’t just help with the work. It begins to do the work itself.
At first, that looks familiar. A strategist uses AI to conduct research and prepare a first draft. A lawyer uses it to review documents. A recruiter uses it to screen candidates. A consultant uses it to summarise research and build a presentation. Most people still think of this as productivity software with a better interface.
The bigger change comes when the human stops being the main engine of delivery.
Take a mid-sized communications agency. Today it needs account managers, strategists, writers, analysts, media specialists and project managers to deliver a campaign. Over the next few years, much of that work can be handled by synthetic workers being directed by subject matter experts. One agent monitors media and public sentiment. Another drafts a response matrix. Another builds a stakeholder map. Another generates reporting. Another adapts copy for different channels. The human team is still there, but their role shifts towards judgement, client handling and final calls.
That is the first step. The labour model changes.
The second step is harder to see because it doesn’t just affect jobs. It affects the business model itself.
Once AI can reliably perform most of the moving parts of an agency, consultancy, recruitment firm, accounting practice or research shop, someone will package that into a product. Instead of hiring a small agency, a client buys an AI-native service platform that does audience analysis, campaign planning, content generation, stakeholder tracking and performance reporting in one place. Instead of paying for a team of junior and mid-level people, they pay a subscription or usage fee. What used to be a service business starts to look like software.
You can already imagine the pitch.
Upload your brief. Connect your channels. Approve the direction. The platform handles the rest.
The same thing happens in law, recruitment, finance, training, procurement and a long list of other knowledge-heavy sectors. First AI helps the professionals. Then AI reduces the number of professionals needed. Then a new class of SaaS product comes along and offers the whole service model in product form.
It doesn’t end there.
Software as a Service (SaaS) still assumes that the buyer is choosing a product made by someone else. It still assumes fixed workflows, defined features and a vendor deciding how the system should work. That is where things get shaky.
If AI gets good enough to assemble workflows, connect data, manage logic, generate interfaces and run tasks from a plain-language description, then the need for a fixed software product starts to fade. You do not buy a tool built around someone else’s assumptions. You describe the business you want, and an AI operating layer builds it.
Say you want to launch a niche corporate affairs advisory focused on ESG issues in the energy sector. Today, you would need a website, CRM, proposal templates, reporting dashboards, media monitoring, document workflows, research tools, finance systems and a delivery team. In the next phase, you may simply describe the firm you want to create.
Build me a specialist advisory for ASX-listed energy companies. Give it a premium brand position. Set up a website, outbound prospecting, proposal generation, quarterly reporting, media and policy monitoring, a client portal and internal workflow management. Use a conservative tone. Flag high-risk issues for human review. Price work on a monthly retainer plus project fees.
That description becomes the operating model. The system builds the infrastructure, spins up the agents, sets the workflows and starts running the business.
At that point, the old distinction between software and company starts to blur.
This is why I think a lot of current discussion about AI misses the real trajectory. People are debating whether AI will replace workers, or whether software companies will add enough AI features to stay relevant. Both questions matter, but they are still close to the current frame. The more interesting possibility is that software itself becomes less visible because language becomes the way businesses are assembled and run. Software will become democratised, decentralised, and hyper-personalised.
You can see early versions of this already. Someone uses no-code tools to launch a micro-business in a weekend. A solo founder runs marketing, sales support, customer service and finance with a stack of AI tools and a handful of contractors. A small e-commerce brand uses synthetic agents to post social updates, write product copy, answer customer questions and generate sales leads. These are rough examples now, but they point in the same direction: fewer fixed systems, fewer fixed roles, more orchestration.
That changes what matters commercially.
If anybody can create a capable virtual business, then capability itself stops being scarce. Plenty of firms will be able to produce competent strategy, competent copy, competent analysis, competent customer service and competent reporting. Being able to do the work will not be enough because everyone will have access to roughly the same underlying intelligence. The world will be flooded with millions of synthetic businesses.
What becomes harder to replicate is trust.
Why does one advisory get hired over another when both can generate good thinking instantly? Why does one wealth platform keep clients when dozens can offer similar analysis and automation? Why does one communications business stand out when every competitor has access to the same tools and workflows?
Usually because clients are not only buying output. They are buying confidence, credibility, taste, accountability and access. They want to know who stands behind the work. They want to know which judgement calls are being made by a person. They want to know whether the business understands their world, their risks and their politics. In some categories, they will actively pay more for that human layer.
Marshall McLuhan called this the “Reversal": push any medium to its limit, and it flips into its opposite. If I can’t see you breathe, smell you or touch you, I don’t believe you.
Think about a government tender. An AI-built delivery model might produce a strong response, a sensible project plan and a compliant pricing structure. But when the panel is choosing a partner, they are also judging whether the team can handle ministers, stakeholders, public scrutiny, political sensitivity, procurement realities and reputation risk. Those things do not disappear because the work is more automated. In some ways they become more important.
The same applies in medicine, law, public policy, executive search, corporate reputation and high-stakes advisory work. The mechanical part gets cheaper. The relational part becomes more valuable.
So the future may not belong to the company with the biggest software stack. It may belong to the company that knows how to combine machine capability with a point of view people trust.
That is where I think this heads over the next three to five years.
First, synthetic workers take on more of the day-to-day knowledge work inside firms.
Then AI-native SaaS businesses start packaging entire service models and selling them cheaply, quickly and at scale.
After that, even those SaaS products start to feel restrictive, because people can generate their own operating systems for specific business models just by describing what they want.
Once that happens, the barrier to forming a sophisticated business drops dramatically. More firms appear. More niches are crowded. More services become interchangeable.
And when everyone can build a capable business quickly, the real differentiators are not features. They are reputation, judgement, relationships, proprietary data, access to customers, and the confidence that somebody credible is accountable when it matters.