AGI is coming. The future of work isn't ready

As AI becomes more capable, the real business risk is not whether AGI arrives next year. It is whether organisations can redesign work, develop people and keep human judgement ahead of machine capability.

Most business leaders are not sitting around debating whether their organisation is ready for AGI.

They are thinking about revenue, budgets, customers, staffing, delivery and whatever problem has landed in front of them this week. Sometimes they are simply trying to survive the quarter.

That is understandable. But the best businesses do something else as well: they periodically lift their eyes from the immediate and ask a harder question.

What is coming next?

With AI, that question is becoming increasingly urgent.

Much of the media conversation is drawn to extremes. Will AI destroy jobs? Will AGI arrive next year? Could superintelligence become uncontrollable? Could AI eventually pose a threat to humanity?

Those questions attract attention because they are dramatic. Meanwhile, something quieter is happening underneath all that noise.

AI just keeps getting better.

Capabilities that looked unreliable a year ago are becoming useful. AI is moving beyond drafting and summarising into reasoning, research, coding, planning, tool use and multi-step execution.

What could not be automated yesterday increasingly can be automated today. And what cannot be reliably automated today may look very different twelve months from now.

That trend matters more to business strategy than arguing over whether one model can complete one particular task this week.

AI, AGI and what comes next

Part of the confusion is terminology.

Artificial intelligence already performs many tasks that previously required human cognitive effort: writing, analysing information, recognising patterns, solving problems, coding and increasingly taking actions.

Artificial General Intelligence, or AGI, generally refers to AI capable of performing effectively across a very broad range of intellectual tasks rather than being constrained to narrow domains.

There is no universally agreed test for when AGI has arrived. That matters because we may spend years arguing over whether AGI technically exists while increasingly general AI systems quietly transform work around us.

My own view is that some of the capabilities we historically associated with AGI are already beginning to emerge.

Artificial superintelligence is the more speculative stage at which machine intelligence substantially exceeds human capability across most cognitive domains.

The exact dates matter less than the direction. Businesses do not need to predict precisely when AGI arrives. They need to prepare for a world in which AI keeps becoming cheaper, more capable and more autonomous.

The media is debating the apocalypse. Businesses are still buying licences.

There is a strange disconnect in the market.

At one end, AI leaders are debating frontier risk, regulation and the safe pace of development. Anthropic CEO Dario Amodei has argued that the industry should slow the pace of capability development. OpenAI CEO Sam Altman has also talked about the need to "pace the frontier", while Google DeepMind CEO Demis Hassabis has described the current period as a "precious window" in which society can prepare for more advanced systems.

At the other end, many businesses are still asking whether they should adopt AI at all. So, they buy Microsoft Copilot licences. They give staff access to ChatGPT. They run a workshop. And somewhere along the way, access to AI becomes confused with having an AI strategy.

A licence is not a strategy.

Neither is a pilot. Neither is telling employees to "use AI more".

Strategy without tactics is hallucination. Tactics without strategy is the noise before defeat.

Businesses need both.

Strategy means deciding where AI could materially change the organisation: which workflows should change, which roles will evolve, where productivity can improve, which decisions should remain human, what data AI will rely on, what risks are acceptable and what capabilities people and managers will need.

Tactics turn those decisions into reality: training people, redesigning workflows, running pilots, building agents, improving data, establishing controls and measuring outcomes.

Without strategy, businesses end up with disconnected experiments. Without tactics, they end up with impressive presentations and very little change. The strongest organisations balance both.

Stop asking whether AI can replace someone

As an AI workshop facilitator and advisor, I am often asked some version of:

"Can AI replace this role?"

I understand why. Labour is expensive. Businesses are under pressure to improve productivity and reduce costs.

But I think it is usually the wrong first question.

"How can AI make this person significantly more productive, and how should this role evolve as AI gets better?"

Those two questions create very different management behaviour. One begins with headcount reduction. The other begins with redesigning. And that is where the real future-of-work opportunity sits.

What increasingly capable AI will look like

Forget humanoid robots.

Imagine a project manager in an engineering or infrastructure business. Today they may spend hours reading emails, tracking actions, preparing client reports, reviewing programme changes and updating risk registers.

Current AI can already help with pieces of that. Now imagine giving a more capable system the objective:

"Keep this project on track and alert me when something requires my judgement."

It could monitor correspondence, compare progress against the programme, identify emerging risks, prepare stakeholder updates, chase actions and escalate issues that cross agreed thresholds.

The project manager does not necessarily disappear. The work changes: less time assembling information; more time making decisions, managing relationships, resolving ambiguity and applying judgement.

Or consider a finance manager. AI could continuously monitor cash flow, budgets, receivables and anomalies, investigate unusual movements, prepare forecasts and recommend actions.

The finance manager increasingly asks: Why? What are we missing? What happens if this assumption is wrong? Should we act?

That is the shift. AI performs more of the execution. Humans increasingly provide direction, judgement and accountability.

Microsoft's 2026 Work Trend Index describes a similar change: as agents take on more execution, people increasingly move towards setting direction, making decisions and owning outcomes. View the Work Trend Index

Human + AI is the opportunity

Done well, the emerging operating model is Human + AI.

The human understands the business. The human sets direction. The human applies judgement. The human remains accountable. AI provides leverage.

But there is a risk.

If organisations do not deliberately develop their people, roles and workflows, Human + AI can quietly become AI + Human.

The machine performs most of the work. The person becomes the reviewer, approval layer or exception handler around a process they increasingly understand less well.

That might look efficient. It could also create a fragile organisation. Human capability erodes. Institutional knowledge weakens. People become less capable of recognising when the machine is wrong.

The challenge is not simply whether AI capability continues to improve. It is whether organisational and workforce capability can keep pace.

AI capability is advancing faster than workforce capability

Deloitte's 2026 research found that 74% of surveyed leaders expected nearly half their business processes to be redesigned or rebuilt around agents within four years. Yet only 25% considered their workforce prepared or highly prepared for agentic AI, while only 21% said the same about their business processes. 

That is the real strategic challenge.

AI capability is advancing faster than organisational and workforce capability.

So, the question becomes: how do organisations continuously adapt their people, roles and workflows as AI takes on more work?

This cannot be solved through a one-off training programme. Readiness becomes an ongoing management discipline.

We also need to measure whether people are ready

Most organisations still measure AI adoption through licences, usage statistics, course completion or self-reported confidence. Those measures tell us very little about whether someone can perform better with AI.

Can they frame a problem well? Can they decide when AI should be used? Can they provide the right context? Can they recognise a weak answer? Can they challenge AI-generated reasoning? Can they intervene when an agent goes off track? Can they use AI to produce a better business outcome?

That leads to a question I think will become increasingly important:

The next challenge is whether practical workplace AI capability can be measured through realistic AI-assisted tasks - and whether those measurements can be used to adapt learning and predict improvements in workplace performance.

If Human + AI becomes the new operating model, organisations will need a better way to measure the human side of that equation. Not just who has a licence. Who can work effectively with AI?

Five questions to ask now

Before scaling agentic AI, leadership teams should be able to answer five questions:

If those questions are difficult to answer, the organisation is probably not yet ready for highly autonomous AI. And buying more licences will not fix that.

Focus on the trend, not the AGI date

Perhaps AGI arrives in 2027. Perhaps later. Perhaps we never agree on exactly when the line was crossed.

I am increasingly convinced that the date matters less than the trajectory.

AI is getting better. AI is doing more. And the boundary between work performed by people and work performed by machines keeps moving.

The businesses that thrive will not simply be those with access to the best models. They will be the ones that continuously redesign work around that shift - strengthening their data, developing their people, changing roles and workflows, and measuring whether human capability is improving alongside machine capability.

That is why the biggest risk is not that AGI suddenly arrives one morning and wipes out the workforce.

The bigger risk is quieter.

AI capability keeps advancing while organisational capability stands still. Roles gradually change. Workflows quietly automate. People lose touch with how the work is performed.

And by the time leadership recognises how much the operating model has shifted, Human + AI has already become AI + Human.

The future of work is not something businesses can wait for. They must design it.

The organisations that prepare now will get to decide how humans and AI work together. The organisations that do not may eventually have that decision made for them.