The State of Artificial intelligence How Work

Getting Real on AI Context & Challenges

Artificial Intelligence has been around for decades but even in 2025, many organisations still struggle to turn AI from a nice idea into a real, meaningful driver of value for their business.

Right now, we face a core problem: lots of businesses are playing with AI, even experimenting with some great ideas, but actual transformation – the kind of change that really moves the needle across the whole organisation – is still all too rare. Many companies are dabbling in AI, running little pilot projects here and there, but few have cracked the code on how to make AI have a real, lasting impact.

Meanwhile, the hype and expectations around AI just keep on growing. You hear people talking about “AI agents”, “generative AI”, “a workforce revolution”, “automation of everything”, but what’s really working, what’s just noise, and what can actually transform business in a meaningful way?

Without a clear picture of what’s actually working and what’s not, companies risk under-investing in AI, misallocating resources or ploughing way too much cash into tools that won’t deliver.

The bottom line is there’s a huge gap between just ‘trying AI’ and building a sustainable, enterprise-wide AI strategy that actually drives results. In 2025, lots of organisations are still stuck on the early adoption curve – using AI in isolated pockets, without a clear plan or vision – and that’s where all the confusion, unmet expectations and wasted investments come from.

To navigate this moment, we need a grounded, data-driven view of where AI stands now – what’s working, where AI is truly making a difference, what innovations are driving change & what remains speculative or uncertain.

Why AI Matters Now (Before It’s Too Late)

Imagine your company – or your job – mid-2025. You’ve heard about AI tools that can automate tasks, generate content, speed up development, analyse data. You’re convinced – you invest, you buy some AI software, run a little pilot, maybe build a chatbot or adopt an AI-powered analytics tool.

But months go by. The pilot limps along, a side project. No real cost savings, no noticeable productivity gains, no new product. Your team gets back to business as usual. Leadership starts to question the value of all that money spent.

That’s exactly what’s happening in thousands of companies right now. The frustration builds, early enthusiasm evaporates and AI just becomes another tool in a crowded toolbox – not the transformative engine that everyone hoped for.

Meanwhile, competitors who invest in AI more strategically are starting to pull ahead: faster innovation, automated workflows, smarter analytics, better customer experience. For them, AI is finally starting to make a difference.

The core tension is simple: AI holds out a lot of promise, but promise is not the same as delivery. Without a clear strategy, some vision and a realistic plan for scaling, AI just stays a loose experiment – not a business game-changer.

That can mean wasted budget, disappointed leadership, missed opportunities – and worst of all, a growing scepticism about AI’s real value.

If that becomes the norm, investment in AI will dry up. The broader potential of AI – across business, industry, geographies – will get stalled long before it has a chance to really change things.

That’s not just a business risk – it has implications for jobs, for economies, for societies as a whole.

So the question becomes urgent: which areas of AI are actually delivering in 2025? Which innovations are really driving change? Where should businesses put their time and money – and where should they be cautious?

Answering that matters for business strategy, for resource allocation, for worker skills – and for long-term sustainability.

What to Do About AI Now (A Clear Picture of Where We Are)

Data from recent surveys and reports offers some clarity. The reality of 2025’s AI landscape is a bit of a mixed bag: there is some real, meaningful growth – but it’s concentrated in certain areas. Understanding which areas matter can guide some really smart decisions.

This section breaks down the state of AI in 2025 in three dimensions: adoption & scale, emerging innovations (especially AI agents), and transformation potential & limitations. After that, I outline what organisations, leaders and individuals can do now to make the most of AI.

1. Adoption & Scale — AI is Everywhere, But Rarely Enterprise-Wide

Widespread adoption, but limited scaling up

According to a major 2025 survey by McKinsey & Company, 88% of companies are already using AI in at least one area of their business – that’s up from 78% the year before. Lighthouse AI Enablement
But only about a third of organisations say they’ve actually started scaling up AI across the whole business. Lighthouse AI Enablement
So many AI initiatives are still just pilots or isolated projects, not fully integrated into the core of how the business operates. McKinsey & CompanyBusiness Acumen

Thus: AI is no longer a niche thing – it’s mainstream – but getting the whole organisation to transform is still a tough ask.

Generative AI adoption is picking up pace

One of the fastest-growing bits of AI tech right now is Generative AI (GenAI):* Generative AI usage in business functions has skyrocketed to 71% in 2025 , way up from 33% in 2023. (According to Kenility & Netguru)

That’s a pretty big indicator that companies are getting a lot more adventurous with AI – and using it for all sorts of tasks that go way beyond just analytics and automation – think content generation, code , design, text – pretty much anything across business functions. (Stanford HAI & Netguru)

The rise of generative AI shows that business is starting to move away from viewing AI as just a tool for specialist jobs – instead its becoming a normal part of business life – content creation, marketing, product dev, all the usual comms stuff.

Still, business impact is surprisingly limited

Only 39% of people surveyed by McKinsey in 2025 could actually put their finger on some measurable impact from A.I on their bottom line (EBIT). (McKinsey & Company)
And even among those that do say AI has made a difference, the impact is usually tiny – in many cases we’re talking less than 5% of EBIT attributed to AI efforts. (Business Acumen)

So, even though everyone’s using it, a real financial transformation thing isnt happening for many organisations.

2. Emerging Innovations – AI Gets Agentic, Workflows Get Smarter

One of the defining themes of 2025 is the rise of agentic AI – AI that can do more than just answer questions: it plans, acts, gets stuff done, and integrates into our workflows.

The AI Agent Boom

In a 2025 survey of some of the biggest names in the AI space – founders, researchers, investors, and engineers – 65% said “AI Agents” was the most important trend on their radar. And they’re not just talking about the fancy stuff, either. SuperAI, Singapore
According to McKinsey, 62% of respondents are already experimenting with AI agents in their business. And a quarter of them have actually scaled an agentic AI system up and running in some part of the business. McKinsey & Company+1

This tells us that the way we think about AI is shifting: from just a backend analytics or automation tool, to something that actually works alongside us in the business.

When Agents, Generative AI, and Multimodal Come Together

AI agents are often powered by foundation models and generative AI systems. When they combine, you get the ability to tackle some pretty complex tasks: whipping up content, crunching data, scheduling, customer service, code generation, and decision support. And with multimodal AI – systems that can handle text, code, images, or structured data – agents can actually work across different domains.

All of this flexibility opens up a whole new world of possible applications: marketing, design, product development, customer interactions, and internal operations.

But, There’s a Caution Flag Waving Here

While agents are on the rise, not every agentic AI project pans out. Experts are warning against “agent-washing” – the tendency to label a simple chatbot or prompt tool as a full-fledged “agent”. Reuters+1

And to make matters worse, most organizations are still in the testing phase – only a small minority have managed to scale agentic AI across the enterprise. McKinsey & Company+2

Even where agents have been adopted, we’re still a long way from seeing any real business transformation – whether that’s a boost in profits, efficiency, or structural change.

3. Transformation: Where the Rubber Meets the Road

What’s Working – Early Signs of Real Change

For the companies that are really taking AI seriously – not just as a bolt-on tool, but as a driver of workflow change – the early signs of transformation are pretty encouraging. According to respondents in 2025:

64% said AI is enabling innovation at use-case levels. McKinsey & Company+1
Some functions – like content creation, customer service, data processing, and internal operations – are already seeing some pretty clear gains.
In sectors like retail, user tests have shown that generative AI enhancements can actually increase sales or improve conversion rates. For example: a recent big field experiment in online retail showed GenAI-based changes increased sales conversion rates. arXiv
And in software development, generative coding tools are becoming increasingly popular: one 2025 research study estimated that AI-assisted coding is writing 30.1% of new Python functions among U.S.-based developers. That’s up from earlier years, and a pretty impressive number. arXiv

These early successes show that where AI is adopted consistently, with some serious workflow redesign and alignment, it can actually deliver some real output – not just some fancy experimentation.

But, There Are Still a Lot of Constraints to Navigate

Only a tiny fraction of companies report that AI has had a material impact on their EBIT. Most gains are still incremental, or limited to specific functions. McKinsey & Company+2
Many organizations are struggling to scale beyond pilot projects. We haven’t yet seen much evidence of companies fully integrating AI into their core operations, multiple functions, or organizational workflows. Lighthouse AI Enablement+1
And then there are the risks – data privacy, regulatory compliance, explainability, integration challenges, change management, and talent gaps. According to 2025 survey data, efforts to mitigate these risks are growing, but still fall short in many organizations. Business Acumen+1
Not all AI agent projects are going to succeed. Some are getting scrapped as unproductive or too costly. Reuters

So, AI holds promise, but scaling up is going to be tough. Without some serious organizational buy-in, strategy, governance, and risk management, many AI investments may not deliver what they promised.

4. What the State of AI in 2025 Means for Business Leaders, Professionals, and Organizations

So what should businesses, leaders, and workers make of this mixed picture – widespread adoption, rising innovation, but limited transformation? Here are some practical takeaways.

a) Treat AI as a Strategic Tool, Not Just a Tactic

AI pilots are okay, but to get any real value out of them, you need to embed them into your workflows, roles, and processes – not just treat them as some kind of side tool.
Think about where AI can really change how work is done – content creation, customer interaction, internal operations, product development. And design those use cases to align with your business objectives, not just convenience.
For agentic AI, start with workflow-friendly tasks (customer support, scheduling, reporting, email triage, internal automation), then scale gradually rather than jumping to high-risk complex tasks.### b) Making the Most of Generative AI
Pairing AI with the right tools is key: we’re talking about combing it with data pipelines, and making sure you’ve got quality training data and domain expertise on tap. And don’t forget to keep an eye on it – oversight is crucial.
Don’t overlook the importance of governance: this is especially true when AI decisions are going to be impacting customers or stakeholders, so you’ve got to make sure you’re on top of things when it comes to things like privacy, compliance, and explainability.
Investing in your people is just as important as investing in AI – we know that machines can’t replace human judgment entirely, but they can definitely help augment it. So put some time and effort into training your teams to use AI responsibly and effectively.

c) Where AI Can Really Make a Difference

The numbers (from 2025 data, and also from McKinsey & Company) are pretty clear: companies that are really getting the most out of AI are the ones that are more focused on growth and innovation than on just cutting costs.
It’s worth saying out loud that AI’s greatest value is often when it’s being used to extend capabilities – so go ahead and use it to create new products, services, or user experiences. Don’t just try to automate the grunt work that you’re already doing.

d) Embracing the Ups and Downs of AI

You need to keep an eye on the metrics that really matter – that’s savings, revenue growth, productivity gains, time savings, reduction in errors, and how happy your customers are.
The last thing you want to do is assume that all functions are going to benefit equally from AI. Some areas – like content, comms, or customer service – might see really quick gains, while other areas – say, complex decision-making domains – might be more of a slow burn.
And let’s be real, not every AI experiment is going to be a winner. But that’s okay – treat those dead ends as learning opportunities and move on with the next idea.

5. Why 2025 is a Pivotal Year — What’s Different Now

Several factors make 2025 stand out compared to previous years. This isn’t incremental; it’s transformative — but uneven.

The rise of agentic AI

For years, AI largely meant analytics, pattern detection, or narrow automation. In 2025, we see a shift toward agentic AI — systems that can plan, act, and carry out multi-step workflows across domains. The majority of AI experts surveyed in 2025 cite AI agents as the top future trend. SuperAI Singapore+1

This shift changes the nature of AI from reactive tools to active collaborators — capable of running tasks, making decisions (within bounds), and integrating across systems. For businesses, that means potential for broader impact.

Generative AI moving mainstream

Generative AI is no longer experimental or niche. Its adoption jumped from 33% in 2023 to 71–78% in 2025 (depending on the study). Fullview+3Kenility+3Stanford HAI+3

That includes a wide variety of functions: content, communication, design, code, analytics, even customer interaction. This broad applicability makes GenAI a core part of many companies’ AI strategy rather than a side tool.

Infrastructure, tools, and ecosystem matured

AI tools, APIs, platforms, cloud infrastructure, compute availability — all of these matured by 2025. That lowers the barrier to entry: even medium-size businesses can use AI meaningfully, without building everything from scratch.

At the same time, governance, risk awareness, regulatory concerns are rising. Companies are increasingly aware that AI isn’t magic — it needs oversight, evaluation, human judgment. Risk mitigation efforts — for privacy, explainability, compliance — are more common than they were just a few years ago. Business Acumen+1

6. What to Watch in the Near Future (2025–2028)

Based on current trends, here are some developments likely to shape AI over the coming 3 years:

More AI agents in enterprise workflows — beyond pilots, into customer service, operations, scheduling, data workflows, decision support. As agents prove value, their use will become more structured and widespread.
Blending of AI and domain-specific workflows — AI will increasingly be embedded in domain workflows (e.g., retail, manufacturing, services), not just tech sectors. Multimodal and agentic AI will power domain-specialist solutions.
Rise of AI governance frameworks, compliance, and ethics — as adoption grows, so will scrutiny from regulators, stakeholders, and the public. Successful AI deployment will require governance, transparency, privacy, and risk management.
Skill shift and workforce adaptation — AI will not just automate tasks but shift roles: more oversight roles, AI-augmented work, hybrid human-AI collaboration. Continuous learning and upskilling will be key.
Uneven global adoption & digital divide — regions with strong infrastructure will lead; others may lag. This may widen the gap between countries or industries in who benefits from AI.

Conclusion — What 2025’s AI Reality Tells Us (and What You Should Do if You Care)

In 2025, AI is no longer a fringe technology. It is embedded in business everywhere: used by the majority of companies, across industries, for many functions. Generative AI and agentic AI are leading the way, turning AI from narrow tools into workflow-integrated systems.

Yet, for most organizations, this is still a journey — not the destination. Real transformation remains rare. Many AI initiatives remain pilot projects; very few produce enterprise-wide impact. Risk, governance, workflow alignment, human adaptability — these remain major hurdles.

That means 2025 is a pivotal year — a moment when businesses can choose whether AI becomes a core part of strategy or just another experiment.

If you care about AI — as a leader, professional, entrepreneur, or simply a curious individual — here’s what matters now:

Be realistic. Don’t treat AI as magic. Treat it as a tool that needs integration, governance, oversight, and human judgment.
Start with workflow — not hype. Identify areas where AI can improve processes, create value, or enable new capabilities. Build pilots, but plan for scale.
Invest in people, data, and processes. AI success depends not only on models, but on data quality, team readiness, governance, and operational integration.
Measure results, iterate, adapt. Track performance, ROI, and business impact. Learn from failures. Scale what works; abandon what doesn’t.
Stay aware of risks — governance, bias, privacy, regulation — especially as AI becomes more autonomous.

2025’s state of AI shows both opportunity and caution. AI is transforming, but not by default — by design. For those willing to build carefully, thoughtfully, and strategically, it remains one of the biggest levers for innovation, growth, and transformation in this decade.

The State Of AI In 2026: Agents, Innovation, And Transformation