Artificial intelligence began as a way to make sense of huge amounts of data. Back in the late 90s and early 2000s, companies like Tesco and Amazon harnessed Machine Learning (ML) models to predict which products or services customers are likely to buy next. They started to use this insight to create personalised recommendations, drive marketing campaigns and improve operational forecasting.
For two decades, AI and ML systems focused on prediction. Today they are dramatically more involved, making content, answering questions and even running parts of businesses without waiting for commands.
This journey has unfolded in three clear phases: prediction, creation and independent execution. If you run a business today, it is vital to understand each phase and know how to make it work for you.
The First Wave – Machine Learning’s Predictive Power
The very first introduction of AI in business was way back in the early 2000s. At that time, Machine Learning (ML) models became common in retail and logistics. They were used to forecast demand, optimise pricing and personalise recommendations. For instance, Netflix began using predictive models for suggestions in 2006. Similarly, Uber introduced surge pricing in 2012. By 2015, companies across the banking, energy, and warehousing sectors relied on forecasting tools.
Today, these systems are not futuristic. They are essential. Cloud services from Amazon, Microsoft and Google simplified implementation. Small businesses no longer need specialist data science teams to run these models.
A useful example is Great Railway Journeys in the United Kingdom, as reported in the case study published by us at Enablis. In 2025, they started using a pricing model using Amazon AWS platform services. The AWS platform offers a range of well known ML models which were trained and tuned to support the customer context. By analysing booking history, weather patterns and holiday schedules, Great Railway began adjusting prices instantly. That effort led to a twelve per cent rise in pricing accuracy.
The Second Wave – Generative AI & LLMs
The next seismic shift came in late 2022 with the explosive debut of ChatGPT. Practically overnight, a new breed of generative AI – known as Large Language Models (LLMs) – burst onto the scene, capable of crafting emails, documents, and content with human-like fluency in response to business queries.
In a flash, routine tasks like writing marketing copy, replying to customer inquiries, or summarising internal reports could be handled in seconds. The response was electric: ChatGPT rocketed to an estimated 100 million monthly users by January, making it the fastest-growing consumer application in history.
In 2023, global spending on generative systems reached 154 billion dollars, according to IDC. And close to 100 million users began interacting with chat tools in early 2023. By mid-2024, Clay AI, DeepSeek and other firms had built dedicated tools for sectors such as medicine, law enforcement and education.
In call centres, for example, an estimate from October 2024 showed that automated systems were handling seventy per cent of routine queries. That means fewer hold times, lower labour costs and happier customers.
The point is clear. These tools are ready now and offer real improvements in efficiency and quality.
The Third Wave – Agentic AI & Automated Action
We are now entering the third phase – tools that not only predict and create but also take action. In late 2024 the Model Context Protocol (MCP), an open standard created by Anthropic (behind the Claude family of models) was released to streamline how AI models connect with tools, systems, and data, think of it like a universal “USB-C” for AI. By enabling seamless, model-agnostic integration enables a new era of AI interoperability, making it a key consideration for any forward-looking tech strategy.
Tools operate with autonomy. They perceive business conditions, weigh options and make accurate and intelligent decisions with minimal human interaction. In supply chains, these systems show strong results. A McKinsey report found that organisations using independent decision systems cut logistics costs by 15%.
In finance departments, similar tools automatically process invoices and employee expenses. Some retailers have tools that automatically arrange deliveries when warehouses run low. That kind of automation frees up staff to concentrate on higher-level tasks where human judgment still matters.
This is no longer about making informed suggestions. It is about systems carrying out tasks and improving speed as well as reliability.
What Businesses Need to Do to Get Ready
According to the Gartner Hype Cycle, artificial intelligence has moved past the Peak of Inflated Expectations and is now entering the Slope of Enlightenment, where real business value starts to emerge. Mature models and scalable cloud platforms are making it easier for companies to apply AI to solve practical problems.
Machine learning (ML) has largely settled into the Productivity Plateau, delivering consistent, proven results in applications like fraud detection, forecasting and recommendations.
Large language models (LLMs) and agentic tools have moved beyond the initial wave of hype into a more mature and grounded phase. The technology and tooling are now significantly more reliable, and practical use cases—like task automation, content generation, and intelligent assistance—are much better understood and are actively being deployed. With a clearer grasp of both their capabilities and limitations, businesses are now positioned to harness the real-world value of generative AI.
Businesses should continue to adopt well-understood and proven ML for reliable while carefully exploring LLMs and AI agents with realistic expectations. The goal is to balance proven AI capabilities with cautious innovation in emerging technologies.
There are three key areas to get right.
- Team
First, put together the right team. You will need data specialists, prompt engineers who shape inputs to produce useful outputs. Also, you need system integrators who connect tools to your websites or apps and ethics-focused people who monitor fairness, transparency and compliance.
- System
Second, your systems and data must be ready. That means clean, well-organised data that is properly stored and easily accessed. It means your operations, from supply chains to customer service, can plug into these tools quickly. To get started quickly we recommend using Cloud platform service from vendors including AWS and Azure they provide a range of models and offer advanced features such as Resource Augmented Generation (RAG)
- Mindset
Third, adopt a careful and considered mindset. Make sure you have oversight of what the systems do. Watch for bias or unfair outcomes. Design a clear process that helps people understand how and when tools act. Start small, monitor results and expand slowly.
Conclusion – The Tech Isn’t New – But What It Can Do Is
Systems that forecast, craft and act form a powerful trio for business. The tools are in place. ML models allow businesses to predict and see what is coming. LLM tools can explain, summarise, communicate. Agentic AI systems can perform tasks and take actions without delay.
The companies gaining the most are those that are already using these tools and are ready to adapt to the latest trends in AI technology that are yet to come. They have prepared data and systems. They have built the right teams. They monitor responsibly. They do not see these systems as extra features. They see them as partners in performance.
So now you face a choice. You can wait and hope your competitors don’t move first. Or you can take action and embrace the AI technology trends. Start small. Build capability. Grow steadily.
Because this technology is ready. And, it’s only going to get more advanced as time goes on.
