AI is the defining technology of the early-21st century. As advances in AI fuel grander expectations of its potential, business leaders are racing to integrate the technology into their operations.

But no two organisations are the same. They have different business models, resources, regulatory responsibilities, levels of organisational complexity and so on. Therefore no single guide can be applied perfectly across industries and organisations; nor can it be truly comprehensive. Nevertheless, when it comes to AI integration, best practices are emerging.

Through the use of case studies, data-led research and expert opinions, not to mention countless interviews with business leaders, academics and consultants, this guide provides business decision-makers with insights and practical guidance on adopting and deploying AI across the organisation. 

The guide is divided into four chapters, each with multiple sections. Section one of this guide briefly discusses the importance of AI for ambitious businesses: why does it matter? And do you really need to care about it? Section two serves as a blueprint for an AI-ready organisation. It covers many of the preliminary considerations that must be made before committing to AI adoption. For instance, what processes and safety protocols should be established to ensure that AI is used safely and effectively? What kind of training should be provided to staff? And what should business leaders expect in terms of implementation timelines and return on investment?

Section three looks at AI deployment across five core business functions. It highlights function-specific use cases for AI, as well as any unique challenges for AI adoption, and examines the potential of the technology to transform various business functions. Section four provides an overview of the risks of AI adoption. These include workforce disruption, regulatory uncertainty and cybersecurity risks, among others.

Tips on using this guide

This guide isn’t the kind of thing you’ll read in one sitting. Fortunately, readers who are logged in will be able to save their progress.

Scattered throughout the guide you’ll find expandable boxes containing full-length interviews, opinions, case studies, debates and explainers. These help to provide context for the main discussion points. The case studies in particular offer concrete examples of AI success stories, which can help business leaders better understand how the practices and principles laid out in this guide can be applied in real life. Readers are strongly encouraged to explore these sections. 

There are several ways to navigate the guide. When you reach the end of a chapter, you’ll see an option in the bottom right-hand corner to progress to the next chapter in the guide. Readers can also jump to different sections or chapters using the floating “contents” menu on the right side of the screen, or the expandable navigation menu in the upper left-hand corner. You can also access each section in the table of contents on the guide landing page.

Helpful terms to get you started

Readers will encounter plenty of technical terms in this guide. Some of these are explained further in short pieces, interviews and opinions we’ve published, which can be accessed throughout the guide.

Before diving in, however, readers may find it beneficial to review a few basic terms:

• Accuracy: Accuracy is a measure of how often an machine learning (ML) model correctly predicts an outcome – the share of predictions that the model got right. Accuracy should be as close to 100% as possible. An accuracy level of 50%, for example, would be the same as random chance.
• Algorithm: An algorithm is a finite set of rules or specifications that defines a sequence of operations. Put another way, it is a set of instructions that can be used to perform calculations or solve certain types of problems. Artificial intelligence (AI) systems use algorithms to process data and complete tasks or arrive at conclusions. 
• Artificial intelligence: AI is the capability of a computer system to mimic human cognitive functions, such as problem-solving, language comprehension and visual interpretation. AI systems use data and algorithms to simulate human reasoning, make decisions and learn.
• Automation: In this guide, “automation” is usually used in the context of business process automation, such as robotic process automation (RPA), where software robots (bots) complete repetitive, rules-based tasks; for instance, issuing invoice reminders or processing service requests. Automation is not synonymous with AI. Automation bots are strictly limited to pre-determined workflows. If a task is not explicitly outlined in its instructions, the bot cannot complete it. AI systems can learn to complete new tasks that have not been outlined in their instructions.
• Bias: Bias refers to the prejudices that may impact an AI’s responses. We recognise biases in the outcomes an AI generates. A bias is often the result of the data that a model is trained on, but could also be a product of the algorithms underlying the model. Biases are most commonly thought of in the context of unfair outcomes that reflect and perpetuate social stereotypes.
• Data: Data are pieces of information that can be processed or analysed to gain insights. They are basic units of meaning, which can take many forms, including images, sounds, numbers and values, all of which can be interpreted formally. Data is the information that is fed into algorithms, which are used by AI to build a knowledge base and complete tasks. Company financials, customer interactions and imagery used in marketing are examples of data sources that could yield business insights.
• Explainability: Explainability is the extent to which humans are able to ascertain the reasoning behind an AI’s decisions. AI systems are sometimes referred to as ‘black boxes’, meaning humans only understand the inputs and the outputs, but are unable to explain why the machine arrived at this or that particular conclusion.
• Generative AI: Generative AI (GenAI) is a type of AI system based on a generative model – probabilistic models that generate an output, often in response to a prompt. GenAI systems are capable of creating original data, including text, images and videos. ChatGPT is an example of a GenAI tool.
• Machine learning: ML is a field of study in AI, which focuses on statistical algorithms that can ‘learn’ from vast amounts of data and generalise to new data that is ‘approximately correct’.
• Model: A model is a representation of a system as a set of rules. For instance, a language model is the probabilistic representation of the rules, derived from data, that describe the structure and operation of a natural language. Large language models (LLMs) are language models that have been trained on vast amounts of data, enabling natural-language processing (NLP), including text and speech recognition, language generation and information retrieval. LLMs underpin many of the most common generative AI applications, such as ChatGPT.
• Prompt: A prompt is a natural-language command or query that describes the task an AI should perform. For instance, a user might command an AI to “summarise these inventory reports.” That simple text command is a prompt.

Business leaders must assume that organisations in the future will rely heavily upon AI. The technology is already transforming industries such as healthcare and software development, but it will have a more fundamental impact on businesses, forcing leaders to question the relevance and competitiveness of their current business models. 

But implementing the technology safely and effectively requires considerable time, effort and investment. Many business leaders will inevitably wonder whether the benefits of using AI justify the cost of adoption.

Of course, the technology’s full potential has not yet been realised. But there’s been a “surprising acceleration” in the development of AI over the past few years, according to Erik Brynjolfsson, senior fellow at Stanford University’s Institute for Human-Centered AI. He believes AI will ultimately be a “quantum leap” for productivity.

For most businesses, AI’s transformative power is in its data processing and automation capabilities, which could benefit any business, regardless of industry or size. For instance, Biffa, a waste-management company, uses an AI-powered automation tool to manage invoice payments and assess credit risks; JPMorgan reportedly can complete thousands of hours’ worth of legal-contract reviews in just seconds using an AI-powered contract-analysis tool; and Phoenixfire, a graphic design agency, uses GenAI to assist with developing rough drafts of creative work.

Anecdotes aside, about three in four (74%) organisations say their investments in GenAI and automation have met or exceeded expectations, according to 2024 research by Accenture. And, more than three in five (63%) plan to further develop their AI capabilities over the next few years.

Many senior leaders emphasise the importance of developing a strategy and adopting AI tools as early as possible, as doing so will enable organisations to experiment and become familiar with AI. This is important as AI systems are likely to progress rapidly if the past few years are any indication. 

However, many experts urge caution. Peter van der Putten, assistant professor of AI at Leiden University, says any business considering whether to implement AI must ask: “Will this tool automate manual work at sufficient levels of quality? Will it improve life for customers and employees? Will it lead to better business outcomes?”

Understanding the risks

Some business leaders can provide positive answers to these questions. This guide includes plenty of real-world success stories of businesses that are using AI to increase productivity, reduce costs and improve commercial performance.

But AI tools must be deployed carefully. There are significant risks that must be considered in any AI strategy. Implementation cannot be a haphazard effort.

The technology has opened up a new frontier in cybersecurity, for instance. Not only is AI empowering cybercriminals, but it is also a new attack surface through which an organisation can be breached.

Even when AI is kept secure, its outputs can’t always be trusted. Sloppy training data can produce biased and inaccurate outcomes. At best, such outcomes could be slightly embarrassing for an organisation. At worst, they could help to perpetuate harmful stereotypes and exacerbate social inequalities.

Then there’s the threat to the labour force. Dr Carl Frey, an economist, and Michael Osborne, a professor in machine learning, both at the University of Oxford, believe that at least 40% of current jobs will be lost to automation by 2050.

Moreover, integrating AI systems into existing workflows can be expensive and time-consuming. There is a worrisome shortage of AI skills at organisations and many leaders believe this skills gap is delaying their firms’ AI efforts.

Most experts agree however that these risks should not lead C-suite leaders to shun AI. According to Simon Baxter, principal analyst at TechMarketView, a technology research firm, AI is not all hype. “It’s quite clear that the gains you get from the current range of GenAI solutions is very real,” he says.

These dangers however highlight the need for careful planning and robust governance. AI has the potential to unlock new planes of productivity and commercial opportunity. But ill-considered implementation could be ruinous. Leaders must put in the effort if they are to maximise the benefits of the technology.