Technology in a Digital World

1 The fourth industrial revolution

1.1 Introduction

You will be familiar with the effects of the first three of these. The fourth industrial revolution is no longer 'on the brink' – it is happening now, and it is reshaping what organisations, and their finance functions, do.

YouTube videoPlay video

1.2 Characteristics and dynamics

The fourth industrial revolution differs from the first three in more than just its technology:

• Speed: it is developing exponentially rather than steadily. Generative artificial intelligence, for example, went from a research topic to an everyday business tool in only a few years.

• Scope: it affects every industry and every function – not just manufacturing, but services, healthcare, government and, of course, finance.

• Convergence: the technologies reinforce one another. Sensors (the internet of things) generate big data, which is stored in the cloud, analysed by artificial intelligence and presented through data visualisation. Each technology multiplies the value of the others.

• Data as an asset: earlier revolutions were about making physical products more efficiently. This one is built on digital assets – data, and the ability to turn it into insight and action.

1.3 The key technologies

The World Economic Forum's digital transformation initiative (DTI), and the major advisory firms, identify a core set of technologies that define and drive the digital world:

• Cloud computing: computing power, storage and software supplied as a service over the internet rather than installed on the organisation's own machines. The cloud is examined later in this chapter.

• Big data analytics: analysing extremely large, fast-moving and varied data sets to reveal patterns, trends and associations. Examined in detail in Chapter 4.

• Process automation: whole processes carried out with little or no human intervention. When you order goods online, the picking, packing and despatch of your order is largely automatic. Within finance, robotic process automation (RPA) uses software 'robots' to do rule-based clerical work – matching invoices to orders, posting journals, reconciling accounts – faster, more cheaply and with fewer errors than people.

• Artificial intelligence (AI): machines that can learn from data and perform tasks that used to need human judgement – recognising faces and speech, predicting what customers will buy, detecting fraudulent transactions, and now generating text, analysis and computer code. AI, machine learning and generative AI are examined in the next section.

• Data visualisation: presenting data as interactive graphs, charts and dashboards so that patterns become visible at a glance. With so much data available, visualisation is the art of making it understandable. Examined with business intelligence in Chapter 4.

• Blockchain: a shared, tamper-evident digital ledger. Records ('blocks') are linked cryptographically and copies are distributed across participants, making unauthorised alteration difficult and detectable under the network’s governance. Uses include cryptocurrencies, registers and tracing goods through supply chains.

• Internet of things (IoT): everyday devices fitted with sensors and connected to the internet. A storage bin that senses its own weight can reorder raw material automatically; components moving through a factory can report exactly where they are; a delivery vehicle can report its location and fuel use continuously.

• Mobile: smartphones and tablets connected everywhere through Wi-Fi and 4G/5G networks mean that employees, customers and devices are permanently connected – sales are booked at the customer's premises, approvals happen on the move, and customers expect to transact at any time.

• 3-D printing (additive manufacturing): building objects up layer by layer rather than cutting or drilling them out of raw material. It allows rapid prototyping, one-off custom components and small production runs that would otherwise be uneconomic.

Other technologies often added to the list include autonomous vehicles, robots and drones. You do not need deep technical knowledge of any of them: the syllabus requires awareness of what each does and, more importantly, an understanding of how they change what the finance function does.

2 Artificial intelligence, machine learning and generative AI

2.1 Predictive and discriminative machine learning

Machine learning (ML) is a branch of AI in which software learns patterns from data rather than being explicitly programmed with every rule. ML includes predictive and discriminative models, which score, classify or forecast, and generative models, which create new content. Show an ML model millions of past credit-card transactions, some fraudulent, and it learns to score how likely each new transaction is to be fraud. The same approach powers demand forecasting, credit scoring, customer-churn prediction and the recommendation engines used by online retailers and streaming services.

The key point is the task: this traditional, predictive (or 'discriminative') form of ML classifies things and forecasts numbers based on patterns in past data. Its output is a prediction – a score, a category, a forecast – rather than new content.

2.2 Generative machine learning and generative AI

Generative AI (GenAI) is not a separate technology beyond ML – generative models are themselves machine-learning models, trained for a different task: creating new content (text, analysis, images, computer code) in response to instructions ('prompts') written in ordinary language. The best-known examples are large language models (LLMs), trained on enormous volumes of text, which can draft, summarise, translate, explain and answer questions in fluent prose. The useful distinction is therefore the task, not a hierarchy: predictive ML classifies and forecasts; generative ML creates content. Specialist-built models and easy-to-use end-user tools exist in both categories.

2.3 How finance uses AI

2.4 Limits and risks of AI

AI output looks confident whether or not it is right, so finance professionals must understand its limits:

• Hallucination: GenAI can produce fluent, plausible statements that are simply wrong – invented figures, invented references. Every factual claim in AI-drafted work must be verified before it is relied on.

• Confidentiality: typing company data into a public AI tool may send it outside the organisation's control, breaching confidentiality obligations and data protection law.

• Bias: a model trained on biased historical data repeats the bias – a credit-scoring model trained on past lending decisions can discriminate unlawfully without anyone intending it to.

• Explainability: many models are 'black boxes'. If finance cannot explain why the model reached its answer, it cannot properly defend decisions based on it.

• Accountability: responsibility for an automated output rests with the humans and the organisation that use it – 'the model said so' is no defence to a regulator, an auditor or a customer.

3 How finance uses digital technology

Chapter 2 introduced the 'information to impact' framework: finance collates data, analyses it to produce insight, communicates that insight to influence decisions, and supports implementation so the organisation gets the intended impact. The technologies in this chapter are not a separate topic – each one plugs into a stage of that chain:

Notice the pattern: technology hits the bottom of the chain hardest. Collecting and processing data is exactly the high-volume, rule-based work that machines do better than people – which is why, as Chapter 1 explained, the routine base of the finance function is shrinking, and why the syllabus insists that finance professionals focus on using data rather than gathering it (the subject of the rest of this chapter).

3.1 Ethics of technology usage

Finance professionals are bound by their ethical code (Chapter 12) whether a task is done by a person or a machine. Using technology raises specific ethical duties:

• Competence: do not rely on automated output you do not understand or have not reviewed – professional competence and due care apply to AI-assisted work.

• Integrity of data and models: biased, incomplete or manipulated data produces misleading output; finance has a stewardship role over the quality of the data and models the organisation relies on.

• Transparency: be open about when and how automated tools were used to produce figures and reports, and be able to explain their conclusions.

• Privacy and surveillance: just because data can be collected – about customers or employees – does not mean it should be. Collection and use must be lawful, proportionate and fair.

• Accountability: automation never transfers responsibility. The organisation, and the professionals within it, remain answerable for automated decisions.

4 IT infrastructure choices

4.1 Introduction

Every organisation has to make a set of basic choices about how its IT is put together:

• How should machines be connected? (networks)

• Wired, Wi-Fi or mobile connectivity?

• On-premise or cloud?

• Centralised or decentralised processing?

4.2 Networks

Computers in business are almost never stand-alone: work requires sharing information, cooperating on tasks and communicating.

• Local area networks (LANs) connect machines over a restricted area such as an office, hospital or campus, with shared servers for files, applications and printing.

• Wide area networks (WANs) link sites in different cities and countries over public telecommunications networks. Because traffic crosses networks the organisation does not control, it can be intercepted – so information must be encrypted before transmission and decrypted only by the authorised recipient.

4.3 Wired, Wi-Fi or mobile?

Within a site, machines connect by cable or by Wi-Fi. Away from the office, employees connect over the mobile networks using 4G and 5G technology (3G networks are being switched off). Mobile connectivity is vital for staff who work at customers' premises – sales can be booked and technical information accessed on the spot – and it is what makes real-time, anywhere access to cloud systems possible.

4.4 On-premise or cloud?

Traditionally each organisation bought its own servers and installed its own copies of every application – the 'on-premise' model. In the cloud model, software and data are hosted on a provider's remote servers and accessed over the internet; the local machine is simply a window onto processing that happens elsewhere. Most modern business software – office tools, accounting packages, whole ERP systems – is now sold this way, as a subscription ('software as a service', or SaaS).

The advantages of the cloud approach:

• Pay for use: a subscription scaled to actual usage replaces large up-front hardware and licence purchases, and capacity can be scaled up or down as the business changes.

• Access to powerful processing: demanding applications (data analytics, design work) run on the provider's powerful machines, so local devices can be simple and cheap.

• Always up to date: the provider updates the software centrally, so every user is on the current version – which also closes the security holes that unpatched, out-of-date software leaves open.

• Easier maintenance: the complex processing runs centrally, where the provider's specialists maintain and troubleshoot it.

• Work anywhere: authorised users can reach the software and data from any location on any device – the basis of remote and mobile working.

Disadvantages:

• Complete dependence on the internet connection and on the provider's service staying up.

• Another organisation has custody of the software and data, raising confidentiality, security and regulatory concerns – including which country the data is physically stored in.

• Subscription costs continue for ever, and moving to another provider later can be difficult and expensive ('lock-in').

4.5 Centralised or decentralised?

In a centralised system, a large computer does all the processing and this is connected to smaller ones which really only act as user interfaces.

In a decentralised system each computer in the system does its own processing. A decentralised system is sometimes known as a distributed system because processing power is distributed, or spread, over many machines.

The advantages of one system tend to be the disadvantages of the other.

Advantages of centralised/disadvantages of decentralised

• All data and processing is centralised so is easier to control and safeguard. For example, access and regular backups are centrally controlled. In a decentralised system you would be relying on many users to back-up their files and to ensure that passwords were changed as necessary.
  

• Generally cheaper than a distributed system.
  

• Data is all in one place so is easy to share.
  

• Centralisation will mean that there is less risk of incompatible hardware or software being added.
  

Disadvantages of centralised/advantages of decentralised

• Faster and more flexible response to user needs. Each user can be given power to install suitable software. In a centralised system budget and administration processes can slow down the adoption of new centralised applications.
  

• More resilient to breakdown. If a centralised computer breaks down no user will be able to work. In a decentralised system a machine breakdown will probably affect only one user.
  

5 The ‘web’, the Internet, intranets and extranets

The terms ‘Internet’ and the ‘Web’ are often used interchangeably. Strictly they are different.

The Internet is a vast network of interconnecting networks spanning the globe in which any computer can communicate with any other computer as long as they are both connected to the Internet. Information that travels over the Internet does so via a variety of languages known as protocols.

The World Wide Web (or Web) is a way of accessing information over the Internet. One of the languages used to send information over the Internet is HTTP (Hypertext Transfer Protocol); web services use HTTP to allow applications to communicate in order to exchange and share information. Email uses the internet but is not part of the Web: it uses a language called SMTP (Simple Mail Transfer Protocol).

The Web uses browsers such as Chrome, Edge or Safari to access web-pages that are themselves linked to each other.

Intranets are ‘internal internets’. The only data accessed and displayed is data from within the organisation, but the distribution of the data is via HTTP and is displayed through a browser.

An extranet is when one intranet is given access to another. For example a supermarket system might be given access to a supplier’s system so that stock and orders can be more easily managed.

6 Test your knowledge

Two quick checks before you move on: work through the flashcards to fix this chapter’s key terms and definitions, then sit the objective questions for exam-style practice. Both mark themselves and explain the answers as you go.