AI terminology for business people

Agentic AI

Autonomous, independent AI systems which can make decisions and take actions. They can learn and improve. Unlike task-specific AI, which follows predefined instructions, agentic AI plans, adapts and responds to the environment to achieve goals.

AGI (Artificial General Intelligence)

AI capable of understanding, learning and applying knowledge in a similar way to human intelligence. Some people are saying that AGI is about to happen, whilst others say that it may not ever be achieveable.

AI agent: see Agentic AI

AI adoption

The process by which an enterprise integrates AI into its business operations, workflows and way of thinking.

AI alignment

In artificial intelligence, alignment refers to guiding systems so that they act in line with your goals, preferences and ethics. A system is aligned when it supports your aims, whilst a misaligned system follows unintended ones.

AI avatar

An AI avatar is a digital portrayal of a human within a virtual environment. Within a marketing context, they can be used for customer service, product or service demonstrations, personalised shopping and sales presentations.

AI Council

A group of people within an organisation, probably from different departments, who guide the development of AI within the organisation. The Council ensures that AI is aligned with business goals and standards.

AI expertise

Development of in-house AI knowledge. This can be achieved through the development of an AI knowledge centre, training programmes and hiring AI professionals.

AI generated content

Marketing content such as blogs, images or videos which have been produced by generative AI (i.e. Claude, ChatGPT).

AI governance

Policies, procedures and controls to ensure responsible, ethical and compliant use of AI within an organisation.

AI leaders

According to IBM, “Leaders look for the intersection of opportunity, need and internal capabilities to develop an action-oriented roadmap; foster organisation-wide alignment through clear and authentic communication; and understand that a strong data foundation delivers the flexibility to customise AI.”

AI ROI

The value produced from an AI investment, in relation to costs incurred.

AI strategic plan

Organisations of all sizes should create an AI strategic plan. This is not just another software implementation. It requires a change management approach.

Algorithm

An algorithm comprises rules that a computer follows to complete a task. It takes an input, like a dataset and generates an output, such as detecting patterns in the data. Algorithms are used in chatbots and social media platforms.

Anthropic

The company behind Claude AI.

API

(Application Programming Interface): A set of protocols and rules which allow different software applications to communicate with each other.

Artificial Intelligence

Computer systems capable of executing tasks that usually necessitate human intelligence, such as visual comprehension, recognising patterns, speech recognition, language translation, learning, reasoning, problem-solving and decision making.

ASI (Artificial Super Intelligence)

AI that goes beyond human intelligence. It would have the ability to improve itself rapidly and outperform people in most cognitive tasks. ASI remains theoretical. See IBM’s article: What is artificial superintelligence?

Augmented reality

AR adds digital items including images, sound, text or animations on top of what you see in the real world. It does not replace reality, like Virtual Reality does. Instead it mixes real life with digital content. You can use AR with devices such as smartphones, tablets, smart glasses or headsets.

Automation

AI automation is the use of artificial intelligence to carry out tasks or processes automatically, without the need for human intervention. It involves applying AI methods, such as machine learning, natural language processing to automate repetitive or complex activities. See also: Intelligent Automation

Big Data

Datasets of such substantial size or complexity that conventional data processing applications cannot handle them effectively.

Bitcoin

Bitcoin is a type of digital currency that does not rely on banks or governments. It was launched in 2009 by an unknown person or group under the name Satoshi Nakamoto. New bitcoins are created through a process called mining, where computers solve difficult mathematical problems. All transactions are stored on a public record known as the blockchain, which is shared and maintained by a network of computers worldwide.

Bitcoin can be used to pay for goods and services, send money across borders, or kept as an investment. Its price is highly volatile, which makes it appealing to some but risky to others.

Bitcoin and blockchain are being investigated and applied to support transactions in AI-related systems, includinh potential uses such as paying for AI-generated tasks, powering data marketplaces and serving as a transaction currency for AI agents in the future. There are initiatives exploring how Bitcoin’s features, such as its decentralised ledger and capacity for micro-transactions could be integrated into AI applications and emerging digital economies.

Brain Machine Interface

A BMI, or brain–computer interface (BCI), enables direct communication between the brain and an external device. The objective is to translate neural activity into commands that can control computers, prosthetics or other machines. In some instances,  information can be fed back into the brain. Research combines neuroscience, computing and bioengineering.

Business Intelligence

BI comprises technologies and processes which collect, analyse and present business data to support decision making. Increasingly, BI is enhanced by AI capabilities.

Chatbot

A chatbot is a software application created to engage with people via text or voice, emulating human-to-machine conversation.

ChatGPT

(Chat Generative Pre-trained Transformer): A free to use, publicly accessible chatbot. You type in a question, it types out an answer.

Claude

An LLM (like ChatGPT). Built by American company Anthropic.

Codebase questions

Enquiries regarding the structure, functionality and behaviour of a software’s source code. They can range from high-level understanding of the architecture to granular level details concerning specific code sections. The result is a better understanding of how the code works, how to change it and how to solve problems.

Computer vision

Machines which can interpret what they see and make decisions. Examples include quality control for assembly lines and image recognition systems in self-driving cars.

Context aware AI agent

An AI system that understands and adapts based on business or operational context. It can interpret information regarding the situation, person or task and make more infomed decisions or give better answers. For example:
* A context-aware sales assistant agent recalls a customer’s previous conversations and mentions a relevant special offer
* A customer service agent detects that the customer is not happy, so it changes its questions in order to try and resolve the situation.

Context window

This looks like a search bar within ChatGPT and other generative AI systems. It is where you write your question or enter your command.

Deep learning

Deep learning (which is used by ChatGPT) is a subset of machine learning. It uses structures called neural networks to find patterns within data and generate outputs. Deep learning neural networks draw inspiration from the organisation of the human brain, featuring multiple layers of computational units referred to as neurons. They are particularly well-suited for intricate learning tasks, such as image feature extraction and speech analysis.

Emergent capabilities

These are unanticipated features or abilities that an AI system develop during its design or use. They may involve creating new strategies, interpreting novel data types, or solving problems in unexpected ways. These capabilities result from the system’s interaction with its data and environment.

Large Language Model (LLM)

An LLM understands and generates natural (human) language. It is an Artificial Intelligence (AI) algorithm. Deep learning techniques and large amounts of data are used to generate new content. Examples include ChatGPT from OpenAI and Claude from Anthropic.

Generative AI

See Large Language Model.

GPT

A GPT in ChatGPT is a Generative Pre-trained Transformer, which is a type of artificial intelligence model designed to generate human-like text by predicting the next word in a sequence based on the context of the preceding words.

Hallucination

An AI hallucination is when an artificial intelligence system produces information that appears plausible but is false or misleading.

Human-in-the-loop

An approach where humans oversee, intervene or guide AI processes to improve accuracy, ethics and reliability.

Inference

In AI, inference means the process of using a trained model to make predictions or decisions based on input data.

Intelligent Automation

IA enhances business operations by replacing manual tasks with a combination of AI, robotic process automation, business process management and other technologies. Sometimes called hyperautomation, IA extends beyond routine automation by using advanced methods such as cognitive computing and data mining to support smarter and more adaptable decision-making.

Long-horizon reasoning

The ability of an LLM to comprehend and respond to complex, multi-step tasks or scenarios which require a sequence of steps over a long time period.

Machine learning

Helping machines to discover their own algorithms to solve problems, without needing to be told what to do by human-developed algorithms.

Natural Language Processing (NLP)

Giving a computer the ability to understand human language, both in text and spoken form.

Neural network

A system that comprises a collection of nodes that seek to mimic the processing abilities of the human brain.

OpenAI

OpenAI is the company behind ChatGPT. It has received significant investment from Microsoft.

Optical Character Recognition

OCR converts printed or handwritten text in images or scanned documents into editable, searchable digital text.

Prompts

The instruction or command that you enter into the context window. A great deal has been written about prompt engineering and how to write effective prompts. In broad terms, if you are specific in your prompt, you will get better results.

Replit

Replit is an online platform where you can write, test and run computer programs directly in your web browser, without needing to install anything on your computer. It’s like a digital workspace for coding, making it easy for beginners or anyone to create and experiment with software projects from anywhere.

Responsible AI

The idea that AI developers and users should ensure that AI is developed and used ethically.

Robotics

AI robots are machines that combine physical robotics with AI, enabling them to sense their environment, process information and act with an element of autonomy.

Synthetic data

Artificially generated data that emulates the properties of real-world data. It is created via algorithms, simulations, or machine learning models rather than collected from the real world. Synthetic data is used in fields such as machine learning, data analysis and testing.

Tokens

Large Language Model tokens are the textual units used by models such as ChatGPT to comprehend and analyse language. Tokens can be as short as a single character or extend to the length of an individual word. The total count of tokens within an input text impacts several aspects, including cost and processing duration. Most LLMs impose a maximum threshold on the quantity of tokens which they can accommodate in a single interaction.

Unstructured data

Does not conform to a data model and lacks a formal structure. Unstructured data are often more like ‘normal’ business information, for example CRM notes.

Use cases

A usage scenario, or in other words, how AI / ChatGPT can help you. The term is often applied to software (“it can do this and it can also do that”).

Vector databases

Vector databases store data as numerical vectors, allowing for efficient similarity searches and pattern recognition in AI and machine learning applications. Unlike traditional databases, vector databases are optimised for handling high-dimensional data and identifying similar items based on context, rather than exact matches. This makes them essential for tasks such as recommendation systems, semantic search and LLMs.

Vibe coding

Vibe coding is an AI-driven method of software development. Developers and non-technical users write natural language (i.e. English) commands to direct the AI in generating, improving and debugging code. Rather than writing code line by line, the focus is on describing the desired outcome (the ‘vibe’) whilst the AI manages the technical details. This creates an iterative, experimental workflow that resembles a collaborative ‘pair programming’ partnership between developer and AI.

Vibe coding is accelerating software development; lowering barriers to coding; enhancing creativity and experimentation; improving collaboration; supporting rapid prototyping; enabling personalised software solutions; transforming education.

Weights

In AI, especially in neural networks, weights are numbers that control how much one part of the network affects another.

When the AI learns, it changes these weights to get better at its task.

Every connection between neurons has a weight.

The weight decides how strong the signal is from one neuron to the next.

Simply put, weights help the AI decide what information is important and how to use it.

I hope that you have found AI terminology for business people helpful. Nigel Temple AI Futurist speaker.