AI and automation glossary

Software that carries out a multi-step task on its own: reading an email, working out what it means, and taking the next action, not just answering.

When an AI model states something confidently that is actually wrong or made up, rather than saying it doesn’t know.

A cloud service from Amazon that gives businesses access to several AI models in one place, with strong security controls built in.

A defined way for two pieces of software to exchange information, like a CRM sharing contact data with an AI agent.

A program that replies to messages in a conversation. Most chatbots can only answer questions, unlike an AI agent, which can also take action.

An AI agent that runs on remote servers rather than on a business’s own computers, so it’s accessible anywhere and doesn’t need local hardware.

Recording exactly what an AI agent said and did in every interaction, so there’s a clear record if it’s ever reviewed.

The amount of information an AI model can consider at once when generating a response.

Ongoing work to keep an AI agent’s running costs reasonable, by watching usage and fixing anything wasteful.

The software a business uses to track contacts, leads, and customer interactions in one place, like HubSpot or Salesforce.

A numeric representation of a piece of text that captures its meaning, used behind the scenes so an AI agent can search by topic.

Google’s cloud platform for building and running AI models, another option AI Abstraction uses depending on what a project needs.

A setup where an AI agent handles most of a task but hands off to a person for anything it shouldn’t decide alone.

The collection of a business’s own documents, like manuals, FAQs, and policies, that an AI agent is trained to answer from.

The AI model that reads and writes language. It is the “brain” behind an AI agent, trained on huge amounts of text to understand requests and generate responses.

How long it takes an AI agent to respond after receiving a request. Lower latency means a faster, more natural-feeling interaction.

Automatically filling in missing details about a new contact, like their company or job title, so a sales team has full context right away.

Sorting new leads by how likely they are to become a customer, so sales time goes to the ones worth pursuing first.

The broader technology where a computer improves at a task by learning patterns from data, rather than being programmed with fixed rules.

The ongoing service of running, watching, and maintaining an AI agent after it launches, so it keeps working correctly without a business having to manage it.

The broader field of AI focused on understanding and generating human language, which large language models are a modern, powerful version of.

The instructions given to an AI model that tell it how to behave, what to focus on, and how to respond.

The ongoing process of adjusting an AI agent’s instructions so its answers stay accurate as a business, its products, or its policies change.

A method that lets an AI agent look up a business’s own documents before answering, instead of guessing from general knowledge alone.

A written commitment for how fast a problem will be responded to, based on how serious it is.

A small chunk of text, roughly a word or part of a word, that an AI model processes and that usage costs are typically billed by.

Automatically checking that an AI agent is running correctly, around the clock, so problems get caught before a customer notices.

A specialized way of storing documents that lets an AI agent search by meaning, not just by matching exact words.

An AI agent that handles phone calls, answering questions or taking action by voice instead of through text or chat.

A message one system sends to another automatically the moment something happens, like a new form submission, so nothing has to be checked manually.

Setting up a repeated business task, like data entry or follow-up emails, to happen automatically instead of by hand every time.

Connecting multiple business systems, like a CRM, calendar, and email, so one action updates everywhere it needs to, automatically.