AI in financial services: terminology, decoded.

This term refers to the situation where many people do not have access to professional financial advice that could help them make better decisions about their money.

The Advice Guidance Boundary Review (AGBR) is a joint initiative between the FCA and HM Treasury examining the regulatory boundary between regulated financial advice and other forms of consumer support. It aims to close the UK's advice gap by introducing new ways for firms to help consumers make decisions about pensions and retail investments. Targeted support and simplified advice are the two main propositions emerging from the review.

An agent is an AI system designed to perform specific tasks autonomously on behalf of a user. Unlike simple chatbots that respond to queries, agents can take actions, make decisions within defined parameters and complete multi-step workflows without constant human intervention.

Agentic AI refers to artificial intelligence systems that can act autonomously to achieve defined goals, making decisions and taking actions without requiring step-by-step human guidance.

AI governance encompasses the frameworks, policies and processes that organisations establish to ensure artificial intelligence is deployed responsibly, ethically and in compliance with regulatory requirements.

AI hallucination refers to when an artificial intelligence system generates false or incorrect information, presenting it as if it is true.

An API (Application Programming Interface) is a set of protocols that allows different software systems to communicate and share data with each other.

An API key is a unique identifier used to authenticate and authorise access to an API. It acts as a secure credential that verifies the identity of the application or user making the request and controls what data or functionality they can access.

Artificial Intelligence (AI) is a branch of computer science that develops computers and machines to process tasks that would otherwise be performed by natural (human) intelligence.

Aveni Assist is an AI productivity tool built for financial advisers. It captures and analyses adviser-client conversations, generates suitability report drafts, fact-find summaries and meeting notes, and integrates with Intelliflo Office and Xplan, alongside MS Teams, Google Meet, Zoom and Webex. Aveni Assist is designed to reduce the administrative burden on advisers so they can spend more time with clients.

Aveni Detect is an AI-powered quality assurance and compliance monitoring platform for financial services firms. It analyses 100% of adviser-client interactions across voice and digital channels, surfacing risk, vulnerability indicators and Consumer Duty evidence at scale. Detect helps QA, compliance and risk teams move from sample-based reviews to full-population oversight.

Bias in AI systems occurs when models produce systematically skewed outputs that unfairly favour or disadvantage certain groups or perspectives.

Chatbots are computer programmes that can have conversations with humans using natural language. They are designed to simulate how a human would behave as a conversational partner, and can respond to questions, requests, and general conversation.

The Financial Conduct Authority’s (FCA) Consumer Duty regulation is a set of rules that requires financial firms to prioritise their customers’ interest when providing services and offering financial products.

The Consumer Duty cross-cutting rules are three behavioural standards that sit under Principle 12 of the FCA Handbook. They require firms to act in good faith towards retail customers, avoid causing foreseeable harm, and enable and support customers to pursue their financial objectives.

When it comes to the Financial Conduct Authority’s (FCA) Consumer Duty regulation, customer outcomes refer to the impacts that financial products and services will have on its customers.

Deep Learning is a type of Machine Learning (ML) that uses Artificial Intelligence (AI) to create neural networks that are modelled after the human brain.

FG21/1 is a guidance issued by the Financial Conduct Authority (FCA) focused on the fair treatment of vulnerable customers by financial firms.

Financial inclusion ensures customers are able to access useful and affordable mainstream financial products and services that best meet their needs.

Generative AI refers to Artificial Intelligence capable of creating new content inspired by, but not identical to, human-generated content.

Hallucination occurs when an AI model generates information that appears plausible but is factually incorrect or entirely fabricated.

Horizontal AI, often called generic AI, is an AI system designed to perform broadly across multiple domains instead of being specialised for a single area, such as vertical AI.

Human in the Loop (HITL), in the context of Machine Learning (ML), is a model or programme that requires human interaction and intelligence to ensure the correct results.

A Large Language Model (LLM) is a type of Artificial Intelligence (AI) that can process human language.

Machine Learning (ML) is a type of Artificial Intelligence (AI) that allows software applications to carry out actions that they’ve not been explicitly programmed to do.

A meeting agent is an AI system designed to automatically capture, analyse and document adviser-client conversations.

The Mills Review is a forward-looking FCA review into the long-term impact of advanced AI on retail financial services through to 2030 and beyond.

Misalignment refers to AI systems producing outputs that deviate from financial services regulatory requirements, industry standards or firm policies, even when technically functional.

In Natural Language Processing (NLP), a model is an AI system trained to understand, interpret, and generate human language.

Natural Language Processing (NLP) helps machines to better understand human language. Computers can easily identify keywords and understand the literal meaning of words, but it lacks the nuance of human speech patterns, so it misses things like sarcasm.

Outcome Testing Teams (OTT) are teams within financial services organisations that assess and analyse the real-world impacts of Artificial Intelligence (AI).

Regulatory Technology or “Regtech” helps financial institutions to navigate the ever-changing landscape of regulations and compliance. It ensures that they are always operating to the most up to date requirements.

Retrieval Augmented Generation is a technique that enhances AI model responses by combining generative capabilities with real-time information retrieval.

A risk taxonomy is a structured framework that defines and categorises the risks an organisation faces. In financial services, it groups risks into categories such as conduct, operational, financial crime, regulatory and technology risk, with each category broken down into specific risk types.

A sandbox is a controlled testing environment that allows firms to trial new products, services or technologies under regulatory supervision but without the full set of authorisation requirements that would normally apply.

Sentiment analysis is a technique in Natural Language Processing (NLP) used to identify the emotional tone of spoken or written words.

SM&CR (Senior Managers and Certification Regime) is the UK regulatory framework that holds individuals accountable for their conduct and competence within financial services firms.

Small language models (SLMs) are AI models trained on a smaller volume of parameters than large language models, typically optimised for a narrower set of tasks or a specific domain.

Sovereign AI refers to artificial intelligence systems that operate within a specific jurisdiction's regulatory and data governance framework, ensuring data sovereignty and compliance with local laws.

Speech analytics does exactly what it says on the tin. It analyses speech. It sounds a lot like Natural Language Processing (NLP) because NLP is a tool of speech analytics.

A suitability report is a document produced by a financial adviser that explains why a specific product or recommendation is appropriate for a client’s individual circumstances. It outlines the client’s objectives, financial situation and risk profile, then sets out the adviser’s reasoning, any alternatives considered and the associated costs. Suitability reports provide transparency for clients and evidence for regulators that advice is tailored, compliant and in the client’s best interests.

Synthetic data is artificially generated information that mimics the statistical properties and patterns of real data without containing actual client or business records.

Targeted support is a new FCA regulated activity that allows authorised firms to make ready-made suggestions to groups of consumers who share common characteristics, sitting between generic guidance and full personal advice.

A task registry is a structured catalogue of the specific actions and workflows that an AI agent can perform. It defines what tasks are available, the parameters each task requires and the conditions under which tasks can be executed.

Toxicity in AI refers to harmful, offensive or inappropriate content that a model might generate or fail to identify.

Transient vulnerability is a temporary or short-term period of vulnerability for a customer, such as someone going through a divorce or the loss of a loved one.

Vertical AI are artificial intelligence systems designed for a specific industry. Unlike generic AI, which aims to work well within a variety of platforms and situations, vertical AI focuses on excelling in one narrow area. 

The Financial Conduct Authority (FCA) definition of a vulnerable customer is someone who, due to their personal circumstances, is especially susceptible to harm, particularly when a firm is not acting with appropriate levels of care.