Emerging AI Terms (2026): A Glossary

Part of the Five Stones knowledge base - 27 terms on emerging AI terms for 2026, in plain English.

Part of the Five Stones knowledge base - 27 terms on emerging ai terms (2026), in plain English.

AI-native business - A business built from the ground up around AI-driven processes, rather than one that has bolted AI onto an older way of working.
Agentic commerce - AI agents autonomously completing purchase or booking transactions on a customer's behalf, an emerging pattern in online retail and services.
AI receptionist - An AI voice or chat system handling incoming calls and enquiries the way a human receptionist would, now common in clinics and trade businesses.
Vertical AI - AI products built specifically for one industry (dental clinics, logistics, law firms) rather than general-purpose tools adapted after the fact.
Small Language Model (SLM) - A smaller, more efficient AI language model, often used where speed, cost or running on a local device matters more than raw capability.
On-device AI - AI that runs directly on a phone, till or camera rather than sending data to the cloud, often chosen for speed, cost or privacy.
AI browser / agentic browser - A web browser with a built-in AI agent that can carry out multi-step tasks (research, booking, form-filling) on a user's behalf.
Model Context Protocol (MCP) - An emerging open standard that lets AI models connect consistently to different business tools and data sources.
Agent-to-agent (A2A) protocol - Emerging standards that let AI agents built by different vendors communicate and hand off tasks to each other directly.
Reasoning model - A newer type of AI model specifically built to work through complex, multi-step problems more reliably before answering.
Test-time compute - Extra computing power an AI model uses at the moment of answering (rather than during training) to reason more carefully through a hard question.
Shadow AI - Staff using AI tools at work without company approval or oversight, an emerging governance risk for many SMEs.
AI sprawl - A business ending up with many overlapping, unmanaged AI tools adopted by different teams, without any central strategy.
Composable AI stack - Building a business's AI capability from several smaller, interchangeable tools and models rather than one single locked-in platform.
AI-augmented workforce - Staff whose day-to-day work is meaningfully assisted by AI tools, rather than replaced by them, the most common real-world pattern in SMEs today.
Responsible AI - The practice of deploying AI in ways that are fair, safe, transparent and accountable, increasingly expected by customers and regulators alike.
AI regulation - The growing body of law and guidance (varying by country) governing how AI can be built, sold and used.
Synthetic media - Video, audio or images generated or altered by AI, including deepfakes, increasingly relevant to fraud risk and marketing disclosure.
AI watermarking - Embedding a signal in AI-generated content to identify it as machine-made, an emerging response to synthetic media concerns.
Token efficiency - How economically an AI system uses tokens to produce a given result, directly affecting the running cost of an AI tool at scale.
AI adoption curve - The pattern by which businesses in an industry move from ignoring AI, to experimenting with it, to depending on it, usually far slower than the technology itself changes.
Autonomous workflow - A workflow where AI agents carry out every step end-to-end, with a human checking outcomes rather than approving each action.
Multimodal search - Searching using more than one input type at once, such as a photo plus a typed question, increasingly supported by AI search tools.
AI cost optimisation - Deliberately managing which AI model or tool is used for which task so a business is not overpaying for capability it does not need.
Composable automation - Building automation from small, independent, swappable pieces rather than one rigid, all-or-nothing system.
Trust layer (AI trust) - The combination of guardrails, human review and transparency a business puts around an AI system so its output can be relied on.
Digital twin - A live, data-driven virtual model of a real business process or asset, used to test changes or predict outcomes before they happen in reality.

Looking for a different topic? Browse the full knowledge base, or tell us what's missing.