"Agentic" is the most abused word in the category, so here are the terms with their meanings attached: what the words actually promise, how to tell the real thing from the relabeled one, written by a team building the real thing.
An agentic experience layer is the software layer between your traffic and your website that AI agents can operate directly. It scores each visitor's in-session behavior, decides which experience the page renders, and exposes every operation — segments, variants, placements, measurement — as tools an agent can call. AXO is an agentic experience layer for website personalization.
Read the full definition →Agentic personalization is website personalization operated by AI agents instead of clicked together by hand. An agent builds segments, drafts variants, wires placements, and reads lift through tool calls, while a deterministic runtime decides what each visitor sees in-session. The agent operates the system; it does not improvise in the live decision path.
Read the full definition →Agentic software is software built to be operated by AI agents: its real operations are exposed as callable tools — via protocols like MCP — that an agent can invoke, compose, and verify. Agentic software is not the same thing as an AI agent. The agent is the driver; agentic software is the machine it drives.
Read the full definition →Agent tooling is the set of callable tools a software product exposes so AI agents can operate it: the functions, schemas, and permissions that turn a product into something an agent can drive. For marketing teams, agent tooling determines whether a platform can be run by Claude or ChatGPT — or only by hand, in the vendor's UI.
Read the full definition →A holdout test measures the true impact of a marketing intervention by randomly withholding it from a slice of the audience. The held-out group experiences the site or campaign without the treatment, at the same time as everyone else, and the difference in outcomes between the two groups is the lift the intervention actually caused.
Read the full definition →Incrementality is the portion of an outcome — conversions, revenue — that a marketing activity actually caused, beyond what would have happened without it. It is measured by comparing a treated group against a randomly held-out control group running at the same time. Anything short of that comparison measures correlation, not contribution.
Read the full definition →An agentic CRM is a customer relationship management system designed to be operated by AI agents. The agent does the record-keeping — logging touchpoints from email and calendar, maintaining deal timelines, suggesting pipeline stage moves — while humans approve what changes. What makes a CRM agentic is its tool surface, not its database.
Read the full definition →The Model Context Protocol (MCP) is an open standard, introduced by Anthropic in 2024, for connecting AI models to software. A product ships an MCP server that describes its operations as callable tools; any MCP-capable agent can then connect and operate the product. MCP is to AI agents what the API was to integrations.
Read the full definition →In-session personalization adapts a website for a visitor using only the behavior they exhibit during the current visit — pages viewed, scroll depth, dwell time, sequence — rather than a stored profile or resolved identity. Because it requires no login, CRM record, or third-party cookie, it works for the anonymous majority of traffic.
Read the full definition →Cookieless personalization adapts website experiences without third-party cookies or cross-site identifiers. Instead of recognizing who a visitor is from a stored profile, it reads what the visitor is doing in the current session and decides what to show from that behavior — an approach unaffected by cookie deprecation, tracking prevention, or consent-driven data loss.
Read the full definition →Personalization flicker is the visible flash that occurs when a webpage first renders its default content and then swaps in personalized content a moment later. Sometimes called FOOC — flash of original content — it degrades user experience and contaminates experiments, because visitors briefly see the variant they were not meant to see.
Read the full definition →A decision layer is the part of a marketing stack that chooses, per visitor and in real time, which experience a page renders. It sits between traffic and site, scoring in-session behavior and selecting among authored variants. It is distinct from the content supply chain, which produces the variants but does not choose between them.
Read the full definition →Website personalization is the practice of showing different visitors different content on the same site — headlines, offers, product recommendations, calls to action — based on what is known or observed about each visitor. Approaches range from hand-written audience rules to machine-learning models to in-session behavioral scoring, and the results are credible only when measured against a randomized holdout.
Read the full definition →A personalization engine is the software component that decides which experience each website visitor receives: it ingests signals, evaluates segments or scores behavior, selects among variants, and returns a decision the page renders. Engines differ on where the decision runs, whether they cover anonymous visitors, and whether results are proven against a holdout — differences that matter more than the shared label.
Read the full definition →Real-time personalization selects and renders a tailored website experience at the moment a visitor interacts — in milliseconds — rather than acting on precomputed batch segments. The defining property is decision latency: signals are evaluated live and the page adapts while the visitor is still looking at it. Batch personalization improves the next visit; real-time personalization changes this one.
Read the full definition →Same-session personalization acts on visitor signal within the visit that produced it: behavior observed moments ago shapes what the site shows next, before the visitor leaves. The term became common as data warehouse and CDP pipelines got fast enough to activate within a visit; for anonymous visitors with no profile to fetch, same-session and in-session personalization describe the same practice.
Read the full definition →