The AI GTM Glossary

A reference guide to the key terms used in AI-powered go-to-market. These definitions reflect how the concepts are implemented in production systems, not just theoretical descriptions.

AI agent

An autonomous software program that executes tasks independently without requiring a human to prompt each action. An AI agent has a defined role, access to tools it can act through, persistent memory, and the ability to make decisions within its scope. Key distinction from a chatbot: an agent acts on a schedule; a chatbot responds when prompted.

AI SDR (Sales Development Representative)

An AI agent that performs the prospecting, outreach, and qualification work of a human SDR. Runs on a heartbeat cycle, finds ICP-fit prospects across multiple databases, writes personalised outreach, manages sequences, and hands off warm replies to the Account Executive with full context.

AI-native

Built from the first commit for AI agents to operate directly. An AI-native platform exposes typed schemas agents can read and write, attributes every action to an agent, and routes signals into a memory layer — as opposed to a traditional platform that added AI features on top of a human-designed data model.

Brain (memory layer)

The structured memory system that agents load before each heartbeat cycle. Organised in three tiers: Personal Brain (per agent — episodic and procedural memory specific to one agent's work), Team Brain (shared across the team — ICP, brand voice, shared playbooks), Company Brain (workspace-wide — institutional knowledge that survives turnover). Memory blocks are typed JSON records with importance scores, not vector-embedded text.

Closed-loop GTM

A go-to-market execution model where every action produces an outcome, every outcome produces a signal, and every signal feeds back into the team's memory before the next cycle. Contrasted with open-loop GTM, where signals aren't systematically captured and fed back. Closing the loop is what enables compounding performance over time.

Coaching (as memory)

A plain-English instruction written to an agent that is stored as a typed procedural memory block and reloads on every relevant future cycle. Written by a human or a supervisor agent. Permanent unless removed. One coaching note can inform dozens of future cycles.

Delegation

The process by which a supervisor agent creates tasks for worker agents. A Sales Director might delegate: “Research the top 20 healthcare SaaS companies in the US and prepare outreach for their VP Sales.” The SDR picks up this delegated task on its next heartbeat.

Escalation

When a worker agent encounters a decision outside its defined autonomy scope, it creates a review task for its supervisor (human or director agent) instead of proceeding. Examples: a pricing exception, a multi-stakeholder deal, a compliance question. Escalation paths are defined per agent and prevent autonomous action on decisions that require human judgment.

GTM (go-to-market)

The combined motion of sales, marketing, and growth that takes a product to customers and revenue. In a B2B context, GTM typically includes outbound prospecting, inbound marketing, paid acquisition, content, sales pipeline management, and customer expansion.

Handoff

The passing of a prospect or deal from one agent to another with full context. In an AI-native system, handoffs are built into the runtime — when the SDR books a meeting, the AE picks it up automatically on its next cycle, with every prior touch, company data, and sequence history available. No human routing required.

Heartbeat cycle

The fixed recurring execution cycle of an AI agent. Each cycle: scheduler fires, Brain assembles, agent reasons, tool calls execute, memory writes, handoffs queue, agent sleeps. Cadences vary by role: 60 min for SDR and Ad Manager, 90–120 min for AE and Marketing Manager, 120–180 min for director agents.

ICP (Ideal Customer Profile)

The definition of the company and contact most likely to buy and succeed with your product. Includes firmographic criteria (industry, company size, geography, revenue), technographic signals (tools they use), and behavioural signals. In an AI GTM system, the ICP is stored in the Team Brain and referenced on every prospecting cycle.

Memory block

A typed, structured record stored in an agent's Brain. Fields include: type (coaching, learning, playbook, account_context, pattern), content, importance score (0–1), source, linked episodes, and timestamps. Retrieved by importance × recency × cycle relevance during Brain Assembly. Not vector embeddings — typed JSON records that survive model upgrades.

Outbound sequence

A multi-step, multi-touch outreach campaign. Typically 4–6 steps spaced 5–7 days apart. Each step is a different channel or angle: email, LinkedIn connection, follow-up, break-up. An AI SDR creates sequences, enrolls prospects, and manages execution automatically. Replies trigger handoffs or escalations depending on sentiment.

Supervisor agent

A director-level agent that reviews worker outputs, writes coaching, delegates tasks, and escalates to humans when needed. In a GTM team: the Sales Director supervises the SDR and AE; the Marketing Director supervises the Marketing Manager and Ad Manager. Supervisor agents run at slower cadences (every 2–3 hours) since their work is review and coordination, not high-volume execution.

Workspace isolation

The architecture pattern where each customer or client workspace runs on a dedicated database rather than a shared multi-tenant database. Relevant for agencies managing multiple clients: each client gets their own agents, memory, and data. Coaching one client's SDR doesn't affect another client's workspace. Data is structurally isolated, not just logically separated.