Reference
The GTM & AI agent glossary
Short, plain-English definitions for the terms behind AI agents in the revenue stack, what they mean, why they matter, and where the oversight actually lives.
AI oversight
04 terms
- HOTL, human over the loopHuman on the LoopHuman on the loop is an oversight model where an AI system executes actions on its own while a person monitors it and retains the ability to intervene, veto, or roll back. The human supervises rather than authorises, so the failure is caught during or after the action rather than prevented before it.
- HITLHuman in the LoopHuman in the loop (HITL) is an oversight model where an AI system cannot execute an action until a person approves it. The human sits inside the decision cycle, so the failure is caught before it happens rather than reported after, which is why it is the strongest control for irreversible or customer-facing actions.
- HICHuman in CommandHuman in command (HIC) is the highest level of AI oversight: a person retains the authority to decide whether and when an AI system is used at all, sets the boundaries it operates within, and can shut it down entirely. Unlike approving individual actions, it is the standing control over the system's role.
- AI kill switchKill SwitchAn AI kill switch is a control that stops a running agent immediately when something goes wrong. A real one is central (one place that can pause every agent), complete (it revokes tokens and halts queued actions, not just new ones), and tested, so it works in the seconds you need it rather than the minutes you do not.
AI fundamentals
01 terms
AI security
03 terms
- Excessive AgencyExcessive Agency is an AI security risk, listed as LLM06 in the OWASP Top 10 for LLM applications, where an agent can take high-impact actions it should not. OWASP names three root causes: too much functionality, too many permissions, and too much autonomy. Notably, the model's intelligence is not on that list.
- Prompt InjectionPrompt injection is an attack where malicious instructions are hidden in the content an AI agent reads, such as an email, a web page, or a document, causing the agent to follow the attacker's commands instead of the user's. It is the top-listed risk in the OWASP LLM Top 10.
- NHINon-Human IdentityA non-human identity is any account that acts in your systems without being a person: a service account, an API key, an automation, or an AI agent. These identities now vastly outnumber human ones, and most organizations govern them far less carefully, which becomes a problem when agents start writing to systems of record.
GTM
33 terms
- AI sales development repAI SDRAn AI SDR is software that automates the sales development rep's work: researching prospects, writing outreach, running sequences, and booking meetings. They range from fully autonomous agents that run the whole outbound motion to augmentation tools that speed up a human rep who still owns what gets sent.
- Conversation IntelligenceConversation intelligence is software that records, transcribes, and analyzes sales and customer calls to surface insights: what was said, what was committed, deal risk, and coaching moments. Tools like Gong and Chorus popularized the category, and it now feeds AI agents that draft the follow-up work after a call.
- Revenue OperationsRevOpsRevOps, or Revenue Operations, is a strategic function that unifies and optimizes all processes, systems, and data across marketing, sales, and customer success teams. Its primary goal is to maximize revenue generation, improve operational efficiency, and enhance the overall customer experience by ensuring seamless alignment from lead acquisition through retention.
- Speed to LeadSpeed to Lead refers to the elapsed time between when a potential customer, or lead, expresses interest in a product or service and when a sales representative makes initial contact. Minimizing this time is crucial in B2B sales, as a quicker response significantly increases the likelihood of engaging the lead and moving them through the sales funnel. It is a key metric for sales efficiency and effectiveness.
- Intent DataIntent Data refers to information collected from various digital sources that indicates a company or individual is actively researching or showing interest in specific products, services, or topics. In B2B sales and marketing, it helps identify potential buyers who are likely to be in-market for a solution, allowing go-to-market teams to prioritize their efforts and personalize outreach effectively.
- Lead RoutingLead routing is the automated or manual process of assigning incoming prospective customers (leads) to the most appropriate sales representative or team within a B2B organization. This assignment is typically based on predefined criteria, such as geographic location, company size, industry, product interest, or lead source. Its primary goal is to ensure timely follow-up and maximize the likelihood of converting a lead into a customer.
- ICPIdeal Customer ProfileAn Ideal Customer Profile (ICP) is a detailed description of the type of company that would gain the most value from a vendor's product or service, and in turn, be most valuable to the vendor. It outlines key firmographic, technographic, and sometimes behavioral characteristics of a perfect-fit customer.
- NRRNet Revenue RetentionNet Revenue Retention (NRR) is a key B2B metric that measures the percentage of recurring revenue retained from existing customers over a specific period. It accounts for revenue from renewals, upgrades, and cross-sells, offset by revenue lost from downgrades and churn.
- Email DeliverabilityEmail Deliverability refers to the ability of a sender's emails to successfully reach the recipient's primary inbox, rather than being blocked, filtered into spam, or bouncing. In B2B go-to-market and sales, it directly impacts the effectiveness of outreach campaigns, ensuring that critical messages like proposals, product updates, and follow-ups are seen by target prospects and customers. Maintaining high deliverability is fundamental for efficient communication and revenue generation.
- Sales CadenceA sales cadence is a pre-planned, timed sequence of sales activities designed to engage a prospect over a specific period. It typically includes a mix of communication channels like emails, phone calls, social media interactions, and other touchpoints. The goal is to consistently nurture the prospect, build interest, and ultimately move them towards a sales conversation or a specific action.
- Data EnrichmentData Enrichment in B2B go-to-market and sales refers to the process of adding more detailed, accurate, and relevant information to existing customer and prospect records. This involves appending data points like company size, industry, technology stack, revenue, and contact details from various external sources. The goal is to create a more comprehensive profile, enabling better targeting, personalization, and more effective sales and marketing strategies.
- Lead ScoringLead Scoring is a methodology used in B2B sales and marketing to rank potential customers (leads) based on their perceived value and likelihood to convert into a paying customer. It assigns numerical scores to leads, indicating how interested they are in a company's product or service and how well they fit the ideal customer profile. This prioritization helps sales teams focus their efforts on the most promising opportunities.
- Revenue IntelligenceRevenue Intelligence is the process of gathering and analyzing data to predict and understand potential sales opportunities.
- SEPSales Engagement PlatformA Sales Engagement Platform is a software solution designed to help sales teams manage and automate their interactions with potential customers.
- ABMAccount-Based MarketingAccount-Based Marketing (ABM) is a targeted approach where businesses focus their marketing efforts on specific companies rather than individuals.
- CACCustomer Acquisition CostCustomer Acquisition Cost (CAC) is the total cost incurred by a company to acquire one new customer.
- ARRAnnual Recurring RevenueAnnual Recurring Revenue (ARR) is the total revenue a company expects to receive from a customer over a year for products or services they have already sold.
- SQLSales-Qualified LeadA Sales-Qualified Lead is a potential customer who has shown interest in a product or service and meets specific criteria set by the sales team.
- MQLMarketing-Qualified LeadA Marketing-Qualified Lead is a potential customer who has shown interest in a product or service through specific marketing activities and meets predefined criteria set by the marketing team.
- Sales VelocitySales velocity measures how quickly a team turns opportunities into revenue. It is commonly calculated as the number of open opportunities multiplied by the average deal value and the win rate, divided by the average length of the sales cycle. A higher figure means revenue is being generated faster.
- TAMTotal Addressable MarketThe total addressable market is the entire group of potential customers who could be reached and served by a company's products or services.
- Win RateWin Rate is the percentage of sales opportunities that result in a sale.
- Churn RateChurn rate is the percentage of customers who stop doing business with a company over a specific period.
- SDRSales Development RepresentativeA Sales Development Representative (SDR) is a sales professional responsible for identifying potential customers and establishing initial contact to drive pipeline growth.
- LTVCustomer Lifetime ValueCustomer Lifetime Value (CLV, also called LTV) is the total profit a business expects to earn from one customer over the entire relationship. In B2B SaaS it is an estimate rather than a measurement: you project it from average account revenue, gross margin, and how long customers stay, so it tells you how much you can afford to spend acquiring and keeping each account.
- MRRMonthly Recurring RevenueMonthly Recurring Revenue (MRR) is the fixed amount of money a company receives regularly from its customers for ongoing services or products.
- Pipeline CoveragePipeline coverage is the ratio of open pipeline value to the sales target for a period, often expressed as a multiple such as 3x. It tells a team whether it has enough opportunities in play to hit quota, given its typical win rate.
- Sales ForecastingSales forecasting is the process of estimating future sales based on historical data and market trends.
- Golden DatasetA Golden Dataset is a curated, high-quality collection of verified customer and market data used to inform targeted sales strategies in B2B contexts. It serves as a centralized repository of accurate leads, company insights, and engagement opportunities to improve outreach effectiveness and sales efficiency.
- Mutual Action PlanA Mutual Action Plan is a structured agreement between a business and its client outlining shared steps to achieve specific goals. It ensures both parties align on objectives, responsibilities, and timelines to drive successful outcomes.
- Waterfall EnrichmentWaterfall enrichment is the practice of querying several data providers in sequence rather than relying on one. The first provider is asked for a record; if it returns nothing, or nothing usable, the request falls through to the next, and so on down the waterfall. It exists because no single vendor has complete coverage, so chaining them raises match rates while paying only for the lookups that succeed.
- Sales Accepted LeadA Sales Accepted Lead is a qualified prospect that the sales team formally takes ownership of after initial validation by marketing or a lead generation team. It marks the official handoff from lead generation to the sales process, ensuring the opportunity is prioritized and tracked within the sales pipeline.
- RevOps and Sales Terms: A GlossaryRevenue operations and sales run on a dense shorthand of lead stages, deal qualification frameworks, and email deliverability standards. This glossary gathers the RevOps and sales terms that define how a lead becomes an opportunity, how teams stay aligned on service levels, and how outbound mail actually reaches the inbox.
Agents
03 terms
- AI GuardrailsAI guardrails are the rules, checks, and constraints placed around an AI system to keep its behavior safe, accurate, and within policy. They can block unsafe outputs, limit what actions an agent may take, and require human approval on high-consequence steps.
- Tool CallingTool calling, also called function calling, is when an AI model or agent invokes an external function, API, or tool to take an action or fetch information rather than only generating text. The model decides which tool to call and with what inputs, then uses the result to continue the task.
- AI Agents in GTM: A GlossaryAI agents are moving from chat interfaces into the execution layer of go-to-market work: reading a CRM, drafting outreach, and writing records back. This glossary collects the core vocabulary for that shift, from the foundational idea of agentic AI through the protocol, guardrails, and evaluation methods that keep an agent's actions safe and reviewable.
FAQ
Questions
What is an AI agent in a go-to-market context?
An AI agent is software that uses a language model to pursue a goal by taking actions in real systems, rather than only producing text. In a revenue stack that means updating a CRM record, filing a ticket, or drafting a follow-up email. The distinction that matters commercially is not how well it writes, it is that it can act, which is why the terms in this glossary concentrate on oversight rather than on model quality.
What is the difference between human in the loop and human in command?
Human in the loop means a person reviews and approves individual decisions before they take effect, so the human sits inside each consequential action. Human in command is broader: people retain overall authority over the system, including the ability to override or shut it down, without necessarily approving every step. A system can be human in command while still running many actions unattended, which is why the two terms are not interchangeable.
Which of these terms actually change how you deploy an agent?
Five carry real operational weight: excessive agency, which is the risk of granting an agent more functionality, permission, or autonomy than its job needs; least privilege, the scoping discipline that limits it; non-human identity, so an agent acts as itself rather than borrowing a person login and disappearing from the audit trail; prompt injection, because an agent can be steered through its inputs in ways an employee cannot; and kill switch, the ability to stop it. The rest of the glossary is vocabulary. These five are design decisions.
Is Mindlyft one of the tools defined in this glossary?
No. Mindlyft is GTM engineering on subscription, powered by A.S.T.R.A., the approval and audit layer over AI agents operating in revenue systems. It is not a CRM, not an outbound sequencer, and not a meeting-notes tool. It runs on top of the tools a team already uses and executes against them under human approval.