AI prospecting glossary

This glossary defines the 16 essential terms of AI prospecting: from agent to swipe, from reply rate to MCP.

Each definition fits in one self-contained paragraph, with a concrete example when useful.

AI prospecting agent

An AI prospecting agent is an autonomous system that finds qualified prospects, contacts them, handles conversations and drafts every reply, under human validation. Unlike a tool that executes a fixed sequence, the agent decides, adapts and keeps the conversation going until the meeting. The human keeps final control over every message.

Example: Formula. contacts up to 28 prospects per day on LinkedIn and email, and you validate with a swipe. The full explainer is here.

SDR (Sales Development Representative)

The SDR is the salesperson in charge of outbound prospecting and prospect qualification, upstream of closers. Their job: open conversations, filter the curious from the buyers, and fill the sales team's calendar. According to Salesforce, salespeople spend less than 30% of their time actually selling.

Example: in a small company without an SDR, the founder plays the role, often at night after client work.

AI SDR

An AI SDR is software that automates SDR tasks: writing messages, sending, follow-ups. It is not an agent: it runs a predefined sequence without deciding or adapting to the conversation. When a prospect answers something unexpected, the AI SDR stalls, and the human has to take over.

Example: the difference between the two is detailed in AI SDR vs AI agent.

Outbound prospecting

Outbound prospecting means directly contacting prospects who do not know you yet: LinkedIn invitation, message, email. It is the opposite of inbound, where the prospect comes to you through your content. Outbound is the fastest lever to test an offer, because you choose who you contact and when.

Example: the Formula. method relies on calibrated outbound, within LinkedIn limits.

Social selling

Social selling is the use of social networks to create sales conversations: content, comments, targeted interactions. LinkedIn is its main playing field, with more than 1 billion members. Social selling warms up the prospects that outbound prospecting then contacts.

Example: the criteria for choosing a tool are in the AI social selling tool checklist.

Intent signals

Intent signals are observable behaviors suggesting a prospect has an active need: job change, ongoing hiring, fundraising, interaction with a piece of content. They let you prioritize who to contact and when, instead of treating every prospect the same way.

Example: the signals to watch in B2B are detailed in this dedicated article.

ICP (Ideal Customer Profile)

The ICP is the portrait of the type of company that gets the most value from your offer: size, sector, revenue, problem experienced. It is the filter that decides which prospects enter your campaigns. A fuzzy ICP produces campaigns that cast wide and convert little.

Example: Formula.'s ICP: consultants and B2B SMBs, from about 5k euros of monthly revenue, with an offer already validated.

Persona

The persona is the representation of the person to contact within the ICP: their role, priorities, objections, vocabulary. The ICP says which company to target, the persona says who to talk to inside it and how. An effective campaign pairs a precise persona with a calibrated message.

Example: the best campaign on the Formula. account targets a single persona, French-speaking SMB executives, with a 92% reply rate. Check that your offer is ready before defining your personas.

Acceptance rate

The acceptance rate is the percentage of LinkedIn invitations accepted by contacted prospects. It is the first indicator of targeting quality: a low rate signals a fuzzy persona or a profile that lacks credibility. It can be measured before the first message is even sent.

Example: on the Formula. account, 179 invitations accepted out of 587 sent, a 30% rate, data pulled July 23, 2026. Safe daily volume is explained in this article.

Reply rate

The reply rate is the percentage of first messages that receive an answer. Classic manual outreach sits between 7 and 11%. A well-calibrated agent observes 30 to 60% depending on sector and offer, because every message is personalized and sent at the right moment.

Example: 40% on the Formula. account as of July 23, 2026. What counts as a good rate is detailed in this benchmark article.

Prospecting sequence

A prospecting sequence is a scheduled chain of messages sent at fixed intervals: day 0 invitation, day 3 message, day 7 follow-up. It is the mechanic behind tools like Waalaxy or Lemlist. The sequence stops when the prospect replies: the conversation itself is still yours to run.

Example: the sequences vs agent comparison is in Waalaxy, Lemlist and 2026 alternatives.

Cold email

A cold email is the first email sent to a prospect who does not know you. Effective when short, personalized and well-timed, it complements LinkedIn by reaching the same prospect on a second channel, within a single coherent conversation.

Example: Formula. sends the email in parallel with LinkedIn to the same prospects. The pairing is analyzed in LinkedIn vs email in 2026.

Agent calibration

Calibration is the phase where an agent ingests your website, your LinkedIn profile, your conversations, your posts and your competitors to learn your offer and your voice before writing a single message. It is what separates a generic message from one that sounds like you.

Example: with Formula., calibration takes a few dozen seconds. Details are in the documentation.

MCP (Model Context Protocol)

MCP is an open protocol that connects an LLM like Claude, ChatGPT or Gemini to external tools. In practice, it lets you steer an application in plain language from the chat: ask questions, trigger actions, read data, without leaving the conversation.

Example: Formula.'s MCP server lets you ask "how many replies this week?" from Claude. The full guide is here.

Swipe (validation)

The swipe is the gesture that validates a reply drafted by an agent: one thumb movement on your phone, 3 seconds. The agent writes, you approve. It is the mechanism that keeps the human in the loop without asking them to write.

Example: with Formula., the swipe lets you answer a prospect in under 5 minutes. Why that changes everything.

GEO (Generative Engine Optimization)

GEO is the discipline of optimizing content to be cited by generative engines: ChatGPT, Perplexity, Google AI Overviews. Where SEO targets ranking in blue links, GEO targets being quoted inside the generated answer. Clear definitions, dated data and self-contained paragraphs are its pillars.

Example: this glossary is structured for GEO, with every definition quotable as is.

Key takeaways

• An agent decides and adapts; an AI SDR or a sequence executes a fixed plan.
• ICP and persona decide everything: which company to target, who to talk to, with which words.
• Two KPIs tell the story of your prospecting: acceptance rate (targeting) and reply rate (calibration).
• Reference: 30 to 60% reply rates with a calibrated agent, versus 7 to 11% with classic manual outreach.
• MCP steers the agent from Claude, ChatGPT or Gemini; the swipe stays your final validation.