Knock AI

The AI agent carries the conversation: it engages buyers around the clock, answers questions, runs a qualification chat against the customer's rules, routes buyers and books meetings once they qualify. A substantial product works without it, since buyers can chat with human reps over Slack, LinkedIn or WhatsApp links, and identification, enrichment, intent scoring, scheduling and CRM sync all run as conventional features.

The customer decides when the agent steps in and when reps take over, conversations follow workflows the customer defines, and the agent books a demo only after the buyer matches qualification and routing rules. Reps watch every conversation live in Slack and can reply at any point, with full context passed at each handoff. LinkedIn connection requests go out automatically when intent triggers fire, and stopping a misbehaving agent or capping automated outreach is not described.

The agent is described by what it does, engaging, qualifying and routing on context and intent, with the language models and providers behind it unnamed. Qualification and routing follow rules and territories the customer writes, so that part of the logic is open to the buyer. The intent score's inputs are listed but its weighting is not, and nothing says whether buyers learn they are talking to software.

Rivery reports a 38 percent pipeline increase within three months, more than its previous chat tool produced in twelve, and a twelvefold return on investment in the same period. Other named customers cite a 30 percent rise in click to qualified lead conversion, a 75 percent cut in lead to qualified lead time and ten times more pipeline than form fills. Baselines, sample sizes and how pipeline and return were measured are not given.

Website tracking runs subject to the customer's own cookie compliance, and buyers decide what to share in a conversation. Knock sends automated LinkedIn connection requests to high intent visitors and carries conversations over WhatsApp and Telegram, yet consent for those messages, LinkedIn's messaging rules and opt out handling are not addressed. Lawful outreach is left to the customer.

A privacy notice explains the data Knock collects from buyers who engage with its customers, including conversation transcripts and device data from customer site cookies, and the professional profiles it builds from public and third party sources. It names legitimate interest as a basis, lists access, erasure, objection and portability rights, and relies on standard contractual clauses for transfers. A DPA for customers, a subprocessor list and a lawful basis for the profiling are not offered.

Buyer identification and enrichment draw on professional profiles that Knock builds from publicly available sources and data supplied by third parties, adding verified contacts, firmographics, technology stack, headcount and funding. Lead generation companies and LinkedIn appear as sources for Knock's own marketing contacts. Suppliers, licensing terms and the legal footing for the enrichment data are not named.

Knock Outreach triggers LinkedIn connection requests automatically from visitor intent, and conversations run inside LinkedIn, WhatsApp, Telegram and the customer's Slack workspace. Knock takes no position on how its LinkedIn automation fits a platform that prohibits automated activity, and it does not say whether WhatsApp runs on the official business interface. Google Workspace data follows Google's Limited Use rules, and HubSpot and Salesforce connect natively.

Conversation transcripts, buyer details and enrichment data flow through Knock and into the customer's CRM, and personal information is kept for as long as its original purpose requires. Whether those transcripts or customer data train the agent's models, and which model providers receive them, is not addressed. Google Workspace data stays within Google's Limited Use rules.

Knock's marketing features a customer praising the agent because some prospects chatted with it for 30 to 40 messages without noticing it was AI, presenting undetected AI conversation as a selling point. Buyers reach the agent through ordinary looking Slack, LinkedIn and WhatsApp conversations, and nothing describes a default notice that they are talking to software. Knock takes no position on the disclosure duties in Article 50 of the EU AI Act.

HubSpot and Salesforce connect in one click with field mapping, and enriched contacts, account data, intent signals and conversation activity sync in real time, triggering routing, alerts, tasks and workflows once intent thresholds are met. Slack serves as the rep workspace, with LinkedIn, WhatsApp and Telegram as buyer channels and custom integrations on Enterprise. Sync direction, conflict rules and a public API are not described.

Personal information may move outside its country of origin under standard contractual clauses, and hosting providers support the service. Where Knock runs, which cloud and regions hold conversation transcripts and enrichment data, and whether any residency choice exists are not stated anywhere a buyer can read.

Knock claims an AICPA SOC attestation and GDPR compliance without naming the report type, audit period or auditor, and it runs no trust center or security documentation beyond its terms of use and privacy notice. Enterprise adds advanced permissions, workspace controls, audit logs and enterprise security settings.

Pipeline Foundation starts at $2,000 a month and Pipeline Acceleration at $3,500 a month, each with its features listed, while Enterprise is quoted. Both starting prices leave room for usage based increases, and limits on identification and enrichment apply below Enterprise, where they become unlimited, without saying what those limits are or what exceeding them costs. Billing terms and contract length are not shown.

Conversations, enrichment and activity sync continuously into the customer's HubSpot or Salesforce, so most working data already sits in the customer's own CRM. Individuals can request portability of data they provided. Contract terms covering export of transcripts, deletion after termination and renewal notice are not public.

Knock avoids cold email: buyers start conversations by clicking a chat link, and follow ups run inside Slack, LinkedIn, WhatsApp and Telegram. The one automated sending surface is LinkedIn connection requests triggered by intent. Daily limits on those requests, and how the product protects the sending account, are not described.

High growth and enterprise B2B sales and marketing teams are the buyers, especially those with limited SDR capacity, and the plans are built around inbound pipeline rather than cold outreach. Named customers are mostly B2B software firms, among them Rivery, Optimove, Cyera, Port, LSports, Descope and the logistics company Veho. Languages supported by the agent and the geographic reach of its enrichment are not stated.