Attention
Attention is a sales conversation platform that records every meeting without a bot, reads the email around each deal and turns both into CRM updates, scorecards, forecasts and next steps. Calls are transcribed in more than 100 languages, each one is scored against MEDDIC, BANT or a custom framework, and forecasts are built from what buyers said rather than what reps typed. Recordings can also be pulled in from Gong, Aircall and other tools.
Its agents watch each deal and propose the next move: a follow up drafted in the rep's voice, outreach to other buyers on the account, a nudge for a stalled deal or a proposal built from the pricing discussed on the call, each sent after one click of approval. Teams describe their own agents in plain language, such as competitor reports, win and loss reviews or escalation alerts, and a Super Agent answers questions across calls, emails and deals. An MCP server gives Claude, ChatGPT and Cursor more than 70 read and write tools.
Salesforce, HubSpot, GoHighLevel, Snowflake, Outreach, Salesloft and Slack are among more than 200 integrations, and customers include Lovable, Preply, Avoca and Abridge. Preply reports a 70 percent rise in close rates and ten hours a week saved per rep on CRM entry.
Pricing is per user and set through a sales conversation. Attention holds SOC 2 Type II, supports HIPAA, stores data in the United States and keeps customer data out of general AI model training. The company behind it, Benwick Inc., is based in New York.
Capability Axes
Capability grades
17 of 17 axes rated · 10 graded A or B
Transcription, framework scorecards, forecasts built from buyer language, CRM autofill and the agents that draft follow ups, proposals and outreach all run on models. Remove them and only a call recorder would be left. Customers build new agents by describing a play in plain language.
Agents propose follow ups, buyer outreach, nudges and proposals, and nothing goes out until a rep approves it with one click. Every agent action is logged and traceable, and MCP access carries granular permissions. CRM fields fill automatically after each call, and how a wrong update is caught or a running agent is paused goes unaddressed.
The models behind transcription, scoring, forecasting and drafting are not named, and how deal health or call scores are calculated is not explained beyond the chosen framework. Recording runs without a bot, so participants learn about it only through the optional pre call email. Emails drafted in the rep's voice carry no marker that AI wrote them.
Preply reports a 70 percent rise in close rates and ten hours a week saved per rep, starting from a pilot with two multilingual reps, and estimates 20 to 30 minutes saved per call. Lovable, Avoca and Abridge report gains in CRM fill rates, lead conversion, ramp time and coaching completion. The measurement period and comparison baseline are not given.
An optional email goes to every invitee a few minutes before a recorded meeting with a legal note and a way to opt out, and Attention seeks consent from every participant whatever the state rule. Its consent guidance spans one party and all party US states and the rules in Canada, the UK, Germany, Ireland and the EU. Outbound outreach drafted by agents comes with no stated handling of opt outs or anti spam rules.
Attention honors access, correction and deletion requests, has a California section and follows Google and Microsoft Limited Use rules. For EU users it relies on consent, legal obligation and vital interests, without contract or legitimate interest, and no processor role, DPA, subprocessor list or transfer mechanism is given. Attention holds that no data comes from third parties even as it draws on deep Google and Microsoft account access.
Everything Attention analyzes comes from the customer's own meetings, email, calendar, files and CRM, plus recordings imported from tools such as Gong. Email bodies are processed to extract sales insight, and only the derived analytics are kept. No outside contact data is added.
Gmail, Calendar, Drive and Microsoft 365 connect through official APIs under Google and Microsoft Limited Use rules, and CRMs and dialers use their own integrations. Meetings are captured without a bot joining the call. How that bot free capture fits each meeting platform's recording rules is not addressed.
No user data, including anything from Google or Microsoft accounts, trains general AI or machine learning models, and full email content is not kept after insights are extracted. Data is held for no more than six months after an account ends. Whether outside AI providers are bound by the same training exclusion is not stated.
Buyers receive follow ups and outreach written in the rep's voice, approved by the rep, with nothing telling them AI drafted the message. Meeting participants hear nothing from a bot because none joins, and the pre call notice is optional. Attention takes no position on Article 50 of the EU AI Act.
More than 200 integrations include Salesforce, HubSpot, GoHighLevel, Snowflake, Outreach, Salesloft, Aircall, Clay and Slack, and any CRM can receive data back. The MCP server exposes more than 70 read and write tools to Claude, ChatGPT and Cursor with granular permissions. Which CRM objects and fields are written is not spelled out.
Data is stored in the United States with third party storage providers that go unnamed. No EU or other regional option is offered, and the hosting provider and tenancy model are not stated.
An independent audit confirmed SOC 2 Type II for security, confidentiality and availability, and Attention supports HIPAA with a business associate agreement. Annual penetration tests cover network and application layers, and data is encrypted with TLS 1.2 and 1.3 in transit. The auditor, audit period and a way to obtain the report are not given, and there is no trust center.
Attention shares no prices, plan names or seat economics, and every path to buy runs through a demo request. Plans exist, with the MCP connector included on every one, but none carries a price.
Data is kept for no more than six months after an account ends, and deletion or a copy can be requested by email. The API and MCP tools can read data out, while a bulk export of recordings and transcripts and the renewal and termination terms go unaddressed.
Approved follow ups, nudges and multi threading emails go out through the rep's own Gmail or Outlook, so sending reputation rests on the rep's mailbox. Volume limits, pacing and monitoring for that outreach go unaddressed.
Attention serves revenue teams, sales leaders, reps and customer success, with named customers from consumer apps to healthcare AI. Recording and transcription cover more than 100 languages, and HIPAA support opens regulated buyers. Company size focus and the geographic mix of customers are not stated.
Alternatives to Attention
Vendors in the same categories as Attention whose graded profiles look most alike, with a line on what each product does.
Similar is not the same, so check that each one does the job you need. No vendor pays for placement.
Revenue execution platform built around call intelligence, founded in 2022 as Glyphic by two former Google DeepMind research scientists and rebranded to Airspeed on 20 May 2026.
Conversation intelligence built specifically for business to business selling rather than for general meeting notes, and the distinction holds up on inspection.
Otter.ai records and transcribes meetings, then turns each conversation into a summary, action items and a searchable record.
Conversation intelligence for sales teams, covering call and meeting recording, transcription, and analysis of what was said.
Zoom Revenue Accelerator is Zoom's conversation intelligence product for sales and other customer facing teams, sold separately from Zoom Workplace.
Notta is an AI note taker that records and transcribes meetings in 58 languages.
Ordered by how closely the graded profiles match across the seventeen axes, inside the same categories. See how each axis is graded in the methodology.
Pricing
What this vendor charges, what it commits to in writing, and where the bill can move. Figures the vendor publishes itself are labeled Vendor Published. Figures labeled Estimated come from other sources and the vendor has not confirmed them.
No pricing data has been verified for this vendor. Pricing information will be published here once confirmed through vendor disclosure or third-party estimation.