Attention

What is Attention?

Attention is an AI-powered conversation intelligence platform designed for sales, customer success, and revenue management teams. It records meetings and calls, converts conversations into text and structured insights, automatically updates CRM systems, generates follow-up content, and triggers business processes.

The platform currently focuses on AI sales agents that are capable of learning from successful sales conversations. It is not just a tool for creating meeting minutes; it also analyzes transactions, customers, competitors, and the performance of sales staff across multiple calls.

Core functions

• Automatic recording and transcription: Captures sales conversations from online meetings, phone calls, and external recording systems.
• AI Summary: Identifies customer needs, objections, decision-makers, next steps, and action items.
• Automated CRM updates: Write call or transaction-related fields into systems such as Salesforce and HubSpot.
• Follow-up email: Generate a personalized email after the meeting based on the customer’s original words.
• AI sales agent: Identifies red flags, missed tasks, and process gaps, and triggers notifications or updates.
• Real-time guidance: Displays scripts, battle cards, and prompt questions during the call.
• AI scoring card: Automatically evaluates the performance of each call and sale in accordance with the team’s methodology.
• Cross-call analysis: Aggregates multiple interactions within the same transaction to generate insights at the transaction level.
• Natural language queries: Ask the Super Agent for information on calls, transactions, customers, and team data.
• Workflow Builder: Configure triggers, conditions, AI steps, and actions for external applications.

What can AI sales agents do?

Attention’s AI sales agents are used to turn insights from calls into actionable steps, rather than merely generating a summary. Teams can configure different agents for things such as transaction risks, CRM fields, sales methods, and tasks to be carried out after a call.

• At the end of the conversation, identify the clear next steps, the person responsible, and the deadline.
• Identify risk signals such as budget issues, decision-making processes, competitors, or security audits.
• Check whether the sales staff have completed the key problem-finding and methodology steps.
• Synchronize structured results to opportunities, contacts, or other CRM objects.
• Generate subsequent emails, internal briefings, and client handover information.
• Remind the relevant personnel to handle it through Slack, email, or CRM task alerts.
• The decision regarding transactions is updated across multiple calls, rather than relying on a single meeting.

Call records and transcriptions

Attention can record sales interactions and convert audio or video conversations into searchable text. Teams can also import records from external recording sources such as Zoom Phone and Gong, and then use Attention’s analysis tools and workflows.

Automatic extraction of CRM fields

Administrators can define free-text or option-based fields, and provide extraction guidelines for each field. Fields can be used to analyze only the current call, or to aggregate all calls related to the same opportunity.

1. Create field configurations in the settings and select the appropriate sales team.
2. Add the field name, output type, and range at the call level or transaction level.
3. Specify the extraction rules, the output format, and the approach to handle cases where insufficient information is available.
4. Use historical conversation test prompts to check whether the results match the CRM options.
5. In the workflow, select \"End call\" or \"User manual action\" as the trigger.
6. Connect to the target CRM object and map the Attention field to the corresponding field.
7. Start with manual confirmation mode, and gradually introduce automatic writing.

Comparison of call-level and transaction-level fields

AI scoring cards and sales coaching

Attention allows for the automatic evaluation of all calls or calls of a specific type, using custom scoring cards. Managers can define questions and scoring criteria based on methods such as Sandler, MEDDIC, MEDDPIC, or their own proprietary approaches.

• Check whether the salesperson has completed the introduction, discovery, demonstration, handling of objections, and identification of next steps.
• View performance trends by team, employee, call type, and time.
• It is necessary to identify the skills that require priority coaching, rather than randomly selecting a small number of recordings for review.
• Compare the behavioral differences between successful and failed conversations.
• Use key excerpts, comments, and ratings for one-on-one tutoring.
• Use consistent standards to accelerate the implementation of new sales processes and methodologies.

How to design an effective scoring sheet

1. First, determine the type of call corresponding to the scoring card, such as first discovery, demonstration, or renewal.
2. Only retain the questions that can be determined from the dialogue evidence, avoiding the evaluation of tasks that were not heard.
3. For each question, specify the conditions for passing, failing, and being irrelevant.
4. Blind testing is conducted using historical calls with different participants and outcomes.
5. Have the sales manager review the AI score and document the reasons for any discrepancies.
6. Adjust the hints, weights, and applicable teams based on misjudgments.
7. Calibrate the scoring criteria regularly to prevent the team from speaking mechanically just for the sake of getting scores.

Real-time call assistance

Salespeople can access real-time battle cards, talking points, and prompt questions during calls. The system is useful for providing reminders about product information, competitive comparisons, and methodological steps, but it should not force salespeople to read aloud the text generated by the AI mechanically.

• Display product, price, or security information based on the customer’s questions.
• Show differentiated messaging when competitors are mentioned.
• Remind sales staff to bring up any issues that have not yet been addressed.
• Provides a framework for handling objections and approved responses.
• Help new sales staff become more familiar with the products and sales methods more quickly.
• Reduce the time spent switching between documents, CRM, and knowledge bases during calls.

Follow-up emails and action items

Attention generates an email after the call, based on the content of that call, and tries to use the customer’s actual needs and wording from the conversation. Sales staff can use templates, review the content before sending it, thereby avoiding missing any commitments or sending emails that are not consistent with what was discussed during the call.

• Summarize the business objectives and issues discussed by both parties.
• List the specific next steps, responsible persons, and estimated timeline.
• Cite the features, risks, and decision criteria that are of concern to customers.
• Select the appropriate email template based on the sales stage.
• Before sending, sales staff check the facts, tone, and attachments.

Super Agent natural language queries

Super Agent enables users to search for and analyze an organization’s conversation intelligence data using natural language. It can explain why a particular transaction has been delayed, identify which competitors appear frequently, determine the common objections raised by teams, and provide relevant evidence from calls.

• Search for calls by customer, transaction, sales representative, topic, or time range.
• Summarize the historical progress and unresolved issues of the same opportunity.
• Batch update multiple Salesforce records.
• Compare the common patterns in conversations related to winning and losing bids.
• Identify the reasons for low scores on the rating sheet and find opportunities for guidance.
• Return dialogue evidence supporting the conclusion from the original snippet.

Attention Builder workflow

Builder is used to take the completion of a call, user actions, or data changes as triggers, thereby initiating AI analysis and actions in external applications. The new version of Agent Builder organizes these processes further into reusable agents.

Appropriate use cases

• Sales representative: Automatically records meetings, updates the CRM, and generates follow-up actions after the meetings.
• Sales Manager: Arrange coaching using scorecards, snippets, and trend data.
• Sales operations: Improve CRM completeness and standardize field definitions and sales processes.
• Revenue Manager: Examine transaction risks, the implementation of methodologies, and the reasons for successes and failures.
• Customer Success Team: Records handovers, renewal risks, and customer commitments.
• Product team: Collects customer requirements, feedback on concerns, and insights from competitors.
• Marketing team: Extracts customer language and content themes from real conversations.
• Corporate training: Use excellent call examples and standardized scoring to help new employees get up to speed more quickly.

Main integration

The official website states that Attention supports over 200 types of integrations, covering CRM systems, meeting tools, telephone services, email systems, collaboration platforms, and other session intelligence systems. The specific read and write permissions, as well as the available objects and fields, depend on the connector used and the enterprise’s configuration.

MCP and API capabilities

The official documentation provides API references, and it introduces the Attention MCP Server, which enables AI assistants that support MCP to search for calls, analyze transactions, view rating cards, and identify coaching opportunities. Before deployment in a company, it is necessary to restrict the scope of data that can be accessed as well as the user permissions.

• APIs use keys or account authorization to access data within an organization.
• It can read calls, transcripts, individuals, teams, comments, and related objects.
• MCP supports searching for sales sessions and transaction information using natural language.
• Sensitive customer data should not be accessed through personal accounts or unauthorized AI clients.
• Interface calls, exports, and bulk CRM update operations should be recorded.

Price and procurement methods

The Attention website does not currently list any fixed package options or prices based on the number of seats; sales are carried out through scheduled demonstrations, customized implementations, and direct inquiries from enterprises. The costs are generally influenced by factors such as the number of seats, volume of calls, CRM integration, data migration, AI agents, implementation services, and the level of support provided.

What needs to be compared before making a purchase?

Safety and compliance

Attention states that it has passed the SOC 2 Type II audit and provides information regarding corporate security. However, this certification does not automatically address the requirements related to recording consent, labor management, healthcare, finance, or data cross-border transfers in various regions.

• Before the recording begins, inform the participants in accordance with applicable laws and obtain the necessary consent.
• Configure rules for robot participation and recording by region, customer, and meeting type.
• Limit the range of calls that sales staff, managers, operators, and administrators can view.
• Audit records are maintained for CRM writing, batch updates, exports, and deletions.
• Set the intervals for saving and deleting recording files, transcriptions, summaries, and training data.
• When dealing with medical or financial clients, it is necessary to verify contracts, data processing, and industry-specific requirements.
• Employee evaluations and coaching should be transparent, and major personnel decisions should not be made solely based on AI-generated scores.

Product advantages

• Bring call records, analytics, CRM updates, coaching, and automation together on one platform.
• It provides insights at the call level as well as at the level of transactions across calls, thereby reducing the number of isolated meeting summaries.
• Custom fields and prompts can be adapted to various sales processes and CRM structures.
• The scoring sheet covers all qualifying calls, thereby reducing the bias associated with random sampling.
• Real-time battle cards and post-meeting emails serve both during and after calls.
• Over 200 integrations are available to facilitate connection with existing revenue-generation technologies.
• Super Agent and MCP provide managers with a natural language interface for handling data.

Usage restrictions

• There is no fixed public price; the costs associated with procurement and implementation need to be confirmed with the sales team.
• Errors may occur in transcription, speaker recognition, summarization, and field extraction.
• Automatic writing to CRM may treat assumptions as facts, hence testing and review mechanisms are necessary.
• Too many real-time prompts can interfere with salespeople’s ability to listen and communicate naturally.
• AI scoring cards may favor quantifiable criteria and are unable to fully assess relationships and judgment skills.
• Call recordings are related to regulations on consent, privacy, and labor laws that vary significantly from region to region.
• Value depends on the quality of CRM data, the coverage of meetings, and the level of adoption by the team.
• Cross-call analysis may combine outdated and up-to-date information, requiring time and evidence rules.

Implementation suggestions

1. Select a sales team and a specific scenario for a pilot project, such as updating the CRM after calls.
2. Organize the existing CRM fields, sales methods, call types, and requirements for recording approvals.
3. Use historical calls to establish artificial benchmarks for transcription, fields, summaries, and ratings.
4. First, have the sales staff verify the AI results, and then enable automatic writing for some low-risk cases.
5. Mandatory manual approval is reserved for sensitive fields, predictions, and customer commitments.
6. Track time saved, field completeness rate, adoption rate, and changes in transaction results.
7. After stabilization, expand to more teams, agents, workflows, and external systems.

GitHub and open source

As of the time of verification, no official complete platform source code repository that could be cross-verified with the Attention official website and documentation was found. A large number of projects on GitHub bearing the name Attention are related to machine learning attention mechanisms and have nothing to do with this product.

Attention should be classified as a commercially hosted SaaS solution, rather than an open-source session intelligence system. The vendor provides API and MCP connectivity options, but the availability of these interfaces does not mean that the backend functions related to recording, AI agents, scoring, and CRM automation are open source.

Basic information

Recommendation score

The comprehensive recommendation score is 4.4 out of 5 points. Attention is suitable for revenue teams that wish to improve both sales and administrative efficiency, as well as the quality of CRM systems and the level of managerial support, but it is necessary to verify accuracy through actual calls and to ensure strict compliance with recording regulations.

Frequently Asked Questions

What is Attention mainly used for?

It records and analyzes sales conversations, automatically updates the CRM, generates follow-up materials, identifies risks, and provides support for sales coaching.

Does Attention automatically record meetings?

It can connect to meeting and phone tools to record conversations, and it is also able to import data from certain external recording systems.

Which CRMs are supported?

The official documentation provides primary support for Salesforce and HubSpot; custom mappings for other CRM systems require confirmation with the team.

Can CRM fields be filled in automatically?

Yes, administrators can define fields at the call level or transaction level, and choose to trigger synchronization automatically or manually.

Does Attention support real-time tutoring?

Supported: battle cards, script suggestions, and methodological guidance can be provided during calls.

What is a Super Agent?

It serves as an interface for natural language queries; it allows users to search for calls, analyze transactions, and retrieve relevant conversation evidence.

How is Attention charged?

The official website does not list any fixed packages; a quote must be requested based on the team size, features, integration requirements, and scope of implementation.

Is the recording legal?

Depending on the location of the participants and the business context, companies must carry out notifications, obtain consent, and conduct data protection assessments.

Is Attention open-source?

It is not open-source; the official party provides APIs and MCP connections, but the complete platform is a commercial SaaS solution.

Can AI scores be used directly for evaluating employees?

It is not recommended to use it on its own; the manager should make a decision by taking into account call records, business outcomes, and manual verification.