Anthropic's sales team increased their closing rate by 2.5 times using Claude's automated replies.



Carl Johnson, Sales Development Leader at Anthropic, explains how the AI, Claude, now handles most inquiries and how this has changed the way the sales team works.

How Anthropic's sales team rebuilt inbound with Claude Managed Agents | Claude by Anthropic

https://claude.com/blog/how-anthropics-sales-team-rebuilt-inbound-with-claude-managed-agents



Anthropic receives inquiries from companies interested in implementing Claude, but according to Johnson, they receive tens of thousands of inquiries every month, and there was a problem where it sometimes took several days to receive a reply after submitting a question through the form. Most of the questions were covered in the official documentation and support articles, but in many cases customers wanted a quick explanation from a representative.

Therefore, Anthropic's sales team used ' Claude Managed Agents ,' a tool for building and deploying cloud-hosted agents at scale, to create purchasing agents that support prospects from the initial stage of 'I want to implement Claude in my company' to the completion of the purchase.

'Claude Managed Agents,' a tool that allows for easy deployment of Claude-based AI agents into production environments, has been released - GIGAZINE



The following five points were cited as reasons for using Claude Managed Agents:

• Rapid implementation in actual operation
The initial purchasing agent was built by a single engineer in just a few weeks.

• Technical and non-technical contributors
While the engineering department owned the code, sales and content leaders could directly review and edit system prompts on the console, and changes were first reflected in a test environment, allowing anyone to try out fine-tuning before it was rolled out to customers.

Focus on the domain, not the harness.
Claude Managed Agents handles agent loops, sessions, and hosting. The sales team focused their time on three sales-specific elements: prompts, tools, and a knowledge base.

- Version control reduces the cost of iterative development.
All changes made to agents are saved as their own versions, and they can easily revert to previous versions if needed. In the case of purchasing agents on the Anthropic sales team, they reached version 7 about a week after internal testing began, and continued to update it afterward.

・Flexibility to accommodate a wider range of use cases
While the basic model involves users visiting a site, finding an agent, and using their services, Claude Managed Agents allows for the automated and periodic execution of agents. This enables companies to explore alternative usage scenarios, such as delivering agents to users at specific times.



The purchasing agents implemented by the sales team using Claude Managed Agents handle thousands of conversations a day, answering customers' most pressing questions and completing the purchase process. Because customers handed over to the team by these agents are well-informed and ready to make purchasing decisions, they are more than twice as likely to develop into deals compared to leads from traditional forms, and the time to closing is shortened by approximately five days.



Please note that the purchasing agent service is opt-in; customers making an inquiry can choose whether or not to speak with an agent or sales representative first.

Inquiries can be made regarding Enterprise plans for large companies with extensive business operations. You can ask questions via AI chat by selecting 'Chat with a purchasing specialist' from the pricing plan page, or by selecting 'Chat with a purchasing agent' from the inquiry page .



Johnson stated, 'Purchasing agents are designed to perform the same tasks as sales representatives. This means they aim to understand the problems customers are trying to solve, their current usage of Claude, and how we can best support them.'

Furthermore, lessons learned from building a purchasing agent include the fact that it is more effective to explain goals to Claude rather than rules, that the 'less is more' principle was the most effective after trying various prompts, and the importance of leveraging experts in the development loop, such as engineers and sales teams spending more time together.

in AI, Posted by log1e_dh