Replay Webinar – Unlock AI in Pipedrive

🤖 Connecting Pipedrive to AI with MCP

Artificial intelligence can now go much further than simply generating text or analyzing information. Thanks to the MCP offered by Pipedrive, it is possible to connect your CRM directly to assistants such as Claude, ChatGPT, Gemini, or Perplexity.

The objective is simple: allow users to interact with their CRM using natural language. Instead of navigating through different records and menus in Pipedrive, users can ask their AI assistant to find information or, when permissions allow it, perform an action directly.

In other words, AI can become an intermediary between the salesperson and the CRM.

🔌 What is Pipedrive MCP?

MCP (Model Context Protocol) allows an AI tool to access certain features and data available in Pipedrive.

This approach is different from the Sales Assistant built into Pipedrive. The built-in assistant can provide information and help users work with the CRM. With an external AI connected through MCP, the possibilities go further: depending on the features available to the assistant, it can read, create, and modify data in Pipedrive.

This means the AI can potentially act as a real operational assistant for CRM management.

⚙️ How to Connect Pipedrive to Claude

In Pipedrive, the feature is available from the personal preferences area, under Pipedrive MCP. Pipedrive provides an MCP URL that can be used to establish a connection with a compatible AI assistant.

In Claude, the process can also be simplified through built-in connectors. At the time of the webinar, Pipedrive was already available as a Claude connector, alongside services such as Google Drive, Canva, Notion, and Figma.

Users simply need to select Pipedrive, choose the relevant account, and authorize the requested access.

However, this process can evolve quickly. Some AI assistants already offer a ready-to-use Pipedrive connector, while others may still require the MCP URL provided by Pipedrive.

🔐 A Major Consideration: Access Rights

This is probably one of the most important topics discussed during the webinar.

At the time of the demonstration, MCP access could be enabled at the company level and was enabled by default in the configuration shown. The speakers therefore recommend that administrators check this setting before allowing users to connect their own AI tools.

When a user establishes a connection with an application such as Claude, Pipedrive administrators receive an email notification informing them that the application has been installed.

The main limitation concerns the granularity of permissions. The speakers explain that it is not yet possible to manage access rights as precisely as they would like directly from Pipedrive.

For example, a company might want to allow salespeople to create information through AI without allowing them to modify certain existing data. This level of granular permission management was not available in the setup demonstrated during the webinar.

🛡️ Controlling Permissions Inside the AI Tool

Claude nevertheless allows users to control individual tools and actions.

For a specific action, it is possible to choose between several behaviors:

• Automatically allow the action;
• Request approval before executing it;
• Block the action.

For example, creating a new activity can require approval. Claude then explains what it intends to create and asks the user for confirmation before making the change in Pipedrive.

This approval process provides an additional layer of security, although it does not replace a proper permission policy configured directly within the CRM.

🔎 First Use Case: Searching for Information in Pipedrive

The first demonstration consists of asking Claude a very simple question: which deal has the highest value in the sales pipeline?

Claude queries Pipedrive and identifies the corresponding deal.

However, this demonstration reveals a fundamental point: prompt quality matters.

Simply asking, “What is my biggest deal?” can be ambiguous. The AI might include won or lost deals, or deals belonging to other salespeople.

A more precise request can specify the relevant pipeline, deal status, owner, or time period.

This level of precision helps generate more relevant answers and, according to the speakers, can also avoid unnecessary searches and excessive token usage.

📞 Turning a Sales Call Summary into CRM Data

The second example clearly demonstrates the potential time savings.

After a meeting, the salesperson simply tells Claude that they have just met a prospect. In the demonstration, they provide the contact’s name, company, meeting context, and explain that a follow-up should take place in two weeks.

Using this information, Claude checks the data already available in Pipedrive. The company already existed in the CRM, so the AI avoids creating a duplicate.

It can then create the contact, create the lead, and schedule the follow-up activity, after receiving the necessary approvals.

🎙️ A Particularly Useful Mobile Workflow

This becomes even more useful when combined with voice input.

A salesperson leaving a meeting could simply dictate a summary to their AI assistant:

“Just met this prospect. Add them to Pipedrive and schedule a follow-up in two weeks.”

The assistant can then transform that information into structured objects inside Pipedrive.

This can eliminate the need to navigate through multiple screens in the mobile application to create an organization, contact, lead, and activity separately.

🔗 Making Multiple Tools Work Together

MCP becomes even more interesting when the AI assistant has access to several connectors.

One example discussed during the webinar involves a calendar. If the AI can access both the user’s calendar and Pipedrive, it could check the user’s availability and then create the corresponding activity in Pipedrive.

The AI effectively becomes an interaction layer between multiple applications.

However, it remains important to clearly specify the relationships between different objects, such as contacts, organizations, deals, leads, and activities. An unclear request could cause the assistant to create information in the wrong place or fail to establish the expected relationships.

🌐 Enriching Pipedrive with Perplexity

Another demonstration is carried out using Perplexity, which is also connected to Pipedrive.

One of Perplexity’s strengths is its ability to perform web research. Starting from a company associated with a deal, Perplexity can search for additional information about the company and generate a summary.

When the connector provides access to the required action, the assistant can then add this summary as a note directly to the Pipedrive deal.

This creates interesting possibilities for preparing sales meetings. A salesperson could ask the AI to retrieve the existing CRM context while also researching recent or additional information about the company.

🧩 Capabilities Depend on Each Connector

Not every action is available through every connector.

During the Perplexity demonstration, certain modifications were not accessible through the connector being used. The assistant could retrieve information and perform some actions, but it could not necessarily modify every desired object.

The same operation was then tested with Claude, which had access to the tools required to perform the requested modification.

The webinar therefore shows that it is not enough for an AI assistant to simply be “connected to Pipedrive.” You also need to understand which specific actions the connector makes available.

Need	Example of AI Usage
CRM search	Find the main deals in a pipeline
Data creation	Add a contact, lead, or activity
Updates	Modify certain deal or organization information
Enrichment	Research a company and add a summary to Pipedrive
Sales follow-up	Automatically schedule a follow-up
Pipeline cleanup	Identify and close old deals

🧹 Automatically Cleaning Up Your Pipeline

The final use case presented is particularly useful for teams with a cluttered pipeline.

The assistant receives an instruction to identify deals that have remained in a specific stage for more than three months, mark them as lost, add a lost reason, and close any activities that are still pending.

The AI analyzes the data and performs several operations in sequence, following the required approvals.

During the demonstration, eight deals were marked as lost and five pending activities were closed.

An operation that would normally require several manual steps can therefore be performed through a single natural-language request.

⚠️ Automation Still Requires Control

This level of automation also introduces risks.

An unclear instruction could modify multiple deals even though the user did not properly define the intended scope. This is particularly important for bulk operations such as changing statuses, closing activities, or updating sales data.

For significant operations, it is therefore preferable to maintain human approval before execution and to write sufficiently precise prompts.

Pipedrive can also retain a record of certain changes. During the demonstration, the change log could be used to identify that an operation had been performed through an API.

🚀 A New Way to Use a CRM

The main takeaway from the webinar is not purely technical.

With MCP, users can gradually move from a workflow where they have to navigate through their CRM to one where they can simply explain what they want to achieve.

AI can search, summarize, create, and modify information while potentially combining data from several different tools.

The possibilities are therefore significant for sales teams, but the speakers emphasize that the technology demonstrated is still evolving. Access management, prompt quality, and control over AI actions remain essential.