Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/vrashmanyu605-eng/Agentic_Sales-Markerting/llms.txt

Use this file to discover all available pages before exploring further.

The CRM update agent closes the loop between a sales conversation and your CRM record. After a meeting with a prospect, you supply the transcript and the client’s name. The agent searches the web for recent news about the client, asks the LLM to extract structured CRM-ready data from the transcript, and writes a row to your Google Sheet. It captures everything a sales manager needs to track an opportunity: what was discussed, what the client needs, objections raised, budget signals, urgency, next steps, and estimated deal value.

​
State inputs

​
meeting_transcript
string
required
The full text of the meeting transcript. The LLM analyzes this to extract all CRM fields.
​
client_name
string
required
The name of the client company. Used to search for recent news and written to the Google Sheet as the row identifier.
​
spreadsheet_id
string
The Google Sheets spreadsheet ID. If not provided, the agent skips the sheet write and logs a warning. The spreadsheet ID is the long alphanumeric string in the sheet’s URL.
​
spreadsheet_range
string
The cell range to write to. Defaults to "Sheet1!A1" if not provided. Use standard A1 notation (e.g., "CRM!A2") to target a specific sheet and starting cell.

​
State outputs

​
crm_update
string
required
A JSON string containing the full extracted CRM data. See output schema for all fields. This is stored in state regardless of whether the Google Sheet write succeeds.

​
What it does

1

Search for recent client news

The agent constructs the query "recent news {client_name} company" and calls web_search. Fresh news about funding rounds, leadership changes, acquisitions, or product launches can reveal new buying signals or objections not mentioned in the transcript.
2

Extract CRM data with the LLM

The agent sends client_name, meeting_transcript, and client_news to the LLM. The LLM extracts seven structured fields: meeting summary, client requirements, objections, budget signals, urgency level, next steps, and estimated opportunity value.
3

Write to Google Sheets

The agent parses the LLM’s JSON response and builds a single-row array with five columns: client_name, meeting_summary, client_requirements, opportunity_value, and next_steps. If spreadsheet_id is present in state, the agent calls update_google_sheet to append the row. If the ID is missing, the agent logs a warning and skips the write.
4

Store the result in state

The raw LLM response string is stored as crm_update regardless of whether the sheet write succeeded. This lets you inspect or reprocess the extracted data without re-running the LLM.

​
Tools used

web_search

Fetches recent news about the client company to supplement transcript analysis with current context.

google_sheets

Writes the extracted CRM row to the specified Google Sheet using the Sheets API.

​
Google Sheet columns

The agent writes exactly one row per invocation. The columns written are, in order:
ColumnSource fieldDescription
Aclient_nameThe client company name from state
Bmeeting_summaryLLM-generated summary of the meeting
Cclient_requirementsRequirements extracted from the transcript
Dopportunity_valueEstimated deal value or revenue opportunity
Enext_stepsAgreed follow-up actions from the meeting
The agent requires a credentials.json file in the project root to authenticate with the Google Sheets API. Without it, update_google_sheet returns an error string immediately and the sheet write is skipped — no exception is raised. See the Google Sheets setup guide for instructions on generating and placing this file.

​
Output schema

The LLM is prompted to return structured JSON only. The expected structure is: 
{
  "meeting_summary": "Discussed Acme Corp's current ERP pain points and explored a phased migration approach. Client expressed strong interest in the cloud-native option.",
  "client_requirements": "Cloud ERP with real-time inventory tracking, integration with existing MES, and a 6-month implementation timeline.",
  "objections": [
    "Concerned about data migration risks",
    "Board approval required for contracts over $100K"
  ],
  "budget_signals": "Finance director mentioned a $150K budget approved for Q3. Looking to start before fiscal year end.",
  "urgency_level": "High — contract renewal with current vendor expires in 4 months.",
  "next_steps": "Send detailed proposal by Friday. Schedule technical deep-dive with CTO for next week.",
  "opportunity_value": "$137,000 first year, $24,000 annual support thereafter"
}
objections and budget_signals are not written to Google Sheets by default, but they are available in the crm_update state field for use in follow-up outreach or internal review.

​
Source code

import json
from llm import llm
from tools.google_sheets_tool import update_google_sheet
from tools.search_tool import web_search

def crm_update_agent(state):
    meeting_transcript = state["meeting_transcript"]
    client_name = state["client_name"]
    spreadsheet_id = state.get("spreadsheet_id")
    spreadsheet_range = state.get("spreadsheet_range", "Sheet1!A1")

    # Search for any recent news about the client company
    search_query = f"recent news {client_name} company"
    client_news = web_search(search_query)

    response = llm.invoke(
        f"""
        You are a CRM Update Agent.

        Analyze the meeting transcript and fresh client news to generate CRM-ready updates.

        Client Name:
        {client_name}

        Meeting Transcript:
        {meeting_transcript}

        Fresh Client News:
        {json.dumps(client_news, indent=2)}

        Extract:
        - meeting_summary
        - client_requirements
        - objections
        - budget_signals
        - urgency_level
        - next_steps
        - opportunity_value

        Return structured JSON only.
        """
    )

    try:
        crm_data = json.loads(response.content)
        # Prepare values for Google Sheet: [Name, Summary, Requirements, Value, Next Steps]
        sheet_values = [[
            client_name,
            crm_data.get("meeting_summary", ""),
            crm_data.get("client_requirements", ""),
            crm_data.get("opportunity_value", ""),
            crm_data.get("next_steps", "")
        ]]
        
        if spreadsheet_id:
            sheet_status = update_google_sheet(spreadsheet_id, spreadsheet_range, sheet_values)
            print(f"[CRM AGENT] Google Sheet Status: {sheet_status}")
        else:
            print("[CRM AGENT] Spreadsheet ID missing, skipping Google Sheet update.")
            
    except Exception as e:
        print(f"[CRM AGENT] Error parsing CRM data or updating sheet: {e}")

    return {
        "crm_update": response.content
    }

Build docs developers (and LLMs) love

Get started for freeTalk to us