Claude Cowork in Action: Completing Sales Automation for 4,000 Customer Account Scoring Overnight

Core Scenario: How AI is Reshaping Sales Management Workflows

Travis Bryant, Anthropic's US Mid-Market GTM leader, recently shared his hands-on experience using Claude Cowork. A GTM (Go-To-Market) leader is a core role in companies responsible for product or service market entry strategy, coordinating functions like sales, marketing, and customer success. Mid-Market typically refers to enterprise customers with annual revenue between $10 million and $1 billion — one of the highest-growth segments for B2B SaaS companies. These customers are large enough to generate substantial contract value, yet don't require the extremely long sales cycles and complex procurement processes of large enterprises.

As a sales leader managing 4,000 customer accounts, Travis's core challenge was: assembling data daily across multiple systems like Salesforce, BigQuery, and Google Calendar, and updating forecast numbers — repetitive operations that used to consume most of his week. The data silo problem across these three systems is the primary bottleneck for sales operations efficiency — sales reps manually copy-paste data between multiple interfaces every day. This "data hauling" work is both time-consuming and error-prone, making it the most direct entry point for AI automation.

Claude Cowork changed this dynamic — it took over the data assembly and formatting work, allowing Travis to devote more energy to customer communication and strategic decision-making.

Sales Automation Across Three Time Horizons

Travis's daily work spans three cadences: daily customer call preparation, weekly forecast summaries, and quarterly full account scoring. Before introducing Claude Cowork, doing all three well simultaneously was virtually impossible.

Daily: Automated Meeting Prep, Saving 90 Minutes

Every morning, a scheduled task automatically scans Google Calendar and books rooms for external meetings missing a room assignment. A second skill runs customer preparation before each meeting — pulling consumption data from BigQuery, pipeline status from Salesforce, and generating a briefing ready for Travis to review the moment he opens his computer.

Travis specifically emphasized a key insight: the scheduler matters more than the skill itself. When preparation no longer depends on human memory but triggers automatically, forgetting is no longer an issue.

Weekly: Forecast Reports Reduced from 3 Hours to Minutes

Sales Forecasting is one of the most critical and time-consuming tasks in sales management. Weekly forecast meetings require sales leaders to report expected close amounts for the current quarter, pipeline health, and at-risk deals to senior leadership — a report that needs to integrate real-time information from multiple data sources. Traditionally, RevOps (Revenue Operations) teams are specifically responsible for connecting data and processes across sales, marketing, and customer success teams, typically requiring dedicated staff to complete cross-system data integration.

Every Friday, a scheduled skill automatically completes the following workflow:

• Pulls opportunity records and submitted commitments from Salesforce's Forecast tab
• Retrieves token consumption data from BigQuery
• Combines internal document notes
• Assembles everything into a single-page web report

The report format fully aligns with Anthropic's sales leadership reading preferences: topline metrics, key deals, changes, and forecast snapshots summarized by each frontline manager. The report is automatically deployed to an internal shared link before Monday's forecast meeting.

Travis's Monday work becomes adding commentary to the report — in his own words: "Claude builds the 'what,' I provide the 'why.'"

Quarterly: Completing 4,000 Account Scores Overnight

This is the largest-scale project Travis runs through Claude Cowork. Every fiscal year, each account in the book needs a score to help Account Executives (AEs) determine priorities. In previous companies and roles, this type of work required hundreds of hours of cross-departmental collaboration between RevOps, FP&A (Financial Planning & Analysis), and marketing teams.

Travis completed all scoring in one night using Claude Cowork.

Dual-Dimension Account Scoring Methodology

Account Scoring is essentially a prioritization mechanism that helps sales teams focus on the most valuable customers within limited time. Traditional Propensity Models are typically trained by data science teams using machine learning models based on historical close data, requiring large amounts of labeled data and specialized technical resources. Travis's approach represents a new paradigm: replacing traditional statistical models with the reasoning capabilities of large language models, defining scoring dimensions and weights, and combining publicly available web information with internal historical data to perform both qualitative and quantitative assessment of each account. The advantages of this approach include no training data required, strong explainability (written rationale for each dimension), and the ability to quickly iterate and adjust scoring criteria.

Travis co-defined two sets of five-dimension scoring criteria with Claude, targeting technology industry customers and traditional industry customers respectively:

Technology Account Scoring Dimensions:

• Agent opportunity potential
• Internal transformation potential
• AI commitment level
• Growth headroom relative to existing consumption
• Industry fit

Industry Account Scoring Dimensions:

• Knowledge worker density (e.g., high for law firms, low for manufacturing)
• Public AI commitment (measured by number of AI-related job postings on career pages)
• Other industry-specific dimensions

Once criteria were set, Claude Cowork conducted deep web research on each of the 4,000 accounts, combined with historical data from Salesforce and BigQuery, generating numerical scores and written rationale for each dimension. The entire process completed automatically overnight.

From Data to Tools: Interactive Sales Dashboard

After scoring was complete, Travis had Claude Cowork build an interactive dashboard based on the results, designed for direct use by frontline sales teams:

• Each AE can click on their territory's pie chart slice to view accounts ranked by score
• Each account displays scores and specific rationale for each dimension
• Hovering over an account reveals potential use cases and comparable case studies, directly usable for sales outreach

This transformed scoring from a static data exercise into a practical tool that frontline salespeople can use every day.

Why Claude Cowork Can Successfully Land in Sales Teams

Claude Code is Anthropic's command-line tool for developers — users interact with AI through the Terminal, directly manipulating file systems, executing code, and calling APIs, suited for technical users with programming backgrounds. Travis candidly shared that he tried Claude Code but never felt comfortable with terminal operations. Claude Cowork represents a different product positioning — wrapping the same underlying AI capabilities in a graphical interface and natural language interaction layer, targeting non-technical users in business functions like sales, marketing, and operations. This "democratization of technical capabilities"