How Crescendo Grew Quarterly Pipeline By $14M  In 4 Months With The Kiln

How Crescendo Grew Quarterly Pipeline By $14M In 4 Months With The Kiln

Overview

Crescendo, an AI customer-experience platform that automates support and contact-center work for mid-market and enterprise companies, is scaling fast and wanted a go-to-market engine built to match. Rather than run one-off campaigns, The Kiln built the system underneath the growth: an ICP model that scores the entire market, contact-level scoring that finds the right buyer, and a repeatable, automated multi-channel motion. Over the partnership, quarterly pipeline climbed from $21M to $35M.

As a company building on frontier AI itself, Crescendo wanted the same from its GTM stack. Claude runs inside the live workflow, screening every sourced company before any paid enrichment, and Claude Code is how The Kiln builds, audits and tests every change to the engine.

The Challenge: Turning Strong Demand Into a Scalable Engine

Crescendo had real demand and an ambitious growth target. The opportunity: turn a large addressable market into a systematic, data-driven motion, one that tells the team exactly which accounts to pursue, reaches the right people inside them, and launches new campaigns at speed. That meant three things: a scoring model that predicts revenue, precision at the contact level, and the automation to run it all at scale.

Scoring an entire market also raises the stakes on getting it right. An AI step running on the wrong companies wastes enrichment spend, a scorer that drifts sends reps after the wrong accounts, and a CRM fault can quietly hold good leads back from sales.

The Kiln's Approach: Build the Engine, Not the Campaign

An ICP model that scores the entire market. The Kiln built a multi-dimensional 1-to-100 scoring model and ran it across 50K+ companies in Crescendo's TAM, scoring the entire database overnight, with a tier and clear reasoning on every account. Validated against a year of closed deals, the top tier converts 3x better, with 6x the pipeline per account, so the team spends its time where it pays off.

Claude as the first filter. Before any paid enrichment runs, Claude Sonnet 5 checks every sourced company's industry and business model against Crescendo's ICP. The Kiln tested models side by side and chose Claude for the steps that need to reason over rich, record-level context. In validation, the Claude screen matched 96 of 100 companies, with no disagreements on industry or competitors.

Claude Code behind every change. The Kiln uses Claude Code to audit, fix and test the scoring engine. It traced a slow score inflation to the model calculating a weighted average by hand and moved the scorer to count-based tiering, removing the drift entirely. A full audit of the scoring sub-functions found matching bugs and dead prompts, and the refactor cut their cost by about 60% per row. To prove each change was safe, The Kiln built a code version of the scorer and shadow-ran it across 13,868 accounts: it matched production on 97.4%, and every difference traced back to a rule the original prompt was breaking. After every update, open opportunities are re-scored and compared so the team knows exactly which accounts moved, and why.

Contact scoring that finds the right buyer. On top of the account score, The Kiln scores and ranks the people inside each account, so reps open with the buyers most likely to engage, not just the right company.

A repeatable, automated campaign engine. Its always-on ABM motion, Map & Attack, runs email, LinkedIn, and calls together as one coordinated play, and event-scraping automation turns each event, speakers, attendees, and engagers, into a sourced, scored campaign in days. Consolidated workflows cut the time to launch a new campaign from weeks to days.

A CRM Claude can read. The Kiln connects Claude directly to Crescendo's HubSpot and Salesforce to pull data, map fields and diagnose routing. That is how the team found a router fault that had stranded 2,654 contacts at MQL [confirm: and got them back to sales]. Scores, tiers and reasoning are written back to about 10,000 matched HubSpot companies, so reps see why each account ranks where it does.

The Results: A Compounding Pipeline Engine

• Quarterly pipeline grew from $21M to $35M, up 67%, over the partnership.

• 50K+ companies scored, with Tier 1 converting 3x better and 6x the pipeline per account.

• A high-performing LinkedIn channel: 39–44% connection-accept rates and a steady stream of high-quality replies from senior buyers on the top campaigns.

• Campaign launch cut from weeks to days, across a repeatable multi-channel motion.

• Every sourced company screened by Claude before a credit is spent on enrichment.

• About 60% lower cost per row on the scoring sub-functions after a Claude Code refactor, with score drift removed.

• 2,654 stranded contacts found through Claude's connection to the CRM.

Conclusion: The System Behind the Growth

Crescendo no longer runs campaigns one at a time, it runs an engine that scores the market, finds the right buyers, and launches at speed, and it keeps compounding as the company grows. With Claude inside the workflow and Claude Code behind every change, it also keeps getting cheaper and more accurate, and Crescendo's own team now builds its own Claude tools on top of it.