Four tools, one question, hours of work

Salesforge.ai is an outbound sales execution platform that brings email, LinkedIn, AI personalization and deliverability infrastructure together in one place. Around it sits the Forge Stack, a family of seven products, from Agent Frank, their AI SDR, to lead search and inbox warming. A Nekt customer since May 2026, Salesforge had one recurring problem: answering a question like who is about to churn meant pulling billing from Chargebee, crossing it with the CRM in HubSpot, checking product events in Mixpanel, and digging through support in Gleap and Jira. That work is gone. The answer is now instant.

What Salesforge built on Nekt

• A churn engine that separates billing from health on purpose, with a rollup that carries MRR-at-risk forward and an append-only history of every risk transition.
• Revenue modeled end to end: MRR waterfalls at the customer and product level, unit economics, and involuntary-churn recovery.
• Claude Sonnet reading every sales call, both the outbound bot's and the human ones, extracting risk, pains, objections and next step straight onto the paying account. The same path classifies the intent of every support ticket.
• Agents that query in plain language. Nekt's MCP gives their AI agents direct, structured access to the lakehouse tables, so the team asks a question instead of opening a dashboard.

On top of that: DNC compliance monitoring, deliverability analytics over their own outbound engine, and acquisition unified across all seven products.

The systems behind it

Twenty data sources feed the layer, unified and modeled instead of left as raw exports. Among them:

From signing to first customer table in 18 days

Salesforge signed with Nekt on May 12, 2026, wanting a BigQuery-connected setup without maintaining ingestion infrastructure, and a way for RevOps to build through Claude Code and the Nekt MCP server; engineering signed off after reviewing Nekt's SOC 2 Type II report. One RevOps builder shipped the first customer table 18 days later, on May 30, without a dedicated data team: one record per customer with MRR by product, status, activation and renewal dates, reconciled against Chargebee before anything was built on top.

Today a completed Chargebee sync triggers the customer journey model, then the MRR changelog and customer health, with multi-input models waiting for every upstream run. The main sequence runs four times a day in about 30 minutes, feeding Forge Control, Salesforge's revenue and retention dashboard, Slack alerts, and an internal question box answered over the Nekt MCP, so plain-English questions and dashboards read the same prepared data. After launch the team halved its daily credit use by moving the main sequence from eight runs a day to four and slower tables to a daily schedule, and clarified that customer health reflects billing status after usage signals did not predict churn in its analysis. The full method is the worked example in how to build a connected GTM tech stack.

Running on the intelligence they sell

Salesforge sells sales software. Now it runs its own operation on the same standard of intelligence it delivers to the market. Nekt isn't just the analytics layer here; it is the data and AI operating system of a company that uses intelligence to decide faster and ship better. It is the same idea behind context engineering for AI agents and a governed MCP server for data: the meaning travels with the data, so agents answer instead of guess. Learn more about Salesforge.