US Small Businesses Spark Salesforce "Unsubscription Wave," Slashing Costs by Building AI-Powered In-House Software

Several small businesses in the US are abandoning traditional enterprise software like Salesforce by leveraging AI coding tools, saving tens of thousands to hundreds of thousands of dollars annually, which has further intensified investor concerns about the long-term growth logic of the SaaS industry.

Recently, according to a report by The Information, in the past six months, at least five startups and small companies with 20 to 70 employees have terminated their contracts with Salesforce or HubSpot, turning instead to AI tools such as Anthropic, Lovable, and Replit to develop replacement applications in-house, achieving significant cost reductions. Although this trend is currently limited to small customer segments, it has raised questions in the market about whether traditional enterprise software vendors can maintain the high growth momentum seen over the past decade.

Enterprise software giants like Salesforce and ServiceNow are adopting a defensive posture, emphasizing on one hand that customer renewal rates remain robust, and on the other hand reinforcing their positioning as an "indispensable middle layer" between enterprises and large AI models.

However, AI-native companies represented by Anthropic and Palantir are growing significantly faster than these traditional software vendors, fueling ongoing discussions about a "SaaSpocalypse."

Small Business "Exodus" Cases: Million-Dollar Costs Compressed to Thousands

Several specific cases reveal the scale and speed of this trend.

Greenleaf Management, an Atlanta-based real estate investment management firm with about 55 employees, previously not only paid subscription fees to Salesforce but also hired a full-time employee and an external consultant specifically to maintain its Salesforce account.

Dave Codrea, a partner at the firm, stated that the company used Replit and Claude Code to develop a custom application to replace Salesforce's CRM functions, while also terminating contracts with real estate software companies Entrata and Yardi, saving approximately $100,000 annually, with the new application's monthly maintenance cost being only about $300.

Adrian Balfour, owner of the Seattle professional rugby team Seawolves, said that four engineers within the team of 70 developed an application using Claude Code in four months to replace the Salesforce CRM and the AXS ticketing system, reducing software expenditures by about $100,000, with revenue increasing by approximately 25% year-over-year since the start of the season in March this year.

Hank AI, a South Carolina-based medical software company (with 24 full-time employees), did not renew its $40,000 annual contract with Salesforce this January, opting instead to build its own application using Claude Code, with an estimated annual fee of about $500. Utah-based startup Atonom (45 employees) similarly abandoned its $40,000 annual Salesforce contract in January, switching to Lovable to build a custom CRM, with projected annual operating costs of only $1,200.

Large Enterprises Are Also Testing the Waters, But Face Greater Resistance

This trend is not entirely confined to small businesses; some larger companies have begun exploring similar paths but face more complex challenges.

Sanofi, the French biopharmaceutical giant with approximately 75,000 employees, is cutting back on its use of ServiceNow, according to Emmanuel Frenehard, its Chief Digital Officer. The company is building AI agents via Claude Code and Cursor, combining software from Utah-based startup Elementum and data stored in Snowflake to handle tasks such as equipment failure reports.

Sanofi aims to offload 80% of the workload currently handled by ServiceNow, other software, and outsourcing firms, saving at least $10 million annually. Frenehard admitted that this transformation has "encountered considerable resistance" within his team.

Bobby Mukherjee, CEO of IT consulting firm Loka, stated that a few clients have reduced annual software costs by 40% to 80% through self-built AI alternatives, but he usually advises clients to retain existing SaaS applications because the replacement process is slow and "diverts engineering resources from things that truly create business differentiation."

He said, "The smarter approach is usually to build on top of existing systems," emphasizing that "no serious person is predicting the death of HubSpot or Salesforce."

SaaS Giants Push Back: Enterprise-Grade Reliability Cannot Be Replaced by "Vibe Coding"

In response to external skepticism, traditional software vendors like Salesforce are actively defending their business models, supported by renewal data.

Srini Tallapragada, President of Salesforce, stated in an investor conference call last autumn that corporate customers "have all tried doing it themselves, but they realized that you cannot achieve enterprise-grade reliability and security through vibe coding." ServiceNow stated that most of its customers are "expanding their use of ServiceNow rather than leaving," citing a 97% contract renewal rate in the first quarter of this year as evidence, a figure that has remained stable in recent quarters.

However, market confidence in this narrative is not solid. The Information previously reported that OpenAI executives stated in an investor presentation earlier this year that they expect future products to replace software from companies like Salesforce, Workday, Adobe, and Atlassian.

Meanwhile, some large enterprises are shortening contract terms with traditional enterprise software vendors or demanding more favorable terms while increasing their budgets for AI suppliers like Anthropic.

True Replacement Cost: Data Migration Remains the Biggest Barrier

Although the replacement cases among small businesses are notable, analysts point out that the technical and organizational obstacles to large-scale migration should not be underestimated.

Bobby Mukherjee pointed out that the "true total cost of ownership" for enterprise software is typically four times the list price, but "completely replacing it is still a last resort." The core stickiness of platforms like Salesforce lies in the customized code (i.e., "workflows") that enterprises have continuously layered on over the years. These workflows track everything from product catalogs to pricing to customer commitments. Many large enterprises even employ dedicated teams to maintain these workflows, further driving up migration costs.

Demetri Salvaggio, an executive at Engine, a corporate travel management application provider, said his company has about 1,000 employees and has been a Salesforce customer for nearly ten years. He estimates that migration would take at least a year and there are no current plans to leave.

For investors, the core question now is whether the improving capabilities of AI coding tools will spread this trend from small businesses to medium and large customer segments. If AI continues to advance in programming and handling large databases, the influence of SaaS vendors on large customers also faces the risk of erosion.