At least five US startups have replaced Salesforce and HubSpot with AI-coded alternatives in the past six months, threatening the growth model of enterprise software giants.
At least five US startups with 20 to 70 employees have terminated Salesforce or HubSpot contracts over the past six months, replacing them with custom applications built using Anthropic's Claude Code, Replit and Lovable — saving as much as $100,000 a year.
"We just recently canceled our Salesforce contract because we have an internal CRM that was vibecoded," Fred Turner, chief executive officer and founder of health insurer Curative, said on the "20VC with Harry Stebbings" podcast. The contract cost Curative $600,000 a year.
Greenleaf Management, a 55-person real estate firm in Atlanta, saved about $100,000 annually after replacing Salesforce with a custom app built on Replit and Claude Code, with monthly maintenance of roughly $300. The Seattle Seawolves, a professional rugby team with 70 staff, cut software spending by about $100,000 and reported a 25% revenue increase since the March season start. Hank AI, a 24-person medical software company, replaced a $40,000-a-year Salesforce contract with a Claude Code-built alternative costing about $500 annually. Atonom, a 45-person Utah startup, swapped a $40,000 Salesforce contract for a Lovable-built CRM with an estimated $1,200 yearly operating cost.
The trend threatens the long-term growth narrative of enterprise software giants that have relied on sticky multiyear contracts and high switching costs. Salesforce, ServiceNow and HubSpot have seen their stocks fall 20% to 50% since the start of 2026 on investor fears that AI coding tools will let companies build their own software rather than buy it. Salesforce Chief Executive Officer Marc Benioff has pushed back, saying the company still sees "incredible demand" and that its products have "just got better with agents."
Large enterprises test the waters
The shift is not limited to small companies. Sanofi, the French drugmaker with about 75,000 employees, is reducing its use of ServiceNow by building AI agents with Claude Code and Cursor that handle equipment failure reports. Chief Digital Officer Emmanuel Frenehard said the company aims to move 80% of the workload currently handled by ServiceNow, other software and outsourcers, targeting at least $10 million in annual savings. He acknowledged the effort has faced "considerable resistance" internally.
Curative plans to cut about 80% of its total SaaS spending this year, redirecting the budget to AI tools. The company's monthly spending on Anthropic has grown from tens of thousands of dollars to millions over the past six to seven months, Turner said. He pointed to Gwen, a bespoke AI agent that negotiates contracts with healthcare providers, as an example: completing one contract previously cost $1,500 to $2,000 on average; Gwen's average cost is about $70.
SaaS vendors defend their turf
Salesforce President Srini Tallapragada told investors last fall that enterprise customers "have tried to do it themselves, but they realize you can't vibe code enterprise-grade reliability and security." ServiceNow reported a 97% contract renewal rate in the first quarter, which it said shows customers are expanding their use of the platform rather than leaving.
But the defense has not fully reassured investors. OpenAI executives told investors earlier this year that they expect the company's future products to replace software from Salesforce, Workday, Adobe and Atlassian, according to The Information. Some large enterprises are shortening contract terms with traditional software vendors while increasing budgets for AI providers like Anthropic.
IT consulting firm Loka Chief Executive Officer Bobby Mukherjee said a small number of clients have reduced annual software costs by 40% to 80% through AI-built alternatives, though he generally advises against full replacement. "The smarter approach is usually to build on top of existing systems," he said, adding that "no serious person is predicting the death of HubSpot or Salesforce."
The real barrier to mass migration is data migration. Enterprise software's "true total cost of ownership" is typically four times the listed price, Mukherjee said, but the accumulated custom workflows — tracking product catalogs, pricing and customer commitments — make switching costly and slow. Engine, a travel management app with about 1,000 employees and a Salesforce customer for nearly a decade, estimated a migration would take at least a year and has no plans to leave.
For investors, the question is whether AI coding tools will improve enough to make the switch viable for mid-market and enterprise clients. If they do, the structural advantages that have supported SaaS multiples for more than a decade — high switching costs, expanding seat counts and annual price increases — could face their most serious challenge yet.
This article is for informational purposes only and does not constitute investment advice.