The Information reported that five startups and small companies with roughly 20 to 70 employees ended Salesforce and HubSpot contracts in the past six months and replaced parts of those workflows with AI-built apps from Anthropic, Lovable and Replit, according to PYMNTS. The reported savings were 40% to 80%, while Gartner separately estimates that $234 billion in enterprise application spending could be exposed to agentic arbitrage by 2030. For AI and data teams, the signal is selective build-vs-buy pressure: narrow internal workflows can become cheaper to rebuild, but maintenance, security review, integrations, and data migration still decide whether those savings survive beyond the prototype.
The durable signal is not that SaaS is disappearing. It is that AI-assisted development is making buyers reprice the narrow workflows where seat licenses, vendor configuration, and unused platform features have become visibly expensive. For practitioners, the question shifts from build versus buy in the abstract to which workflows are simple enough, owned enough, and stable enough to justify an internal replacement.
What happened
The Information reported that five startups and small companies stopped using Salesforce and HubSpot contracts over the past six months and used AI tools from Anthropic, Lovable and Replit to build replacement applications. PYMNTS summarized the report and said the firms cut software costs by 40% to 80%. EMARKETER added examples from the same reporting, including companies replacing CRM, ticketing, and other narrowly scoped systems with custom apps.
Market context
The broader market pressure comes from agentic arbitrage, where agents complete work across systems without forcing every user into a paid interface. Channel Dive, citing Gartner, reported that as much as $234 billion in SaaS spending through 2030 could be exposed to this shift. That does not mean every buyer cancels core enterprise software. It means vendors have to defend seat-based pricing against workflows that can be rebuilt, automated, or metered by outcome.
For practitioners
The savings claims are most credible for small firms replacing overbuilt tools, internal dashboards, workflow automations, or lightweight CRM processes where the data model is understood and failure modes are low. The same logic is weaker for regulated systems, deeply integrated workflows, or products that carry audit, uptime, permissions, and support obligations. AI can compress the first build, but it does not remove product ownership.
What to watch
Watch whether these replacements survive maintenance after the first cost-cutting cycle. Strong evidence would include retention of custom apps after six to twelve months, lower incident load, clear data ownership, and budget moving from licenses to AI usage, integration work, and security review. SaaS vendors will likely respond with agent pricing, usage credits, and workflow-level packaging rather than simple seat discounts.
Key Points #
- 1AI coding tools make selective SaaS replacement plausible when workflows are narrow, data is accessible, and maintenance ownership is clear.
- 2The savings story is strongest for SMBs replacing overbuilt licenses, not enterprises with deep compliance and integration dependencies.
- 3SaaS vendors face pricing pressure as buyers compare seat licenses against usage-based internal applications and agentic workflows.
Scoring Rationale #
The story is notable because it ties concrete SMB software cancellations to the larger agentic-arbitrage pressure on SaaS pricing. The score is moderated because the company sample is small and the strongest claims depend on reported savings rather than long-term operating-cost evidence.
Sources #
Public references used for this report. Practice with real Ad Tech data
90 SQL & Python problems · 15 industry datasets
[Active Search Campaigns by BudgetEasy](/problems/sql/active-search-campaigns-by-budget)
[High CPC Clicks & Poor Landing PagesMedium](/problems/sql/high-cpc-clicks-poor-landing-page)
[Campaign ROAS by Attribution ModelHard](/problems/sql/campaign-roas-by-attribution-model)
250 free problems · No credit card