Introduction
For years, “replacing Salesforce” has mostly been a hypothetical conversation inside big companies that never actually happens — the switching costs, the customization, and the sheer size of the migration keep most enterprise customers locked in. But a smaller, quieter shift is underway at the other end of the market. According to reporting from The Information, small businesses are starting to walk away from Salesforce and HubSpot entirely, replacing them with custom-built applications made using AI coding tools like Claude Code, Replit, and Lovable.
The numbers behind this trend are notable enough to pay attention to: some companies say they’re cutting software costs by 40% to 80% by building their own tools instead of paying for enterprise CRM licenses. Here’s what’s actually happening, who’s doing it, and why this looks more realistic for small firms than it ever has for large ones.
Table of Contents
- The Company That Started the Conversation
- Why This Is a Small-Business Story, Not an Enterprise One
- What “Vibe Coding” a CRM Actually Means
- The Cost Math Behind the Shift
- The Trade-Offs Nobody’s Advertising
- What This Means for Salesforce and the CRM Market
- Conclusion
- Sources
The Company That Started the Conversation
The clearest example in the reporting is Greenleaf Management, an Atlanta-based real estate property and investment manager with about 55 employees. According to Dave Codrea, a partner at the firm, Greenleaf replaced Salesforce’s CRM software entirely with a custom application built using AI tools from Replit and Anthropic’s Claude Code. The result: the company now saves an estimated $100,000 annually, while the custom application reportedly costs around $300 a month to maintain — a cost reduction of roughly 96% compared to what it was paying before.
Greenleaf isn’t an isolated case, either. The broader reporting describes a pattern of small companies and startups across different industries taking the same approach: rather than negotiating a new CRM contract or absorbing a price increase, they’re having AI coding tools build them a replacement from scratch.
Why This Is a Small-Business Story, Not an Enterprise One {#why-this-is-a-small-business-story}
It’s worth being precise about the scope here, because it matters. The largest companies in the world aren’t rushing to rip out Salesforce and rebuild it themselves — the reporting is explicit that this trend is concentrated among smaller firms, not enterprise customers.
That distinction makes sense once you consider what actually makes Salesforce valuable at scale: deep customization, a massive ecosystem of integrations, compliance features, and years of accumulated internal workflows that would be genuinely risky to rebuild from scratch. A Fortune 500 company has too much operational complexity baked into its CRM to walk away easily.
A 55-person real estate firm doesn’t have that problem. Its CRM needs are narrower, its workflows are simpler, and the cost of a full enterprise license is proportionally much larger relative to its size. That’s exactly the kind of gap AI coding tools are suited to close — not because they can replicate everything Salesforce does, but because most small businesses were never using most of what Salesforce does in the first place.
What “Vibe Coding” a CRM Actually Means {#what-vibe-coding-a-crm-actually-means}
The term used in some of the coverage is “vibe coding” — building software by describing what you want in natural language and letting an AI tool generate and iterate on the actual code, rather than writing it by hand. Claude Code is Anthropic’s agentic coding tool, built to let people delegate coding tasks — including building and modifying full applications — through natural-language instructions rather than manual programming.
In practice, that means a business owner or partner without a traditional engineering background can describe the contact records, deal pipelines, and reporting views they need, and have a working internal tool built and refined through iterative conversation rather than a formal software development process. That’s a meaningfully different cost structure than hiring developers or paying for a fully-featured enterprise platform license.
The Cost Math Behind the Shift
The savings being reported aren’t small, and they follow a consistent pattern across the coverage:
| Detail | Figure |
|---|---|
| Greenleaf’s estimated annual savings | ~$100,000 |
| Custom app’s monthly maintenance cost | ~$300 |
| Approximate cost reduction | ~96% |
| General reported savings range across firms | 40% to 80% |
| Tools commonly used | Claude Code, Replit, Lovable |
That range — 40% to 80% — matters more than the single standout example. It suggests this isn’t one company getting a lucky outcome; it’s a pattern showing up across multiple small businesses making similar decisions.
The Trade-Offs Nobody’s Advertising {#the-trade-offs-nobodys-advertising}
It would be misleading to present this as a strictly upside story, because building your own software instead of buying it comes with real trade-offs that don’t show up in a savings figure:
- Ongoing ownership. A purchased CRM comes with a vendor responsible for updates, security patches, and support. A custom-built tool puts that responsibility on the business itself, even if AI tools made the initial build faster.
- No ecosystem. Salesforce’s real value for many customers isn’t the core CRM — it’s the thousands of third-party integrations and pre-built extensions available around it. A custom app starts with none of that.
- Institutional knowledge risk. If the person who built and understands the custom application leaves the company, that knowledge can be harder to replace than switching to a new vendor would be.
- Scaling uncertainty. What works well for a 55-person company may not hold up cleanly at 200 or 500 employees, where the complexity Salesforce is built to handle starts to matter more.
None of these trade-offs make the approach a bad idea — they just explain why this remains a small-business trend rather than something large enterprises are rushing to copy.
What This Means for Salesforce and the CRM Market {#what-this-means-for-salesforce-and-the-crm-market}
For Salesforce specifically, this trend is unlikely to move the needle at the enterprise level anytime soon — its largest customers aren’t the ones being described in this shift. But it does point to a real vulnerability at the low end of the market: small-business customers who were paying enterprise-grade prices for a fraction of enterprise-grade needs.
More broadly, this is a preview of a wider dynamic that AI coding tools are introducing across software categories, not just CRMs. Any product priced primarily for feature breadth and ecosystem depth — rather than for a specific, narrow use case — is more exposed to being replaced by a small, custom-built alternative once building that alternative becomes cheap and fast enough for a non-engineer to do it.
Conclusion
Greenleaf Management’s story is a useful data point, not proof of an industry-wide collapse — Salesforce and HubSpot aren’t losing their large customers over this. But the pattern underneath it is worth watching: when small businesses can describe what they need and have a working internal tool built for a few hundred dollars a month instead of a five- or six-figure annual license, the calculation for paying for more software than you actually use gets a lot harder to justify. As AI coding tools keep improving, this gap between what small businesses need and what they’re paying for is likely to keep narrowing.
