A recent Business Insider report about health insurer Curative replacing a costly Salesforce contract with an internally built CRM has spread quickly because it reinforces a compelling narrative: AI will allow companies to cancel expensive software subscriptions and build exactly what they need instead.
That may prove true for some organizations. For many others, it’s the wrong lesson.
The more important story isn’t that Salesforce is suddenly obsolete, or that every growth-stage company should ask an AI coding tool to rebuild its CRM. The real story is that CRM strategy is changing. The decision is no longer just about choosing the platform with the most features, the largest integration marketplace, or the strongest brand recognition. It’s about whether the system reflects how the business actually creates, manages, and converts customer relationships.
That distinction matters because a CRM isn’t simply a database with a sales pipeline attached. It’s the operating system that governs customer relationships. It determines how leads are captured, how opportunities are qualified, how follow-up happens, how information moves between teams, how revenue is forecast, and ultimately how leadership understands what’s happening inside the business.
When the CRM no longer reflects how the business actually operates, people stop working through it and start working around it. They create shadow spreadsheets, move conversations back into inboxes and Slack, stop trusting reports, add fields no one uses, and buy automation to compensate for process confusion. Before long, the company is paying for a system of record while its real operating logic lives somewhere else.
AI did not create that problem. AI has simply made it harder to ignore.
The build-versus-buy question has become more nuanced
There are good reasons to buy an established CRM. Mature platforms bring permissioning, uptime, reporting, integrations, compliance support, partner ecosystems, documentation, and years of institutional learning. Salesforce, HubSpot, Zoho, Pipedrive, Microsoft Dynamics and other major platforms exist because most organizations do not want to maintain the full burden of customer infrastructure themselves.
There are also good reasons some companies are questioning large SaaS contracts. The costs can compound quickly. Workflows can become rigid. Implementation debt can accumulate across years of rushed configuration. AI features may be layered onto old process assumptions. A company can find itself paying enterprise prices while still relying on manual coordination to get work done.
Salesforce itself is responding to this pressure by pushing deeper into agentic AI. Its Agentforce pricing page shows a mix of per-user, consumption-based and edition-based models, including Flex Credits for AI actions and Agentforce 1 editions. That is not incidental. The platform vendors understand that the next CRM contest is not only about storing customer data. It is about who controls the workflows, the automation layer, and the intelligence around the customer relationship.
That is why the Curative example is useful but should not be treated as a universal playbook. A company with the technical talent, process maturity, security discipline and leadership focus to build and maintain its own CRM is in a different position from a founder-led brand, emerging SaaS firm, hospitality group, wellness company, nonprofit or professional services business that simply wants better visibility and follow-up.
The real decision is not build or buy. It is which parts of the customer operating system should be standardized, which should be customized, and which should be redesigned before any software decision is made.
CRM failure is usually a process failure first
Most CRM problems are diagnosed too late in the technology stack. Leaders see low adoption and blame the tool. Sales teams complain about administrative work and blame the fields. Marketing cannot connect campaigns to revenue and blames attribution. Operations cannot produce clean reports and blames the dashboard.
Sometimes the tool really is wrong. More often, the business never made the underlying operating decisions clearly enough.
Who owns a new lead after it converts? What makes a prospect qualified? Which lifecycle stages are meaningful, and which exist because a template suggested them? What information is required before a handoff? Which follow-up actions should be automated? Which require human judgment? What should leadership be able to see weekly? What should never be automated because it carries reputational or relationship risk?
Those aren’t administrative questions. They’re strategic decisions that determine whether the CRM becomes a source of leverage or merely a place where teams document work after the important decisions have already been made.
This is also where AI changes the economics. If a company has well-defined workflows, clean data, sensible permissions and a clear escalation model, AI can reduce friction across research, routing, summarization, follow-up drafting, intake, reporting and service support. If those foundations are weak, AI can help teams create more activity around a flawed process.
The common thread is that AI search visibility, CRM architecture, workflow design and executive reporting are not separate operational problems. They’re different expressions of the same underlying question: how information moves through an organization, and how that information supports better decisions.
McKinsey’s 2025 State of AI research found that AI use is widespread, but most organizations have not yet embedded it deeply enough into workflows and processes to capture material enterprise-level benefits. That finding tracks with what operators see on the ground. The value is not in adding AI to every screen. The value is in redesigning how work moves.
Custom systems create freedom and responsibility
The appeal of an internally built CRM is obvious. It can match the business exactly. It can remove unused features. It can integrate with internal workflows without waiting on vendor road maps. It can be designed around the language, handoffs and priorities of the actual team.
But custom infrastructure carries obligations that are easy to underprice in the excitement of the first build. Someone has to maintain it. Someone has to document it. Someone has to secure it, monitor it, govern permissions, manage data quality, handle edge cases, train users, rebuild broken integrations, and make sure the system survives employee turnover.
That does not mean custom is wrong. It means custom has to be treated as an operating commitment, not a clever workaround.
Saltwater has written before about the cost of treating AI as a shortcut instead of a system. The same logic applies here. Many of the most expensive AI mistakes start when a company automates a weak process, skips governance or underestimates the maintenance burden behind a promising prototype.
For many companies, the strongest answer will be hybrid. Keep a stable commercial platform for the core system of record. Build lightweight AI-enabled workflows around it where the business needs speed, specificity or better coordination. Use automation to improve intake, routing, reporting and follow-up. Create custom dashboards for leadership. Connect sales, marketing, service and operations data in ways the base platform does not handle well. Remove unnecessary complexity before replacing the foundation.
That approach is less dramatic than announcing the end of SaaS. It is also more useful for most leadership teams.
What leaders should evaluate before making a CRM decision
Before renewing, replacing or building a CRM, leadership should separate software dissatisfaction from operating-system design. A practical review should include six questions.
First, what decisions does the CRM need to support? A system built for executive visibility is different from one built mainly for sales rep activity, customer support, marketing attribution or partner pipeline management.
Second, where does customer truth currently live? If the real context sits in inboxes, meeting notes, proposals, support tickets, spreadsheets and Slack threads, the CRM may be less a source of truth than a partial archive.
Third, which workflows are unique enough to justify customization? Not every preference deserves custom software. The strongest candidates are recurring workflows where specificity creates measurable speed, quality, accuracy or customer experience gains.
Fourth, what are the real costs? Subscription cost matters, but so do implementation, training, administration, integration, maintenance, data cleanup, reporting, internal adoption and the opportunity cost of leadership attention.
Fifth, what governance is required? AI-enabled CRM systems touch customer data, commercial judgment, communications and sometimes regulated information. Gartner has predicted that agentic AI will become increasingly central to brand interactions, while also emphasizing the need for data governance and transparency. That governance burden should be designed early, not added after something breaks.
Sixth, what should remain human? The best customer systems do not automate judgment out of the business. They remove avoidable friction so people can make better decisions, hold better conversations and follow through with more discipline.
The executive takeaway
The companies that benefit most from AI-enabled CRM will not necessarily be the ones that cancel the most SaaS subscriptions. They will be the ones that understand their customer operations clearly enough to know what should be bought, what should be configured, what should be automated, and what should be built.
That is also a version of the innovator’s paradox: the tool that promises transformation can become another layer of drag if the organization does not redesign the conditions around it.
That is the real lesson inside the Salesforce backlash. The old CRM conversation was often vendor-led: choose the platform, implement the fields, train the team, hope adoption follows. The new conversation has to be operating-model-led: define how growth actually works, then decide what technology should support it.
For growth-stage companies, that is a more demanding standard. It is also a better one.
A CRM should not be a place where customer activity goes to be recorded after the fact. It should be a living system that helps the business see clearly, act faster, protect relationships, and scale without losing intelligence.
AI makes that possible. It does not make the strategic work optional.
Ready to rethink your CRM strategy?
Whether you’re evaluating Salesforce, HubSpot, Attio, or a custom-built solution, the first question shouldn’t be Which platform should we choose? It should be How should our business actually operate?
If you’re asking those questions, we’d be happy to help.
→ Schedule a strategy conversation with Saltwater Interactive.
