It's a conversation we're hearing more often: "We asked a developer with Claude Code to build a CRM mockup, and in three weeks we had something that actually works. Why do we need Salesforce again?"
On the surface, it's a fair question. Generative AI has fundamentally changed what an engineer can build alone, and how fast. A skilled developer with LLM assistance can now prototype, iterate and ship features at a pace that felt impossible five years ago. And yes — you can build a CRM. You can build an ERP. You can build a lot of things.
But "can build" and "can run reliably at scale" are not the same thing. And if your company has bet on data and customers being a source of competitive advantage, they very quickly become different problems.
You can build a system in weeks. You'll be running it for years.
The DIY acceleration is real
Let's start with what's actually happened. Claude Code, GitHub Copilot and tools like them have genuinely transformed what individual developers can ship — dramatically accelerating prototyping, iteration and software delivery. A CRM prototype that might once have taken two months to write by hand can now be scaffolded, tested and iterated in a fraction of the time. The productivity gain is not hype — it's real.
At the same time, more companies are asking the question: if we can build it quickly, why are we paying for a platform licence? The maths looks simple. And if you're thinking about a greenfield MVP or a specific vertical use case, the temptation is real. We've explored the wider build vs buy CRM decision separately.
Where this logic breaks down is at scale, and at the moment you realise the cost of maintaining what you've built is not the cost of building it.
The things the demo doesn't show
A three-week CRM prototype looks polished. It feels fast. It handles the happy path beautifully. But move it to production — add real users, real data, real traffic — and you hit costs that didn't show up in the build phase.
Security governance
Can you audit who accessed which records? Do you meet compliance obligations? How do you rotate credentials, manage permissions across teams or handle data residency?
Performance at scale
Works fine with 100 records. Does it work with a million? Does your architecture handle concurrent users, indexing, query optimisation and the network roundtrips between services?
Integration
Your data lives in multiple places: warehouse, marketing platform, financial system. A CRM is not useful in isolation — it's a hub. How do you sync data cleanly and keep it fresh?
Maintenance and handover
You built it. What happens when you leave, or when a teammate needs to work on it? Can someone understand the code, trace a bug or add a feature safely in six months?
Change management and governance
When you push an update to production, how do you avoid downtime? How do you roll back? Who approves the change?
Each of these is solvable. But each one requires discipline, investment and often specialist expertise. The DIY path hasn't skipped these problems — it's deferred them. For AI specifically, we've looked in more detail at security and governance in agentic systems, while mature CRM environments also benefit from clear ongoing governance and ownership.
Platform is not the opposite of speed
The assumption baked into "why pay for Salesforce?" is that platforms are slow, bureaucratic and inflexible, whilst building is fast and lean. That framing is outdated.
Salesforce, and specifically Agentforce, exists partly because this trade-off is real. Yes, platforms come with constraints. But those constraints are the price of something more valuable: a governed, scalable, integrated system that your team can modify and extend without reinventing the fundamentals every time.
Agentforce in particular is built on the idea that AI should accelerate within a platform, not instead of one. You can build agents, custom logic and integrations directly. You get LLM-driven automation. But you inherit — at no extra cost to you — multi-tenancy, security controls, audit trails, backup and disaster recovery, API governance and compliance readiness. These are not features you bolt on later; they are the foundation.
And because Salesforce is vendor-neutral on AI — integrating with models such as Anthropic Claude, OpenAI and your own models — organisations can combine the speed and flexibility of leading AI models with the governance, data and operating foundations of an enterprise platform.
The real choice
The decision is not "custom vs. platform". It's: do we build to demo, or do we build to run?
If you're prototyping a use case, validating a hypothesis or exploring what's possible with AI, tools like Claude Code can be exactly the right choice. They let teams move quickly, test ideas and accelerate development before committing to a production architecture.
But if you're building something your business will depend on — something that will hold customer data, integrate with revenue systems or become a day-to-day tool — that's a different decision. The fact that a developer can build it in three weeks using AI is not the right measure of whether it's ready for production. The right measure is: can your Ops team run it? Can your Security team audit it? Can you scale it without rewriting it?
For most enterprises, that's not just a development decision. It's a platform, architecture and operating-model decision.
What to do next
If you're evaluating this now, ask your team:
- What would it cost to secure, scale and maintain this ourselves, over three years?
- Who owns the ongoing support and optimisation?
- What happens if there's a security incident or a compliance audit?
- Can we integrate this with our existing systems without custom API plumbing for every new tool?
Then compare that cost against the total cost of ownership of a platform like Salesforce — not just the licence, but the built-in security, scalability, compliance and vendor support.
If you're making the same calculation for AI and automation investment, our AI ROI Calculator can help quantify the potential value and build a clearer business case before you commit.
More often than not, the platform wins on total cost and total risk, even if the headline build cost was faster.
Ready to evaluate your options?
If you're weighing a DIY approach against a platform strategy, the total cost of ownership is usually clearer than the build-cost headline. We've worked with teams on both sides of that decision. Let's explore what's right for you.
