1. The matchup at a glance
The decision isn't about which one is "better" in the abstract — it's about how much weight you put on each criterion. Here are the eight that actually matter.
| Criterion | Odoo (buy / extend) | Custom ERP (build with AI) |
|---|---|---|
| Upfront cost | Per-user/month licensing (Enterprise) + integration days. Community is free but limited. | Near zero on licensing. The cost = your scoping time + your AI subscription. Development costs have collapsed. |
| 3-year total cost (TCO) | Grows with user count and customization. Custom module maintenance is an ongoing expense. | Hosting + managed database = a few tens of euros/month, stable regardless of user count. |
| Time to go live | Fast on standard features: existing modules run right away (accounting, CRM, inventory). | Everything needs to be built — but much faster than before: a functional core in days, not months. |
| Business flexibility | Good within Odoo's framework; beyond that, you're building modules and working around its imposed data model. | Total. The software fits your exact business, not the other way around. No forced checkboxes. |
| Maintenance & updates | Core maintained by the vendor, but every major version upgrade can break your custom modules. | 100% your responsibility — but it's a smaller codebase, fully yours, and AI speeds up fixes too. |
| Lock-in | Vendor (roadmap, pricing, terms) + often a partner integrator. | You / your team. No license ransom, but the knowledge must stay documented and shared. |
| Compliance (accounting, payroll) | Covered and kept up to date by the vendor (standards, tax rules). A huge advantage for anything regulated. | Not worth reinventing. The smart move: let an external source of truth handle the regulatory layer. |
| Risk / reversibility | Low technically (proven product), but exiting is costly once you're deep in the ecosystem. | Quality risk if nobody reviews; but you own your code and data, so you can always walk away. |
2. Odoo in detail
Odoo is a very complete modular management suite (CRM, sales, purchasing, inventory, accounting, HR…), available as open source Community and paid Enterprise. It's an excellent default for the vast majority of SMBs: the standard is solid, the community is huge, and the essentials of a typical business are covered without writing a single line of code.
Where Odoo wins: fast onboarding for generic functions, accounting and tax compliance maintained for you, a rich ecosystem of modules and integrators, and a clean exit path when you stay within the framework. At Cobra, that's exactly my choice: an Odoo I extend almost every day with Claude Code, rather than rewriting everything from scratch.
Where it gets tricky: the moment your business goes outside the framework, you're building modules, working around Odoo's data model, and every version upgrade becomes a regression test for your customizations. Community's "free" label melts quickly into integration and maintenance costs.
3. Custom ERP in detail
Building your own isn't the six-figure, 18-month project it used to be. With AI doing the heavy lifting and a human at the wheel, a functional core comes together in days. The point isn't to "do what Odoo does but worse" — it's to model exactly your business, with your UX and your brand identity.
Where custom wins: total flexibility, near-fixed operating costs (no per-user licensing), ownership of the code and data, and an experience tailored to your team — not a generic compromise. That's decisive for a business the market doesn't model well.
Where it gets tricky: everything depends on your ability to decide and review. No vendor safety net — quality, backups, security, compliance are all on you. And you absolutely shouldn't rewrite generic, critical functionality (accounting) just for the pleasure of going custom.
4. Who should choose what
Standard business, team needs to move fast
- Accounting, payroll, inventory, purchasing at the core
- Strong regulatory compliance requirements
- Non-technical team, no in-house AI driver
- Growth through existing modules
Specific business, strong UX requirements
- Processes the market doesn't model (or models poorly)
- You want your own interface and brand, not a vendor's
- A founder / team that knows how to drive AI
- Need for very specific dashboards and automations
And most of the time, the real answer is hybrid: an off-the-shelf foundation for generic and regulated needs, with custom-built pieces on top for what differentiates you. That's exactly what I do in practice.
6. Frequently asked questions
Odoo or custom ERP: which should you choose?
Odoo (or any off-the-shelf ERP) when your business is standard and regulated — accounting, payroll, inventory, purchasing: you don't rewrite accounting logic, and a proven foundation covers 80% of your needs right away. A custom ERP when your business is specific, when no software can model it without awkward workarounds, or when you want your own UX and brand identity. Now that AI has collapsed the cost of customization, going custom becomes rational much sooner than it used to.
How much does a custom ERP built with AI cost?
The cost that's collapsing is development: HF-OS, a fully functional press ERP (CRM, ad management, invoicing, follow-ups), was built in ~16 hours of actual work over 5 days and 96 pull requests, by one person driving Claude Code. Operating costs come to a few tens of euros per month (hosting + managed database + backups) plus the AI subscription. Compare that to per-user licensing and integration days for an off-the-shelf ERP. The real remaining cost of going custom isn't money — it's the time spent scoping and reviewing.
Is Odoo really free?
Partly. Odoo Community is open source and free when self-hosted, but limited (no full accounting or certain apps). The Enterprise version is paid, billed per user per month, and the real bill is mostly integration, customization, and module maintenance. "Free" doesn't mean "no cost": you need to look at total cost of ownership, not just the license price.
Do you need to know how to code to build your ERP with AI?
Not to write the code line by line — AI handles that. But you need to know how to decide and review: define the scope, make architecture calls, set guardrails (backups, access, personal data), and do the reading. It's steering work, not typing work. A founder who knows their business and learns to drive AI can produce the output of a small team; someone who doesn't want to review anything shouldn't risk it on a critical tool.