
Why I'll Never Implement Another ERP
A founder's journey from ERP hell to AI-native operations — and why you should skip the legacy trap entirely.
By Scott Roy Murphy
The Hidden Tax on Growth Companies
Here's a number that should terrify every founder: 75% of ERP implementations fail to meet their objectives. Not fail outright — fail to deliver what was promised. The system goes live, the consultants go home, and you're left with a $500K invoice and a platform that does 40% of what Excel did.
I've lived this. Three times across three companies. Each time, the pitch was the same: "This will be your single source of truth. Everything connected. Everything automated." Each time, the reality was the same: 18 months of configuration, change management theater, and custom workarounds for things the system was supposed to handle natively.
The Five Lies of Enterprise Software
- "It's out of the box." Nothing is out of the box. Everything requires configuration, customization, and consultants.
- "It'll pay for itself in 18 months." It won't. The TCO is 3-5x the license cost when you include implementation, training, and ongoing maintenance.
- "Your team will love it." Your team will build shadow spreadsheets within two weeks.
- "It scales with you." It scales with your wallet. Every new module, every new user, every new integration costs more.
- "AI is built in." A chatbot that searches the help docs is not AI. A recommendation engine trained on 2019 data is not AI.
What Actually Changes When AI Can Learn?
The fundamental problem with ERP is architectural. It's a rigid system that requires the business to conform to its data model. When the business changes — new product line, new market, acquisition — the system breaks.
AI-native operations flip this. Instead of forcing your business into a schema, intelligent agents observe how your business actually works and adapt to it.
| Dimension | Traditional ERP | AI-Native Operations |
|---|---|---|
| Implementation | 12-18 months | Days to weeks |
| Customization | Consultant-dependent | Self-adapting |
| User Adoption | Training-heavy | Natural language |
| Cost Model | License + modules + users | Usage-based |
| Intelligence | Static reports | Proactive insights |
| Maintenance | Upgrade cycles | Continuous learning |
The Post-ERP Playbook
If you're a growth company evaluating operations infrastructure, here's what I'd tell you:
- Don't buy a platform. Deploy agents. Individual, specialized AI agents that handle specific functions (finance, HR, inventory, compliance) and coordinate autonomously.
- Start with the pain. Don't boil the ocean. Pick your worst operational bottleneck and automate it first. Prove value in weeks, not years.
- Demand transparency. Every automated decision should have an audit trail. Every agent should explain its reasoning. "The AI did it" is not acceptable.
- Keep the human gate. Automation without oversight is a liability. The best systems know when to escalate, not just when to execute.
The ERP era is ending. Not because the vendors are going away — they'll sell maintenance contracts forever. But because a new generation of companies will never install one in the first place.
They'll build with agents instead.
See what AI-native operations looks like
Explore the Manifold platform — autonomous agents that run your operations.
Explore Carborundum AI