The Automation Decision
A business process takes 20 hours weekly to execute manually. You need it automated. Two paths:
-
RPA (Robotic Process Automation): Buy a tool (UiPath, Automation Anywhere, Blue Prism). Configure it visually. Deploy bots to click your applications.
-
Custom Automation: Write code. Build integrations. Deploy to your infrastructure.
Both work. Neither is universally better. Choosing wrongly costs 6 months and six figures.
RPA: When It Wins
RPA tools excel at automating legacy systems with no APIs. Your finance team uses a 20-year-old ERP that's closed-source and unmaintainable. Humans log in, enter data, run reports manually. RPA bots mimic this: log in, enter data, run reports, save output.
RPA Strengths:
- No coding needed. Business analysts configure bots visually.
- Works with any UI, regardless of how janky.
- Implementation is fast (weeks, not months).
- Non-technical teams can maintain bots.
RPA Weaknesses:
- Brittle. When UI changes, bots break. Maintenance overhead grows.
- Expensive upfront: UiPath costs $500-5k/month per bot, plus implementation consultants.
- Difficult to scale beyond 10-20 bots without infrastructure sprawl.
- Hidden costs: bot maintenance, UI regression handling, Intelligent Document Processing (IDP) for complex data extraction.
RPA ROI: Best for high-volume, stable processes. If a process will run unchanged for 2+ years, RPA saves money. If it changes monthly, RPA becomes a money sink.
Custom Automation: When It Wins
Your ERP has APIs. Your business process is well-defined: fetch data, transform it, write to database. Custom code integrates services and does the job reliably.
Custom Strengths:
- Durable. Code doesn't break when UIs change (you control the data flow).
- Scalable. One codebase handles 100 instances of the process.
- Maintainable by developers. No vendor lock-in.
- Cheaper long-term. Build once, maintain indefinitely.
- Flexible. Easy to adapt to new requirements.
Custom Weaknesses:
- Requires engineers. Expensive upfront talent cost.
- Takes longer initially (8-12 weeks vs 2-4 for RPA).
- Requires infrastructure (servers, monitoring, alerting).
- Needs ongoing maintenance and testing.
Custom ROI: Best for core business processes that require frequent changes. If process evolution is expected, custom is cheaper long-term.
The Decision Matrix
| Factor | RPA | Custom |
|---|---|---|
| Speed to value | 2-4 weeks | 8-12 weeks |
| Upfront cost | $50-200k | $100-300k |
| Annual cost (years 2-5) | $100-200k | $20-50k |
| UI/System stability | Brittle | Durable |
| Learning curve | Days (for BAs) | Weeks (for engineers) |
| API availability | Not needed | Essential |
| Scaling to 50 processes | Difficult | Easy |
| Regulatory/audit trail | Good | Excellent |
Hybrid Approach
Many enterprises use both:
- RPA for legacy systems: Finance, HR, supply chain systems with no APIs.
- Custom for modern stacks: Cloud apps, databases, APIs.
Start with custom for green-field work. Use RPA only when APIs are unavailable and the process is stable.
Implementation Strategy for Startups
- Audit your processes: Which ones are manual? Do they have APIs?
- Prioritize by ROI: Hours saved × hourly cost. Pick top 3-5 processes.
- Check API availability: Can you build custom integrations?
- If yes: Build custom automation. Hire a contractor if needed.
- If no: Evaluate RPA. Pilot with 1-2 processes. If ROI is positive, scale.
For 99% of tech startups, custom automation is smarter. You likely have engineers. APIs are standard. RPA is insurance for legacy system lock-in, not a primary strategy.
Frequently asked questions
Can we build internal tools instead of RPA or custom automation?
Yes. Build a simple internal tool that non-technical employees use to run processes (upload data, click run, download results). Cheaper than RPA upfront, more maintainable than either. Requires 2-4 weeks of developer time.
What's the typical ROI timeline for RPA?
RPA breaks even in 12-18 months if automating a 40-hour/week manual process. If the process takes <5 hours/week, RPA rarely breaks even. Custom automation breaks even faster for high-volume processes (5-10 years ROI timeline is common).
How fragile are RPA bots when systems update?
Very fragile. A UI redesign, a new field on a form, or a changed button label breaks the bot. Maintenance overhead is 20-30% of total cost. For systems that update frequently, custom automation is far more stable.