A company needed to transfer data from their CRM to their accounting system. Three vendors pitched solutions: RPA vendor: "Our bot will click through both systems and copy the data. $20K, ready in 2 weeks." Integration vendor: "We'll connect the APIs. $50K, ready in 6 weeks." Custom software vendor: "We'll build exactly what you need. $150K, ready in 4 months."
The company chose RPA for speed and cost. Two years later, they're spending $40K annually on bot maintenance, the process breaks every time either system updates, and they're considering rebuilding properly.
Choosing the right automation approach isn't about cost or timeline. It's about understanding what each approach can and can't do.
Understanding Your Options
RPA (Robotic Process Automation)
RPA bots interact with software through the user interface—clicking buttons, filling forms, copying data. They automate what a human would do manually.
Strengths: - Quick to implement (weeks, not months). No API access needed. Can work with legacy systems without APIs. Lower upfront cost. Good for processes that rarely change.
Weaknesses: - Fragile—breaks when UI changes. Slow—limited by UI speed. Can't handle high volumes. Requires ongoing maintenance. Difficult to test and debug. No transaction guarantees.
Best for: Temporary solutions, processes across systems without APIs, low-volume tasks, situations where API access is denied.
API Integrations
Integrations connect systems programmatically through APIs. Data flows between systems without human interaction.
Strengths: - Fast and reliable. Handles high volumes. Survives UI changes. Provides transaction guarantees. Easier to test and monitor. Lower ongoing maintenance.
Weaknesses: - Requires API access. Longer to implement than RPA. Higher upfront cost. Dependent on API quality and limitations. May require vendor cooperation.
Best for: High-volume data transfer, real-time synchronization, critical business processes, long-term solutions.
Custom Automation
Build software specifically for your needs. Full control over functionality, integration, and user experience.
Strengths: - Exactly fits requirements. Optimized for your workflow. Full control and ownership. Can combine multiple integrations. Can add business logic, validation, transformation.
Weaknesses: - Highest upfront cost. Longest to implement. Requires ongoing maintenance. Need internal or external development team.
Best for: Complex workflows, unique requirements, strategic advantage, situations where commercial solutions don't fit.
Decision Framework
Choose based on these factors:
Factor 1: Volume and Frequency
Low volume (<100 transactions/day), infrequent (weekly/monthly): RPA might be acceptable. Low transaction volume means slow UI interaction isn't a problem.
Medium volume (100-1000/day), regular (daily): Prefer integrations. UI automation becomes bottleneck. Need reliability.
High volume (>1000/day), continuous: Integrations or custom. RPA can't handle volume. Need performance and reliability.
Factor 2: Process Stability
Process changes frequently: Avoid RPA. Every UI change breaks the bot. Integration or custom is better—changes are contained to API endpoints or custom logic.
Process is stable: RPA is viable. If UI rarely changes, maintenance burden is manageable.
Factor 3: Criticality
Nice to have, low business impact: RPA acceptable. If it breaks, you can fall back to manual process temporarily.
Important, moderate business impact: Integration preferred. Need reliability but can tolerate occasional failures.
Mission critical, high business impact: Integration or custom. Can't afford downtime or data loss. Need transaction guarantees and rollback capability.
Factor 4: API Availability
No APIs available: RPA or custom screen scraping. Can't integrate programmatically. UI automation is only option (besides rebuilding the systems).
APIs available but limited: Hybrid—integration where APIs work, RPA for gaps. Or custom software that handles API limitations.
Good APIs available: Integration or custom. No reason to use fragile UI automation.
Factor 5: Timeline
Needed immediately (weeks): RPA only option. Can't build proper solution that fast.
Reasonable timeline (1-3 months): Integration feasible. Time to connect APIs properly.
Strategic, long-term (3+ months): Consider custom. Time to build exactly what you need.
Factor 6: Budget
Limited budget (<$50K): RPA or simple integration. Can't afford custom development.
Moderate budget ($50-150K): Integration or simple custom. Can build proper solution.
Substantial budget (>$150K): Custom is option. Can build sophisticated solution.
Common Scenarios
Scenario 1: Transferring Sales Data from CRM to Accounting
Context: 50 sales orders per day, stable process, both systems have APIs.
Recommendation: Integration. Build API connection between CRM and accounting system. Medium volume justifies investment. APIs available mean reliable solution is possible. Stable process means low maintenance.
Why not RPA: Too many transactions for UI automation. Process is important—can't afford frequent breaks.
Scenario 2: Extracting Data from Legacy Mainframe
Context: Monthly report generation, mainframe has no APIs, data doesn't change format often.
Recommendation: RPA or custom screen scraping. No APIs means can't integrate programmatically. Low frequency (monthly) means slow UI automation is acceptable. Consider custom screen scraping if RPA proves unreliable.
Why not integration: No APIs to integrate with.
Scenario 3: Complex Order Processing Workflow
Context: Multi-step process involving data validation, external lookups, approvals, multiple system updates. Unique business logic.
Recommendation: Custom software. Complex logic needs proper programming. Multiple integrations better managed in custom code. Unique requirements won't fit off-the-shelf solutions.
Why not RPA: Too complex, too many steps, too much business logic. Would be maintenance nightmare.
Why not simple integration: Business logic and workflow orchestration too complex for point-to-point integration.
Scenario 4: Temporary Process During System Migration
Context: Moving from old CRM to new CRM over 6 months. Need to keep systems in sync during transition.
Recommendation: RPA (reluctantly) or simple custom sync tool. This is temporary—investments in proper integration won't pay off. After migration, old system is gone.
Key consideration: This must be truly temporary. If "temporary" lasts years, invest in proper solution.
Hidden Costs of RPA
RPA appears cheap but has hidden costs:
Maintenance: UI changes break bots. Every software update requires bot testing and likely fixes. Ongoing cost adds up—often 20-40% of initial development cost annually.
Error handling: Bots can't handle unexpected situations well. Human must monitor, intervene, fix errors. Time-consuming.
Scaling limitations: Adding more bots to handle volume gets expensive. License costs per bot add up. Each bot needs its own virtual machine.
Technical debt: RPA becomes legacy problem. Eventually have to replace it with proper solution. Meanwhile, you've spent years maintaining fragile automation.
When to Upgrade from RPA
Signs your RPA solution needs replacement:
At that point, invest in proper integration or custom solution. You'll recoup investment through reduced maintenance and improved reliability.
The Hybrid Approach
Sometimes best solution combines approaches:
Core process via integration, edge cases via RPA: Use API integration for 95% of transactions. RPA handles exceptions that don't fit API limitations.
Integration platform with custom logic: Use commercial integration platform (like Zapier, Make, Workato) but add custom code for business logic.
Custom software with embedded RPA: Build custom application that uses RPA libraries to interact with systems lacking APIs.
Making the Decision
Use this checklist:
Choose RPA if: - Volume is low (<100/day). - Process is simple and stable. - Temporary solution (really temporary, not "temporary"). - No APIs available. - Budget is very limited. - Timeline is urgent.
Choose integration if: - Volume is medium to high. - APIs are available. - Process is important. - Looking for long-term solution. - Budget allows for proper implementation.
Choose custom if: - Requirements are complex or unique. - Need to combine multiple integrations. - Significant business logic required. - Integration would provide competitive advantage. - Commercial solutions don't fit needs.
The Bottom Line
RPA is not inherently bad. It solves real problems. But it's a specific tool for specific situations—not a universal automation solution.
Integrations are the default choice when APIs exist. Faster, more reliable, lower maintenance than RPA. Only reason not to use integration is lack of APIs.
Custom software is for unique needs or strategic advantages. Costs more upfront but delivers exactly what you need.
Don't choose based on initial cost alone. Factor in ongoing maintenance, reliability, and longevity. The cheapest upfront option often costs most over time.
