The problem: not all automations are worth the same
Your team is operating at 100% capacity running the business. When someone mentions "AI" or "automation", the question that actually matters is not "can we do it?" but "should we do it first?"
Many mid-size companies make the same mistake: they choose to automate what looks most sophisticated or what a vendor pitches best. The result: they invest in projects that take months, generate little ROI, and don't address the real business pain.
The right criterion is simpler: automate what consumes the most time, generates the most errors, and can be implemented quickly. That is what produces measurable ROI in weeks, not months.
Criterion 1: Impact — how many manual hours are actually saved?
Before automating anything, answer this with data:
How many monthly hours does this process consume?
- Manual Excel reporting: 40–80 hours per month (typical for CFO teams)
- Data reconciliations: 20–60 hours per month
- Invoice or purchase order validation: 15–40 hours per month
- Collections follow-up: 10–30 hours per month
How many errors does it generate?
- Reporting errors: 5–15% of records with inconsistencies
- Reconciliation errors: 2–8% of transactions requiring manual correction
- Data validation errors: 3–10% of records duplicated or incomplete
The real impact is not just the time saved. It is the time saved × the cost of error × the frequency of error.
Concrete example:
A distribution company with 15 people in operations spends 60 hours per month on inventory reporting. The process generates errors in 8% of records, causing stockouts and returns. Monthly cost of errors: ~€3,500. Cost of hours: ~€2,400 (60 hours × €40/hour). Total impact: €5,900 per month.
With an AI agent that automates reporting and validates data in real time, that cost drops to ~€800 per month (minimal supervision). ROI: €5,100 per month. Payback: 2–3 weeks.
Criterion 2: Technical complexity — can it be implemented without rewriting your infrastructure?
Not all automations require the same level of technical complexity.
Low complexity (implementation in 2–4 weeks):
- Reports that extract data from existing systems (ERP, CRM, accounting)
- Data validation against simple rules (duplicates, empty fields, value ranges)
- Automatic dispatch of reports or alerts
- Classification of documents or transactions
Medium complexity (implementation in 4–8 weeks):
- Automation of processes requiring logical decisions (conditional approvals, routing)
- Integration with multiple systems simultaneously
- Data extraction from unstructured documents (PDFs, emails)
High complexity (implementation in 8+ weeks):
- Changes to existing business workflows
- Integration with legacy systems without APIs
- Processes requiring frequent human intervention
Practical rule: Always start with low complexity. Automate what is already structured in your ERP, CRM, or accounting system first. Avoid, for now, processes that require organizational changes.
Criterion 3: Implementation time — when do you see the ROI?
Time matters. A project that takes 6 months generates less ROI than one that takes 6 weeks, even if both save the same number of hours.
Prioritize processes that can be implemented in 4–10 weeks:
- Manual reporting: 4–6 weeks
- Data validation: 3–5 weeks
- Collections follow-up: 5–8 weeks
- Simple reconciliations: 4–6 weeks
Avoid, for now, processes that take more than 12 weeks:
- Business workflow redesign
- Integration with complex legacy systems
- Processes requiring changes to policies or governance
The prioritization matrix: Impact × Complexity × Time
Use this matrix to decide what to automate first:
| Process | Hours/month | Errors | Complexity | Time | Priority |
|---|
| Manual reporting | 60 | 8% | Low | 6 wks | FIRST |
| Invoice validation | 30 | 5% | Low | 4 wks | FIRST |
| Reconciliations | 40 | 6% | Medium | 8 wks | SECOND |
| Collections follow-up | 20 | 3% | Low | 6 wks | FIRST |
| Workflow redesign | 50 | 10% | High | 16 wks | LATER |
Select processes that meet:
- Impact: >30 hours per month OR >5% error rate
- Complexity: Low or Medium
- Time: <10 weeks
How to validate before you invest
Before committing to an automation, validate with real data:
- Measure the current process: How many hours does it actually consume? How many errors does it generate? What is the monthly cost?
- Define the expected outcome: How many hours will be saved? How many errors will be reduced? Over what timeframe?
- Calculate ROI in ranges: If you save 40 hours per month at €40/hour, the ROI is €1,600 per month. If you also reduce errors by €2,000 per month, the total ROI is €3,600 per month.
- Set a timeline: When will you see the first results? In 4 weeks? In 8 weeks?
A real prioritization example
A financial services company with 25 people in operations faces three problematic processes:
- Manual reporting: 70 hours per month, 10% error rate, cost of errors €4,000 per month
- Transaction validation: 25 hours per month, 3% error rate, cost of errors €800 per month
- Client follow-up: 15 hours per month, 2% error rate, cost of errors €300 per month
Applying the criterion:
- Reporting: Impact €5,200 per month, low complexity, 6 weeks → FIRST
- Validation: Impact €1,800 per month, low complexity, 4 weeks → SECOND
- Follow-up: Impact €900 per month, medium complexity, 8 weeks → THIRD
With this prioritization, the company implements reporting in 6 weeks, generates €5,200 per month in ROI, and then expands to validation. Within 10 weeks, cumulative ROI reaches €7,000 per month. Without changing the infrastructure or the ERP.
Conclusion: The criterion that works
Don't automate what sounds innovative. Automate what hurts.
Use the simple criterion: impact × complexity × time. Select processes that save >30 hours per month, can be implemented in <10 weeks, and don't require organizational changes.
Validate with real data before you invest. Measure hours saved, errors reduced, and ROI in weeks — not in promises.
If you need help identifying which processes to automate first in your organization, we can run a complimentary 15-minute diagnostic. We analyze your current processes, identify where the greatest impact lies, and present a concrete roadmap with estimated ROI.
