Two tools, very different costs
An AI agent is non-deterministic, expensive per run and great at fuzzy decisions. Traditional automation (Python scripts, n8n, Zapier-style) is deterministic, cheap and great at well-defined steps.
A simple decision rule
If a competent intern could write down the rule on one page, automate it. If they\'d need judgment on every input, use an agent.
Where each shines (real client examples)
- Invoice OCR & bookkeeping export ? traditional pipeline. Same fields every time.
- Inbox triage with reply drafting ? AI agent. Tone and intent matter.
- Daily price scraping of 200 SKUs ? traditional Playwright job. No judgment needed.
- Lead qualification + personalised outreach ? AI agent. Every lead is different.
- Weekly KPI report generation ? hybrid. Script gathers data, model writes the narrative.
Cost reality check
A naive agent that calls a frontier model on every step can cost 50–200x more than the equivalent script. Use the model only where reasoning is required, and keep the boring steps in code.
Need help drawing this line for your own workflow? Book a discovery call.