When we tell people we cut operational costs by 59% for a staffing client using RPA, the most common response is scepticism. Fair enough. That number sounds like marketing. So let us walk through exactly what happened — the boring, technical version.
The situation before we came in
The client was a mid-sized staffing firm in Bengaluru. Around 40 employees. Their operations team was spending roughly 60% of their working day on three things: entering candidate data from email into their internal system, generating compliance reports by copying from spreadsheets, and sending payroll instructions to their accounts team via WhatsApp messages.
Not a broken business. A profitable one. But one where every growth hire added more manual overhead, and where errors were a daily occurrence because humans copying data will always make mistakes.
What we actually automated
We mapped every repetitive workflow over two weeks before writing a single line. This is the step most people skip. The map revealed four distinct automation candidates:
Candidate intake: Emails with CVs were being manually read, summarised, and entered into their ATS. We built a pipeline that parses incoming emails, extracts structured candidate data using a combination of regex and a lightweight LLM classifier, and creates the record automatically — flagging anything below a confidence threshold for human review.
Compliance documentation: Every placement required a compliance checklist — right to work, references, sector-specific certificates. This was being tracked in a shared spreadsheet and emailed manually. We automated the checklist generation, document chase emails, and status tracking.
Payroll instructions: Friday afternoons were chaos. Someone would compile hours from three different sources, build a summary, and WhatsApp it to accounts. We replaced this with an automated aggregation and PDF generation workflow that runs every Thursday at 5pm and emails a structured report directly.
Reporting: Monthly client reports took one person two days. We automated the data pull and report generation — the human now reviews and adds commentary in under two hours.
The 59% number — where it actually comes from
Before automation, the ops function cost the business approximately ₹4.2L per month in fully-loaded staff time for these tasks. After, that number dropped to ₹1.7L — accounting for the time staff still spend on exception handling and review. That's the 59%.
What we didn't count in that number: the reduction in errors (which previously caused delayed placements and client complaints), the faster turnaround on candidate processing (from 48 hours to under 4), and the fact that they were able to scale placements by 30% without adding ops headcount. If you fold those in, the real business impact is considerably higher.
What we'd do differently
The candidate intake pipeline took longer than it should have because we underestimated the variety of CV formats. We built a rigid parser first, then had to rebuild it with more flexibility. The lesson: assume the messiest possible input from day one.
We also underinvested in the exception-handling UI. When the automation flags something for human review, the person needs to see exactly what was flagged and why — in plain language. Our first version was too technical. We rebuilt it.
These aren't failures — they're the normal cost of building something real. But they're worth knowing about before you start.
Is RPA right for your business?
The honest answer: only if you have clear, repetitive workflows with structured inputs and outputs. If your processes are highly variable, or if the rules change frequently, the maintenance cost of automation can outweigh the savings. The mapping exercise we do upfront is specifically designed to tell you which category you're in before you commit to anything.