Automation projects fail when they start from "what could we automate?" They succeed when they start from "where are people doing repetitive work that a computer should be doing?" The second question has better answers.
The test for a good automation candidate
Score a process against these five. Four or more "yes" answers usually means it is worth costing:
- High volume — happens daily or many times a week.
- Rule-based — you can write down exactly what to do in each case.
- Structured input — the data arrives in a consistent format.
- Error-prone when manual — mistakes happen and cost something.
- Currently a copy-paste job — someone moves data between two systems.
The strongest single signal: a person opening two screens and typing what they see in one into the other. That is not work — that is a missing integration, and it is almost always worth automating.
Five processes that reliably pay back
1. Enquiry capture and routing
Leads arrive by web form, WhatsApp, phone, and email. Someone manually logs them into a spreadsheet, or nobody does and they get lost.
Automated: every enquiry lands in one system, is assigned by rules, triggers an instant acknowledgement to the customer, and escalates if nobody responds within a set time.
Typical payback: fewer lost leads, which usually dwarfs the build cost.
2. Order to invoice to accounting
An order is placed, someone re-enters it into billing software, then someone re-enters it again into accounts.
Automated: one entry flows through automatically, with GST applied correctly and the invoice sent to the customer without human involvement.
3. Document data extraction
Supplier invoices, purchase orders, and delivery notes arrive as PDFs or photos, and someone types their contents into a system.
Automated: extraction pulls the fields, flags anything low-confidence for human review, and posts the rest. This is a strong AI use case because the human stays in the loop for exceptions.
4. Recurring reporting
Someone spends the first morning of every week assembling the same report from three systems.
Automated: the report generates itself and arrives by email or WhatsApp on schedule. Low complexity, immediate and visible time saving — often the best first project.
5. Reminders and follow-ups
Payment reminders, appointment confirmations, renewal notices, and abandoned enquiries all depend on someone remembering.
Automated: scheduled messages by WhatsApp, SMS, or email based on rules, with responses logged automatically.
Calculate the payback honestly
| Input | Example |
|---|---|
| Hours per month spent on the task | 40 |
| Fully-loaded hourly cost of that staff time | ₹250 |
| Monthly cost of doing it manually | ₹10,000 |
| Automation build cost | ₹2,00,000 |
| Annual maintenance (~18%) | ₹36,000 |
| Simple payback period | ~26 months |
That example is marginal — worth doing only if error reduction or capacity for growth adds value beyond the hours saved. Run this arithmetic before commissioning anything; it prevents expensive enthusiasm.
If you cannot express the saving in hours or errors per month, you are not ready to automate the process — you are still discovering it.
What to fix before automating
Automating a broken process produces broken results faster. If the current workflow has undocumented exceptions, informal workarounds, and disagreement about who owns which step, fix that first. Automation makes a good process cheaper and a bad process worse.
- Write the process down, including every exception people actually handle.
- Confirm the people doing it agree that is how it works.
- Remove steps that exist only because "we have always done it".
- Decide who owns exceptions once the routine cases are automated.
Start small and prove it
- Pick one process with clear, measurable time cost.
- Measure the before state — hours, error rate, turnaround time.
- Automate the routine 80%, routing exceptions to a human.
- Run both in parallel briefly so you can compare and build trust.
- Measure again, then use the proven result to justify the next one.
Not sure which of your processes is worth automating? Describe where your team loses time — we will tell you honestly which ones pay back and which do not. See also API integration and AI agents for Indian businesses.