Where to start with automation: the people × hours method
Automation starts with a process inventory: write down who does each job, how many hours a day, and how often. Then look at rule clarity and exception rate. The three jobs that take the most hours, have clear rules and few exceptions are the first candidates. One is chosen and tried in a 30-day pilot.
Why start with an inventory?
Automation projects often start from the wrong place. Either the loudest complaint is chosen or the job on a manager's mind. Yet inside a company there are jobs nobody complains about that take hours every day: the Excel file opened each morning, the order copied from email into the ERP, the month-end matching.
An inventory makes these jobs visible. The aim is not a perfect table but a list simple enough to fill in during one meeting. Each row describes a job, and next to each job there is a person and a duration.
An inventory has a side benefit: it shows the jobs' shared input. Three different jobs may in fact be fed from the same folder of PDFs. Then the first automation also prepares the groundwork for the ones after it.
How are people × hours calculated?
The formula is one line: how many people, how many hours a day, how many days a month. If someone prepares a report for 2 hours every morning and there are about 22 working days a month, this job takes 44 hours a month. That is about 500 hours a year. If two people do it, it is double.
This figure is an estimate and should stay one. You ask the person who does the job; you do not use a stopwatch. The aim is not exact accuracy but ranking jobs against each other. It is enough to know that a job said to take two hours is bigger than one said to take half an hour.
How is the process inventory filled in?
One Excel sheet is enough. Each row has these columns: job name, person doing it, daily duration, frequency, input type (PDF, Excel, email, WhatsApp), system used and when it last broke down.
To fill it in, half an hour with one person from each department is enough. Ask accounting, purchasing, the warehouse and sales operations the same question: what are the jobs you repeat on the computer during the day? Most answers gather around the same three or four jobs.
A small example: a purchasing specialist compares the order confirmations in supplier emails with the orders in the ERP every day and spends an hour and a half a day on it. An accounting clerk enters PDF invoices by hand for an hour a day. A manager pulls a sales summary from Excel for half an hour every morning. Three rows, three different durations, three different inputs. This is how the inventory takes shape.
Which four measures are used for ranking?
Four questions are asked for each row.
- Hours: how many people × hours a year does this job take? Big hours, big opportunity.
- Frequency: every day, weekly, once a year? A frequent job teaches the rule fast and shows the gain fast.
- Rule clarity: can the person doing the job explain how they do it? Comparing the total on an invoice with the order can be written down. Price negotiation with a customer cannot.
- Exception rate: how many cases fall outside the rule? If one in ten records is an exception, the approval queue is manageable. If half are exceptions, the job is not yet ready for automation; the rule has to be clarified first.
You can give each measure a score from 1 to 3 and add them up. But the score does not decide; it eases the conversation. The final decision is made by the person who knows the job.
How are the first three candidates chosen?
The first candidates are those with high hours, a clear rule and few exceptions. They usually gather in three jobs: reading documents and entering them in the ERP, matching two lists, and preparing regular reports.
- Document entry: a PDF invoice, waybill or emailed order is read and written into the ERP as a draft.
- Matching: a bank statement is compared with customer/supplier account movements, an order with an invoice, a count result with stock.
- Reports: if the morning sales, collection or stock summary is prepared by hand in Excel, an agent reads the same data and sends it.
Let's score the same example. Order confirmation comparison: high hours, daily frequency, clear rule (do the number and amount match), few exceptions. PDF invoice entry: medium hours, daily frequency, clear rule, exceptions vary by supplier. Morning summary: low hours but a fully clear rule and no exceptions. All three are candidates; your own hours decide the order.
Only one of these three at the top of the list is chosen. The reason is simple: if you start three jobs at once, who approves what gets muddled and the pilot's result becomes unreadable.
Jobs not to start with
A few jobs should not be the first candidate. A rule that exists only in one person's head and has never been written down is not suited to a first trial. Jobs done once a year show their gain late. Correspondence that directly affects the relationship with a customer stays with a person; at most the agent writes a draft.
How is a 30-day pilot planned?
For the one process you chose, sample documents are collected first: real PDFs, Excel files and emails from the last month. At Panda a demonstration is given with these samples within 3 business days; you see how the agent reads the document and what it matches, in your own files.
The 30 days of the pilot divide roughly into three phases. In the first weeks the agent prepares every record as a draft and a person approves all of them. In the middle, corrections turn into rules. In the last week it is decided which records can pass automatically and which stay at approval.
At the start of the pilot a number is written down: how many people × hours this job takes now. At the end the same number is calculated again. If the difference is small, the cause is found; perhaps the exception rate is higher than estimated or the approval queue is waiting. If the difference is large, you move on to the second process.
How is the team involved in this work?
The person who does the job today is the most important partner in the process. They know the rule best and are the first to notice an exception. Automation does not replace the person; the team does not shrink, its work changes. The person who did data entry becomes the person who checks the draft.
That is why, when building the inventory, you ask the person and you do not make decisions about the person. Duration is treated as the cost of the process, not as the person's performance. Otherwise everyone reports their job as shorter and the inventory comes out wrong.
By the end of the pilot your ERP is still the same and data has not left the company. What you have is only a better inventory for the next process and a real measurement.