Why does your back office grow faster than your revenue?

Revenue grows, and so does the number of people needed to process it. Each new customer brings orders to enter, invoices to raise, payments to chase and questions to answer. Finance, the order desk and customer service hire to keep up, and each hire takes months to find and train. By the time the team is at full strength, volume has moved on again, and the margin you expected from growing has gone on salaries.

Our view is that this is mostly a design problem, and a fixable one. Most back-office work is a set of small steps repeated once per order, invoice or ticket. When a person has to do every one of those steps, headcount grows in step with volume. Take the person out of the routine steps and leave them at the exceptions, and the team can grow much more slowly than the business.

Why the back office scales with volume

Sales and marketing effort can reach more customers without growing at the same rate. Back-office work grows with the count. Each order is entered once, checked once, invoiced once and chased if it isn't paid. A tenth more orders means roughly a tenth more of those steps, and nobody is working any faster.

The steps also get slower as the company grows. More customers means more special terms, more exceptions and more systems that don't talk to each other. A step that took five minutes with a hundred customers takes eight with a thousand, because someone has to check which price list applies or look an account up in two places. That's how the back office ends up growing faster than revenue.

How to spot work that scales with volume

Ask one question of each task your finance, operations and service teams do: does it happen once per something? Once per order, per invoice, per ticket, per new customer, per delivery or per refund. If it does, it grows with the business. Most companies have some of each of these.

• Per order: entering orders from email or PDF, checking stock and credit, sending acknowledgements.
• Per invoice: keying supplier invoices, matching them to purchase orders, chasing approvals, raising customer invoices.
• Per ticket: reading and routing requests, looking up the order or account, answering where-is-my-order questions.
• Per customer: setting up the account in the CRM, the ERP and billing, checking credit, collecting documents.

Work that happens once a month or once a year, such as the board pack, the budget or the audit, grows more slowly. It can still be worth improving, but it isn't what's driving the headcount.

Measure the time each unit takes

Once you have the list, put two numbers on each task: minutes per unit and units per month. Multiply them and you have the hours the task costs today. Run the same sum at the volume you expect next year, and you have the hiring plan nobody has written down.

The numbers don't need to be precise. Ask the people who do the work to time ten examples, or use a week of timesheets, and take the middle value. Take a hypothetical order desk where entering an order takes twelve minutes and 2,000 orders arrive a month. That's 400 hours a month, close to three full-time people. If orders grow by half next year, so does that figure, before you count the slowdown from extra complexity.

Then split each task's time into its steps. Most break down into the same few: re-keying from one place into another, checking against a record, chasing someone for a missing piece, and deciding. The split shows where the time goes.

Which steps to automate first

Start with steps that meet three conditions.

• The most total hours: minutes per unit times units per month, from the sum above.
• Little judgment: re-keying, matching and looking things up need care, but someone who knows the customer adds little to them.
• Inputs you can reach: the order arrives in an inbox you control, the invoice in a known folder, the ticket in your ticketing tool.

Re-keying usually comes out on top, because it's frequent, routine and spread across many people. Matching and checking come next. Chasing often shrinks on its own once the earlier steps are fixed, because orders and invoices that are entered correctly and completely the first time leave less to chase. If you're comparing candidates across several teams, which AI use case to fund first has a fuller scoring method.

Keep people for the exceptions

The team stays, and its work changes. Every per-unit process has cases that need judgment: a customer disputing a price, an invoice that doesn't match its order, an order for an unusual quantity. Those go to a person with everything they need in front of them, and the routine cases run without anyone touching them.

That changes what growth costs. Volume still grows, and so do the exceptions, but exceptions are a small share of the total. A team that spent most of its day on routine steps now spends it on the cases that need a person, on customers, and on improvements nobody had time for. You'll still hire as you grow, but the hiring follows the exceptions instead of the volume, and the people you hire are doing work that needs their experience.

What to measure

Track two numbers each month: back-office hours per unit, and units handled per back-office person. If the automation is working, the first falls and the second rises, even as volume grows. Take the reading now, before anything changes, so you have something to compare against. How to measure AI ROI covers how to turn those readings into a figure the board will accept.

What to do this week

1. List the tasks in finance, operations and service that happen once per order, invoice, ticket or customer.
2. For the five biggest, ask the people who do them for minutes per unit and units per month.
3. Multiply them out at this year's volume and at next year's.
4. Split the largest into re-key, check, chase and decide, and mark the routine steps.
5. Before the next back-office hire is approved, check whether its work is on your list.

If you'd like help with the list, that's what a strategy engagement is for. We work through the per-unit tasks with you, size them, and tell you which step we'd automate first and why.

Get the next article when it's published.