How a manufacturing firm cut admin time by 40%

This is the story of a real project. The company is a UK-based manufacturing firm with 45 staff, making specialist components for the construction industry. We have changed the name and some identifying details, but the numbers, the timeline, and the lessons are all genuine.

When they came to us, they were not looking for AI. They were looking for help with a problem: their office team was drowning in admin, and it was starting to affect everything else.

The starting point

The company had grown steadily over the previous five years, from 25 staff to 45. Revenue had doubled. But the back office had not kept pace. They still had the same three-person admin team handling orders, invoicing, supplier management, and customer queries. The workload had scaled with the business, but the team had not.

The result was predictable. Orders took longer to process. Invoices were going out late. Customer queries sat unanswered for days. Mistakes were creeping in because people were rushing. And the admin team was exhausted.

The managing director told me something I hear a lot: "We're too busy to fix it, but we can't afford not to."

What we found

We spent a week mapping their processes, timing tasks, and talking to the team. Here is what the admin workload looked like:

Task	Hours per week	Who did it
Processing purchase orders	12	Office manager
Creating and sending invoices	8	Accounts assistant
Answering customer status queries	10	Office manager + receptionist
Supplier communication	6	Office manager
Data entry across systems	9	All three staff
Report compilation	5	Office manager
Total	50 hours/week

Fifty hours a week of admin across three people. That left almost no time for anything else, no process improvement, no proactive customer communication, no supporting the production team. They were stuck in a cycle of firefighting.

The Make UK annual survey consistently identifies administrative burden as one of the top productivity drags in UK manufacturing. This company was a textbook example.

What we changed

We did not automate everything at once. We picked the three highest-impact areas and tackled them in order.

Phase 1: Order processing (weeks 1 to 3)

Purchase orders arrived by email in various formats: PDFs, Word documents, sometimes just plain text in the email body. The office manager had to read each one, extract the relevant details (product codes, quantities, delivery dates, pricing), and enter them into their ERP system.

We built an AI system that reads incoming purchase orders, extracts the data regardless of format, cross-references it against the product catalogue and customer pricing agreements, and creates a draft order in the ERP. The office manager reviews and approves each one rather than building it from scratch.

Result: Order processing time dropped from 12 hours to 4 hours per week. Error rate fell from roughly 3% to under 0.5%.

Phase 2: Customer status queries (weeks 3 to 5)

Ten hours a week were being spent answering variations of the same question: "Where's my order?" The office manager would check the ERP, check with the production floor, and email or phone the customer back. Each query took 10 to 15 minutes.

We set up an AI assistant connected to their ERP that could handle status queries automatically via email. When a customer asked about an order, the AI checked the system, generated a clear status update, and sent it within minutes. Complex queries (complaints, changes, anything unusual) were still routed to a human.

Result: Customer query handling dropped from 10 hours to 3 hours per week. Average response time went from 4 hours to 8 minutes.

Phase 3: Invoice generation (weeks 5 to 7)

Invoicing was partly automated already through their accounting software, but there was still significant manual work: matching delivery notes to orders, checking quantities, handling partial deliveries, and chasing missing information. The accounts assistant spent most of Monday and Tuesday every week on invoicing.

We connected the AI system to their delivery records and accounting software so that invoices could be auto-generated when deliveries were confirmed. Discrepancies were flagged for human review rather than requiring someone to check every line of every invoice.

Result: Invoicing time dropped from 8 hours to 3 hours per week. Invoices now went out same-day instead of up to a week late, which improved cash flow noticeably.

The numbers

Here is the before-and-after picture:

Task	Before (hrs/week)	After (hrs/week)	Saving
Purchase orders	12	4	8 hours
Customer queries	10	3	7 hours
Invoicing	8	3	5 hours
Total	30	10	20 hours

Twenty hours per week saved across just three processes. That is a 40% reduction in total admin time (from 50 hours to 30 hours). At a loaded cost of £22 per hour, the saving is roughly £23,000 per year.

The total project cost was under £15,000, including our time, the technology setup, and three months of support. Payback period: around 8 months.

According to Deloitte's UK manufacturing outlook, manufacturers that invest in digital tools and process automation consistently outperform peers on both productivity and profitability.

What we learned

Every project teaches you something. Here are the lessons from this one.

Start where the pain is worst

We could have started with any of the six process areas. We picked order processing because it was causing the most visible problems: late orders, errors, and an overstretched office manager. Starting with the most painful problem meant the team saw benefits quickly, which built trust for the later phases.

Involve the team from day one

The admin team were nervous about AI. They worried it meant redundancies. We were upfront: the goal was to take the drudge work off their plates, not to replace them. We involved them in designing the workflows and testing the outputs. By the end, they were the biggest advocates for the project.

Imperfect automation beats perfect manual processes

The AI does not get every purchase order right first time. About 5% need manual correction. But that is still vastly better than manually processing 100% of orders with a 3% error rate. The Office for National Statistics has published data showing that even partial automation delivers significant productivity gains in manufacturing.

Cash flow effects are bigger than you expect

Getting invoices out on the same day instead of a week late had a surprisingly large effect on cash flow. The company estimated it brought forward roughly £40,000 in payments over the first quarter. That was not in the original business case, but it turned out to be one of the biggest benefits.

What happened next

Six months after the project, the company had not reduced its admin team. Instead, the three team members had taken on additional responsibilities: better customer relationship management, supplier negotiations, and supporting the production planning process. The managing director said the office "felt completely different," not because of new technology, but because his team was no longer constantly behind.

They have since asked us to look at two more areas: production scheduling and quality reporting. Both are still heavily manual, and both follow the same pattern: skilled people spending too much time on data handling and not enough on the work that actually needs their expertise.

If you run a manufacturing firm in the north of England or Scotland, the admin-reduction work described above is the most-asked-for piece of work we do. Sector-specific guides cover Manchester, Leeds, Sheffield, Bradford, Newcastle and Glasgow.

Could this work for your business?

If your admin team is stretched, if orders or invoices are going out late, if your people are spending their days on data entry instead of customer-facing work, the same approach will likely work for you. It does not matter whether you are in manufacturing, distribution, services, or retail. The pattern is the same: identify the biggest time sinks, automate the mechanical parts, and free your team to do the work that matters.

Our free AI opportunity report will show you where the biggest savings are in your specific business. It takes five minutes to request and gives you a clear, practical assessment.

Get your free AI opportunity report here