Understand what your business needs before committing to the solution.

The problem

A large retail company was already using an automated AP invoice processing solution.

However, despite the automation, a significant number of PO-based invoices were still failing to match to the correct Purchase Order and were being sent to a queue for manual review.

With a high volume of invoices being processed, this was creating unnecessary workload for the AP team.

What I did

I reviewed the existing invoice allocation and matching rules from the ground up and spoke directly with the staff responsible for reviewing the exceptions.

The key question was:

What are employees checking manually that the system is not checking automatically?

By comparing the existing system logic with the practical decision-making process used by the AP team, I identified an additional allocation rule that could be automated.

The core matching logic was revised and the new rule was implemented into production.

Business impact

Within the first week, approximately 1,000 fewer invoices required manual review.

This allowed the AP team to focus on invoices where human intervention was genuinely required — for example, where a Purchase Order had not been receipted, quantities did not match, or another genuine PO exception existed.

Manual effort was redirected to exceptions that actually required investigation.

The problem

A large retail company had implemented a code-learning feature designed to automatically determine how non-PO invoices should be coded.

Over time, the learning model became unreliable.

Because it was learning from individual employee behaviour, similar invoices could receive conflicting coding depending on how different staff members had processed them previously.

In approximately half of the reviewed cases, the suggested coding was incorrect.

What I did

I completed a detailed review of the existing learning logic, historical behaviour and invoice allocation process.

Following the assessment, I recommended switching off the unreliable learning functionality and replacing it with transparent, rules-based auto-coding.

For example, the system could use information already available on an invoice — such as the Deliver To address — to determine which business site the cost should be allocated to.

Where an invoice clearly identified multiple sites, the automation could also create multiple accounting lines from a single invoice line and split the cost automatically — for example, 50/50 across two locations.

Business impact

The new approach reduced coding errors by approximately 60%.

Invoices relating to a single site could be coded automatically without staff intervention.

Manual review remained where the supplier invoice did not contain enough information for the system to make a reliable decision.

The objective was not to automate every invoice at any cost.

It was to automate decisions that could be made safely and consistently while keeping genuine exceptions with people.

The problem

A business providing third-party services knew that it was losing money, but the source of the problem was unclear.

There was no single process that could immediately be identified as the cause, so an end-to-end review of the business processes was required.

What I did

I reviewed the operational processes, reporting and information available to management.

During the investigation, I identified a major visibility gap.

The existing reporting was primarily showing completed orders.

It did not provide sufficient visibility across the full population of orders still moving through the operational process.

This meant the business did not have the complete picture needed to understand where losses were occurring.

Once visibility across all relevant orders was established, the underlying issue became clear.

Based on the findings, an appropriate business action and policy change were recommended.

Business impact

The business implemented the recommended change.

A few months later, the company recorded a profitable month.

The problem

A business had multiple reports available, but the numbers were not consistently trusted.

Different views of the same operational activity did not always align, and management could not confidently rely on the reporting when making decisions.

The issue was not the absence of data.

The business already had information available, but it was not being presented in a way that created a reliable view of what was actually happening.

What I did

I reviewed the reporting logic, the underlying business processes and the way operational data was being captured and interpreted.

The objective was to understand why different reports were producing different views and which information management actually needed in order to make decisions.

The reporting logic and exception visibility were then revised so that the business could see the relevant activity more clearly and identify where further investigation was required.

Business impact

The business gained a more reliable view of its operations and could spend less time questioning the numbers and more time acting on them.

The value of the work was not in creating more reports.

It was in making the existing information more useful, consistent and actionable.

The problem

A business had already invested in a third-party software solution, but the system was not being used effectively.

Important company settings had not been configured correctly.

Staff were not confident in how the system should be used, and trust in the software had gradually declined.

As a result, employees relied on workarounds and manual processes instead of using the system as intended.

What I did

I reviewed how the software was being used in practice and identified the gap between the intended process and the way staff were actually working.

I then worked directly with the third-party software provider to resolve the configuration issues and clarify the correct system setup.

Once the technical issues were addressed, I created a practical user guide based on the real business process and organised training for staff.

The focus was not just on explaining which buttons to click, but on helping users understand how the system supported their day-to-day work.

Business impact

Staff confidence in the software improved significantly and reliance on manual workarounds decreased.

With the correct configuration, clear guidance and better understanding of the system, the team was able to work more efficiently and make better use of a solution the business had already paid for.

The business did not need another system. It needed the existing one to work properly.