CHP was Highly Commended in the Innovation category at the Housing Technology Awards 2025. Jenny Wilks, head of digital delivery, and Sam Hardy, graduate business analyst, from CHP explain how the housing provider has overhauled its manual invoice processing using robotic process automation.
We’re always looking for ways to improve, including adopting new digital solutions to enhance our services and enable our team to help customers more efficiently. One example of this is implementing RPA for invoice processing in our supply chain.
The challenge
One of the key responsibilities for our supply-chain team is processing invoices from suppliers. The team faces a significant administrative burden, dealing with around 15,000 invoices each year. Each of these invoices has to be manually checked, validated and uploaded through a process of almost 50 steps from start to finish; depending on the size of the invoice, this could take 5-15 minutes per invoice to do.
The process is prone to errors (due in part to the volume of invoices) and often requires one employee to be allocated each day to meet business demand. The process needs a person to review the inbox to identify which documents are invoices, credit notes and statements, before extracting the relevant data from the invoices and creating these against purchase orders in our system, MRI OneContractor. Furthermore, each supplier has its own way of sending invoices, making it challenging for our team. This can result in a range of problems including:
- Missed invoices;
- Suppliers putting us on ‘stop’ until payment is received;
- Newer invoices often being processed before older ones;
- The invoice inbox often exceeding what can be processed in regular working hours, meaning overtime is required.
Solution design
We considered EDI (electronic document integration) to automate purchase orders into MRI OneContractor but a proof-of-concept revealed its limitations. For example, it didn’t align with our internal financial regulation processes and only had a 50 per cent success rate. Furthermore, some suppliers don’t use EDI, preventing its applicability and scalability to all suppliers.
We instead decided to investigate a potential solution using a combination of RPA and artificial intelligence (AI).
Our goal was to read the invoice PDFs and put the extracted data into a work queue which could be checked by our supply-chain team (where required). Once all the details were checked, the RPA bot would find the corresponding purchase order in MRI OneContractor and create the invoice against it. This allows any discrepancies in stock codes and quantities to be identified and trigger an error message, highlighting that an invoice requires manual intervention.
The development challenges fell into four key areas:
- Speed of delivery;
- Integrations with multiple systems;
- Gathering data from unstructured documents;
- Giving our supply-chain team control over the decision-making process.
We decided to combine Toca.io’s low-code RPA platform and Abbyy’s AI power invoice platform to completely overhaul the current process. The AI power invoice platform reads and extracts data from complex, unstructured PDFs and the Toca platform orchestrates the process by uploading the invoice data to MRI OneContractor using the RPA bot.
Implementation
Our first challenge was recognising the different types of documents attached in supplier emails. The RPA bot integrates with a Graph API tool, which allows access to the supply-chain inbox. The bot reviews each email to identify the type of document so they can each be handled in a specific way. This classification uses machine visioning to read and understand the email and key text within the documents. Once completed, the documents are sorted into three categories: invoices, delivery notes and statements.
The invoices are then sent to the AI tool, which analyses and gathers the required data using optical character resolution (OCR) technology. This classifies invoices by supplier and then AI is used to collect the following information: PO number, invoice number, invoice date, net amount, total VAT, gross amount, and stock codes, descriptions, quantities, item prices, total values and VAT rates for each line item.
To ensure high accuracy of the data collected, the AI gives a confidence score on each of the data points. If the AI isn’t confident that it has gathered the data correctly, it’s sent to a ‘validation tasks’ work queue where the supply-chain team can check the details against the corresponding invoice PDF and manually correct the errors.
If the AI gathers the data to a high confidence score and all the data points are read correctly, the invoices are sent to a ‘processed documents’ work queue where they can then be uploaded.
Because MRI OneContractor doesn’t have a fully-accessible API, the supply-chain team uploads the invoices using the RPA bot. The bot logs into MRI OneContractor to process the invoices. The corresponding PO number is found from the extracted data and the product lines on the invoices are matched as goods received against this using stock codes. Specific exceptions are built in here to mark purchase orders as part-delivered if there are outstanding items or to trigger error messages in the Toca app where uploads have failed and require manual intervention. If the invoices match the purchase order correctly, the automation runs fully and creates the invoice using the extracted information.
“Using the RPA tool is an exciting step forward to bring more automation into a team consumed by time-consuming manual processes. By saving these valuable hours, we can focus on contract management, working with our suppliers and providing a better service for our stakeholders.”
Sam Wright, Planning & Logistics Performance Manager, CHP
Business impact
The overhaul of this process has led to a huge time saving of up to 10 minutes per invoice and our supply-chain team now only need to manually correct ‘errored’ invoices. As we increase the number of suppliers we work with, we estimate annual resource efficiencies of over 2,000 hours or at least one full-time employee. Additionally, it has:
- Created an accessible and reusable platform to resolve invoice queries efficiently;
- Ensured manual interventions are kept to a minimum;
- Maximised visibility of outstanding invoices requiring individual attention;
- Significantly reduced delays in payments and fewer email follow-ups;
- Improved performance monitoring and reporting.
As a result of the benefits seen by our supply-chain team, we’re currently in the process of automating our estates management invoices, using the same automation to significantly reduce the administrative burden in another high-volume business area.
Jenny Wilks is the head of digital delivery and Sam Hardy is a graduate business analyst at CHP. The housing provider was Highly Commended in the Innovation category at the Housing Technology Awards 2025.
