Automated processing and registration of incoming correspondence using oMail
The incoming mail function, handled by the Information Services team of a Victorian Council, was overwhelmed by the increase in volumes of correspondence in a context of diminishing numbers of staff and increased volumes of electronic mail. The Council partnered with our organisation to automate the registration, metadata extraction, and workflow management of incoming physical and electronic mail.
The solution took a couple of months to implement and uses the Councils’ existing document management system to validate the data extracted by our AI and machine learning engine. The results include the automation of over 170 incoming mail types, their individual business rules and workflows.
A range of benefits was achieved: the standardisation of mail naming conventions to facilitate searches throughout the Council’s systems:
- The need for only one person half the time to do the work of previously two and a half staff
- A steep increase in the quality of the metadata provided, and
- A fast turn-around for mail processing inside the Council.
In 2017, the Council engaged with us after an onsite review of the current “as is” mail document processing practices. The Council was looking for automation opportunities and efficiency gains in an environment facing increased volumes of mail, with no options to hire additional staff.
A report was provided which detailed the current steps and highlighted the changes needed as well as the benefits (financial and operational) which could be gained by automating the processes.
It took 2.5 staff 6 hours to process 170 daily mail items, leaving them unable to undertake their other duties until after the mail had been processed. Additionally, the Council was engaged in a number of other projects including migration of systems, merging of author’s databases, increasing the number of online services deployments which needed the Information Services team’s availability to ensure governance, risk, and quality, as well as speed, was not compromised.
The proposed solution includes changes made to the manual processing and scanning steps in order to remove as many as possible but still retain the quality assurance required of the department. Once physical mail is digitised, the technical solution, oMail (our inbound mail processing software), automates the classification, registration, and processing workflows of all incoming document types.
The solution processes all mail items through different stages. An intelligent OCR and image recognition process, which does not require the setup of templates, identifies and extracts the metadata from the document.
All extracted metadata is validated against the council’s internal and trusted data for the author’s identity in HPE Records Manager. A set of agreed naming conventions and business rules, learned by oMail using supervised and unsupervised learning, are applied, and the documents are automatically registered into the document management system with all the verified metadata extracted. Workflows are also triggered as part of the solution.
oMail was selected because there was no other solution that could provide a holistic approach to incoming mail automation on the market and there were a number of reference sites across Australia and New Zealand which could attest to the performance of the solution and the dedication and professionalism of the team.
oMail was set up as a pilot project initially but was adopted as a solution after the pilot returned successful results and demonstrated the possibility to return higher-quality work in a shorter timeline with fewer resources.
This means that staff was able to focus a much greater proportion of their time delivering high-value projects to the Council rather than lower value manual tasks.
- The work originally done by 2.5 staff over 6 hours each, is now done by one staff in less than 2 hours.
- The cost reduction per item processed is lowered to less than half the original costs including the processing service cost and the reduction in FTE.
- Data that was not previously included in the dataset for each mail item is now included and automatically inserted as part of the registration process. This represents a significant improvement in the quality and consistency of results returned through searches in the document management system.
- Common mistakes due to manual data entry have been removed reducing errors, typos, and duplication of information (for example when the name of an author is misspelled and therefore the view of the correspondence from this author is incomplete).
- The number of steps to scan and process the mail was reduced significantly because the software does not require sorting and using separator pages. It does not require setting up templates which can change over time and need constant adjustments.
- Staff manage by exceptions only and focus their time on the very small proportion of correspondence that does not fit into a business rule or where the metadata is not clearly verified. This means that staff can utilise their intricate knowledge of the business rather than spending time on tasks that can be fully automated and do not require human intervention.
Artificial intelligence-powered solutions are new to most organisations and the implementation steps, as well as the technology potential benefits, are very much an unknown. At this Council, as with most of our clients, we used an iterative approach with only a few document types implemented through a no-fuss pilot process which allows the team to truly evaluate in-situ the actual capabilities of the solution.
The scope of the pilot is based on an “as is/ to be” process which clearly evaluated the current processes and proposed a new process quantifying the estimated benefits for the solution prior to making a final decision. The pilot measured image quality (clarity, missed pages) and metadata extraction quality (application of naming conventions and business rules, quality of classification and registration, appropriate assignment of workflow, extraction of critical dates, and author information).
Once the results were demonstrated successfully in only 2-3 weeks, the decision to go ahead and implement all document types was taken and the full project implemented within a couple of months. The approach has been to focus the delivery on high volume document types to start with and to continue the training of document types less frequently used over a few months giving the team space and time to adjust their processes and learn the new way of working.
This approach has also helped remove the fear staff expressed of losing their jobs and to allow for trust to develop for the automation processes. Lately, the Council has increased the scope of the automation and is looking at using oMail for the identification, classification, and workflows of customer queries.