AI-powered document classification at PICA Group
PICA Group sought to improve the efficiency and accuracy of its document processing workflows by exploring AI-driven solutions. Acquiring and manually managing a high volume of critical records and invoices was slow, labour-intensive and impacted customer onboarding timelines. Cevo collaborated with PICA Group to evaluate multiple AI-driven solutions, conducting proof-of-concept trials to identify the most effective and cost-efficient approach. The project delivered a clear AI adoption roadmap, demonstrating the potential for scalable automation, improved classification accuracy, and accelerated business processes.
AI/ML proof-of-concept
Solution
93%+ accuracy
Across all solutions
Real Estate
Industry
PICA Group is a leading Australian strata management and property services company that oversees strata properties, body corporates and owners’ corporations, managing over 187,000 lots across residential, commercial, resort and mixed-use properties. With a workforce of approximately 700 employees and a network of 30 branches, PICA Group offers a comprehensive range of services, including strata management, facilities management, receivables management and property developer services.
Business challenge
PICA Group faced growing challenges in managing and processing documents at scale. As a leading strata management company, a significant part of their operations involves handling high volumes of invoices and other critical records, often in unstructured PDF formats, where efficient classification is essential to maintain smooth business processes. However, their existing approach relied heavily on manual effort, creating inefficiencies and bottlenecks.
A key challenge in this industry is that acquiring new business often means inheriting large volumes of paper-based or poorly scanned digital records. While PICA Group digitises these records through a third-party provider, the lack of structured automation requires the team to spend approximately eight hours per week manually classify thousands of documents, impacting efficiency and accuracy.
These inefficiencies had a direct business impact. Customer onboarding through PICA Group’s online portal could take up to six weeks due to document processing delays, negatively impacting the client experience. Similarly, slow invoicing led to cash flow issues with no way to accelerate timelines.
“This is both a customer service and operational issue. For clients, a seamless onboarding experience builds trust. Internally, relying on a team to manually review thousands of documents is costly and time-consuming. This isn’t a one-time issue; it grows as we scale. The more business we win, the more complex and resource-intensive this process becomes. AI presents an opportunity to help automate document classification, improving both efficiency and accuracy.”
Kitty Henning
Executive General Manager – Technology, PICA Group
Given PICA Group’s strategic focus on AI but limited in-house expertise, AWS recommended Cevo as a trusted partner. Addressing this challenge wasn’t just about improving efficiency gains, but was essential for building scalable, future-proof operations as PICA Group continues to expand its market presence.
Solution
Over six weeks, Cevo worked closely with PICA Group’s technology team and subject matter experts to assess AI-enabled solutions. The project commenced by analysing PICA Group’s existing document classification process utilising a jobs to be done framework, uncovering potential areas for automation. Cevo then explored three proof-of-concept (PoC) trials over three weeks, where each trial was tested against a limited set of document types.
To enhance efficiency, documents were first processed through Amazon Textract before classification, enabling faster and more accurate analysis.
Cevo then tested three classification alternatives:
- Amazon Comprehend: An AWS managed service for natural language processing (NLP) including custom text classification.
- Generative AI (Amazon Bedrock): An AWS managed service that allows you to build Generative AI applications by using various Large Language Models without managing infrastructure.
- Machine Learning Classifier (Vector Databases): A custom machine learning model that is trained to distinguish one document from another.
Evaluating performance and cost-effectiveness
Cevo assessed each solution based on four key metrics: latency, cost, accuracy, and complexity. The goal was to identify an approach that minimised processing delays, kept costs manageable, maintained high accuracy, and integrated seamlessly with existing systems.
“We set out to show how AWS AI services could help PICA Group scale customer onboarding. By understanding their document workflows, we were able to effectively discover the best alternatives for them. We explored several potential solutions, balancing accuracy, cost, latency, and complexity, and worked closely with PICA Group to prioritise what mattered most. Their domain expertise made
collaboration seamless, and it was a pleasure working with the team.”
JO Reyes
AI/ML Technical Specialist, Cevo
After a structured evaluation and discussions with PICA Group stakeholders, the machine learning model was selected as the preferred solution for its balance of affordability, efficiency and ease of use. The Bedrock solution was also considered, but the selected model provided the best fit for PICA’s team and their requirements. Cevo then built and deployed a pilot version within PICA Group’s AWS environment, laying the groundwork for broader testing and integration.
“What stood out was Cevo’s detailed comparison of the three approaches, making it easy to assess each solution’s alignment with our goals. It was clear that traditional programming methods wouldn’t have worked, and it was reassuring to see that in the results. Cevo’s demonstrations gave us confidence that AI was the right path forward.”
Kitty Henning
Executive General Manager – Technology, PICA Group
Collaboration and knowledge sharing for seamless integration
Transparency and knowledge transfer were key to the project’s success. Collaboration with PICA Group was highly effective, thanks to the team’s strong technical expertise, enabling efficient decision-making. Weekly demonstrations allowed PICA Group to see each alternative solution in action, using identical input dataset for direct comparison.
“Working with Cevo felt seamless from start to finish. They quickly understood our needs, provided valuable insights and adapted to our evolving needs. Their collaborative approach ensured we were always aligned and confident in our direction.”
Kitty Henning
Executive General Manager – Technology, PICA Group
Outcomes
The successful PoCs demonstrated the potential of AI to tackle a long-standing challenge in time-consuming document classification.
Key outcomes include:
- Proved the feasibility of AI-driven document classification, establishing a scalable framework for further exploration.
- Provided a data-driven comparison of AI solutions, equipping PICA Group with clear insights to support decision-making.
- Explored multiple AI pathways, giving PICA Group confidence in its next steps.
- Ensured high reliability and over 93% accuracy in all solutions, improving classification consistency and reducing errors.
Additionally, the PoCs demonstrated AI’s ability to:
- Accelerate customer onboarding with significantly faster document processing.
- Automate document classification, freeing staff to focus higher-value tasks.
- Increase processing speed and accuracy, significantly reducing document handling time.
- Enhance operational efficiency, minimising manual intervention and streamlining workflows.
- Scale effectively, enabling faster clients onboarding and an improved customer experience.
“The goal was never to replace our team with AI, but to determine which solution would best enhance our capabilities – allowing us to work faster and more effectively. This project has given us the knowledge to make more informed comparisons and strategically apply AI-driven solutions.”
Kitty Henning
Executive General Manager – Technology, PICA Group
With the pilot now live to PICA Group’s AWS environment, the team is expanding testing across more document types. They are also exploring integration with downstream systems, such as their content management system to streamline document management. These next steps will help refine the solution, address skill gaps, and build a strong business case for scaling AI adoption.
Enjoyed this customer story?
Share it with your network!
You may also like
