Why records management needs an update with AI
Tauhid Jalil, Solutions Manager
Ever-expanding enterprise records means increasing risks and inefficiencies to manage. How can you improve your records management system for better records searchability, retention and disposition?
Records management systems play an important role in high-value operations from finance reconciliation, compliance operations to document management. These records act as corporate footprints–pieces of evidence of an organisation’s activities–and hold vital legal, fiscal, evidential and administrative value. Because records are everywhere, organisations are dependent on effective records management to ensure everyone in the organisation has access to the required knowledge at the right time.
Poorly managed records comes with a hefty price tag
As a company grows, so does its business activity, and the records that come with it. With this increased volume of data, comes complication and issues like information misclassification, data overloading, lack of transparency, security lapse amongst others, that create a host of problems from inaccurate access to compliance risk. In the US alone, ransomware access to health records doubled in 2020, resulting in US$20 billion worth of downtime according to HIPAA Journal (Health Insurance Portability Accountability Act).
Thankfully, technology has advanced considerably to deal with these risks and inefficiencies. Some records management systems are already starting to integrate smarter tools like artificial intelligence (AI) and workflow automation. What are the ‘must haves’ for a secure and effective records management system today?
Effective records management starts with classification
Organising document, text, image and metadata is the first step to enhance content discoverability. When records are classified well, users are better able to find them easily, which then makes it easy to use and share them. Information managers can retain and dispose records according to organisational or legal requirements to minimise both risk and needless data storage related costs.
However, this process of classification is typically mechanical, traditionally requiring a user with domain knowledge to manually add identifiers to ‘tag’ the data. When this is done across hundreds and thousands of records in dynamic and growing digital environments, organisations would very likely be exposed to unnecessary risks and spends.
This is where ontology-based AI makes classification more automated
Today, AI can be trained to automatically tag the database of records, using pre-set rules like time, date, duration, author or other qualifiers that the organisation values. These rules are conveyed to the records management system through machine-readable data models, called ontologies.
An ontology formally specifies how a domain is conceptualised, by explicitly defining each data entity within the domain, along with the relationships (and relationship constraints) between each of them. This forms a giant knowledge map that gives a more nuanced view of how the information within the domain exists.
Just like how a human can easily identify that ‘John Doe’ is a client who works for a company called TAIGER, an ontology helps a machine understand these relations, properties and categories with similar granularity.
The machine, or records management system, would then be able to tag records within enterprise repositories, stripping away a process that is typically manual and tedious. Such bulk tagging is often a favoured technique used in managing high volume of records, such as artefacts in a museum, or collections or archives in a military context.
If required, users can also manually edit tags to validate tagging accuracy or expand the ontology. This is a ‘DIY’ process that allows business users with minimal technical knowledge to iteratively refine the ontology via an admin interface.
Connecting data meaningfully similarly aids records discoverability
As we can see, organised records are the prerequisite to accurate data extraction, search results and content discovery. When data is connected meaningfully through ontologies, they provide a layer of context that can dramatically improve information management through the entire lifecycle.
Because ontologies clearly define how different entities and concepts are related, they help users discover information that is associated with the topic without having to depend on the user’s initial input or search refinement. Referring back to the figure above, searching for records on a consultant, Carl Chang, would also bring about search results on the Enterprise Search project he is working on, and ABC company which is whom he is working for. This connected view of associated records helps to quickly reveal what records to retain or dispose of, in a very comprehensive way.
Records management systems must also be able to protect data access
Besides classifying and connecting data entities, giving and restricting access to records is equally important especially in current times where cyber attacks and fraudulent transactions make front page news often.
Providing security tags to records in the form of a metadata field can define security related actions allowed to be performed on that record. Security tags can permit or restrict access for specific users, set a timing for records disposition, on top of improving organisation and extraction. Secured access functionalities like SAML and 2FA are also increasingly adopted as the norm to ensure access is only for authorised users.
Other security considerations also include deploying a ‘process and purge’ method where documents or metadata are scheduled to be cleared.