The importance of data within automation

Here are some of the comments from the roundtables on “The importance of data within automation” at the Financial Services Summit Singapore on April 11, 2019.

1) What difference can a data-first approach make?

• Unfortunately, when it comes to data, the old aphorism of “garbage in – garbage out” still holds true. Valuable insights that can lead to actionable decisions and innovation depend on high-quality data.
• Rather than replicating a process carried out by a human, focus on the data flows so you can optimise the process. Don't look to just throw a bot at it and automate the process as it is, instead use a data-centric lens to challenge and overhaul it.
• By taking a data-first approach you can build in flexibility and agility, enabling your organisation to plan for future changes in data sources, reporting requirements, asset classes, etc

2) What are the root causes of low automation?

• In most cases, inbound data isn't fit-for-purpose.
• Digitising content is key but it’s not the end of the story. It’s not enough to turn a physical document into a digital file – you have to do the hard work of extracting relevant information automatically otherwise the benefit of digitisation is limited.
• Data fragmentation is caused by a variety of formats coming into your organisation (Emails, PDFs, Faxes, scanned images, Excel, Word, XML, SWIFT, …). It’s also caused by an increasing number of channels used to capture the information (emails, fax, websites/portals, shared folders, APIs, Data Lakes, Social Media,…). In addition, you may need to handle a multiplicity of master sources or “golden copies” aka “single versions of the truth”. It is crucial to automate the capture and normalisation of the data needed for a particular process.
• Data fragmentation is the root cause of data exceptions. Resolving exceptions is time consuming and tends to be manual. If data fragmentation is the root cause of data exceptions, data exceptions are the root cause of low automation.

3) What does AI-enabled automation demand of data?

• Quality data first and foremost, otherwise you are building and training models on bad data.
• A lot of AI solutions require a lot of data and a lot of time to train the model - this needs to be a key consideration when looking at AI.
• Involving business people who understand the data and the process is crucial if you want to leverage AI to its full extent. Users can use platforms such as Xceptor to manage and train the models within the same environment where they carry out their daily tasks. This in turn increases the rate of adoption. Leveraging AI should not be just another technical exercise.