RPA – Should I Be Using Cognitive Automation?

There is a lot of hype in the RPA world around cognitive automation. This leads to many questions from those starting their RPA journey – the primary one being, “should I be using this?”.

I’ll begin by providing a brief summary of what I’ll call “process automation” in RPA, perhaps better described as “traditional” RPA, key characteristics of which include:

  • Repetitive, structured and logical tasks
  • Structured data
  • High stability and consistency, low variability
  • Simple rules based, low complexity
  • High volumes
  • Digital inputs, potentially via basic OCR

Essentially, process automation brings speed, accuracy and precision to time consuming, repetitive, “mundane” tasks that typically don’t require the human user’s specialised expertise and decision making.

Typical use case examples for process automation include:

  • Finance: invoice and expense processing
  • HR: joiners, movers, leavers
  • Technology: data migration, data cleansing, password unlock and reset
  • Contact centres: one customer view, updating customer profiles, customer complaints
  • Operations: customer onboarding, account opening, renewals, account closing, change of address

“Cognitive automation” refers to the use of elements of artificial intelligence (AI) to enhance process RPA, allowing the tools to make “judgements” and increasing the ability to automate business processes steps that require specialised expertise and decision making. Characteristics of cognitive automation include:

  • Repetitive, manual, situational and decision oriented tasks
  • Unstructured data
  • Decisions made by Bots or escalated for human input (and further training)

When we refer to AI in the context of cognitive automation, we are essentially referring to natural language processing (NLP), machine learning (ML) and intelligent optical character recognition (OCR). These three elements allow RPA tools to understand data to extract actionable information. Whilst it will never replace the need for human input to resolve more complex conditions, it allows for the scope of RPA in organisations to take a few steps further.

Typical use case examples for cognitive automation include:

  • Mismatches between contracts and invoices
  • Chatbots
  • Banking: customer screening
  • Insurance: assess the impact of policy changes
  • Insurance: make automated claims decisions

So, should you be considering cognitive automation in your RPA journey? The question should perhaps be better asked as “when should you be considering cognitive automaton in your RPA journey?”.

In answer to this, there is no doubt that the ultimate success of RPA in many organisations is based on early stage successes and confidence, providing the levels of adoption, interest (crucially across both business and IT stakeholders) and budgets that require further investment at each stage. Indeed, there are many organisations that have begun their RPA journey with limited success and often significant cost which means that adoption is low, funding for ongoing maintenance and support isn’t forthcoming, and confidence investing further quickly erodes. RPA therefore oftens ends up “on the shelf”.

Don’t Run Before You Can Walk

Whilst there are many reasons as to why this situation can occur, the phrase “don’t run before you can walk” is appropriate in the context of cognitive automation. I’ve yet to work with an organisation that doesn’t have any clear, “quick win”, straightforward opportunities for process based automation, which if implemented well and with appropriate stakeholder buy-in, can have very significant cost, speed and efficiency benefits. This in turn provides the business case and confidence to move to the next stages and levels of adoption, during which cognitive automation will become increasingly relevant.

One final point – the hype and “happy path” sell around cognitive automation can unfortunately often inspire too much initial confidence and scope in some organisation’s RPA pilot projects. More often than not, this is not down to failures in the cognitive technology itself, rather the failure of organisations to fully appreciate the complexities and risks of implementing it.