Using automation to tap unknown and undocumented data
Many organizations are falling behind on automation, while others have proven to be leaders. The difference? The leaders weren’t given a choice.
In many cases, these organizations have grown so quickly that they can no longer successfully operate without automation, specifically Intelligent Automation with an eye toward scaling up to artificial intelligence.
The Three Process Categories
All enterprise processes fall into one of three categories: known and documented; known and undocumented; or unknown and undocumented. Automation follows a similar pattern.
Most automation we see in the workplace now addresses specific business processes that themselves are known and documented. Even the organizations that lag have adopted many of these solutions throughout parts of the enterprise. These processes, like credit card payments and other methods of money transfer, are very much the low-hanging fruit and already in a mature phase of automation.
Organizations in the next phase of automation are tackling the known and undocumented processes, mostly via technology like robotics process automation.
For example, an employee going on vacation can typically setting up a freeze on an expense management tool, a straightforward process that creates an "if/then/except" rule based on incoming requests.
But assume for a moment the request is more complex than a simple "freeze" or "no freeze" rule can handle. Say this vacationing user needs to access a different expense management tool, log in with a user name and password, and access and approve a transaction.
As an extra challenge, assume the transaction amount is above a certain threshold, requiring the user to log into a spreadsheet program to determine the next process and then fire off an email to the auditing department to get final approval.
One can quickly see that this is a very complex, multi-application process that could and possibly should be automated. Despite the complexities, however, every step is known and can be documented. However, these steps haven't been documented as of yet.
RPA is very well on the way toward being able to accomplish this within certain organizations. What RPA does is give those organizations the ability to document these components and create a software robot that can automatically manage all steps, so that which is known but undocumented can move into the category of known and documented.
Unknown and Undocumented
While some organizations have been slow to tackle known and undocumented processes, the real opportunity lies in the unknown and undocumented processes. This is where AI will truly make its mark. These are the processes for which we assume human input is needed.
For example, when a company's collections department receives an invoice but something looks confusing or incorrect, a set of rules does not exist for "confusing" bills. The collections agent needs to make a knowledge-enabled decision based on experience and perhaps a series of investigations.
This is where the opportunity lies. At some point in the future, AI will be able to determine quickly which invoices are most likely to disputed. This type of complex decision-making, based on pattern recognition, is something we all do intuitively thanks to our memories.
But there are ways to automate even these most sophisticated processes. These more complex, multi-application, high-density data environments can be addressed through AI. Getting it right involves leveraging the right people, while taking a holistic view of the entire organization.
How It Will Look
Finding individuals with specific and unique skillsets is crucial. The most important skills include a core understanding of the undocumented processes that need to be automated and a fundamental knowledge of how best to document them.
The reason for this is twofold: Organizations run a real risk of falling even further behind if they don’t get their automation journey right in the first place. All enterprises need to understand that the process of documenting itself isn’t easy.
Automating bad processes — which companies often do when they prioritize getting it done fast over getting it done right — creates an issue in which they must go back and reengineer poor engineering.