Let’s look at recent trends in new products and how their offering might leave buyers disappointed after the novelty of AI wears off.
AI sells! What does that mean for you?
It’s official – AI is everywhere and the push for acceptance has begun. Google has AI-based responses at the top of every results page and Microsoft added their AI-powered personal assistant, known as Co-Pilot, as a new feature as part of Windows. With the addition of AI into digital products we use every day, the rate of acceptance by the public increases. These changes prove that AI is here to stay.
Automation for automation’s sake is alluring and easy to sell. It appeals to anyone desperate to tackle an immediate issue with a supposedly ‘easy’ solution to enable more work without increasing headcount. This is especially true in industries that are slow to adopt technology or are experiencing a spike in workload.
The importance of choosing the right solution will determine if the project will be a success. While that sounds straight-forward, it can be hard to ensure the options provided will meet a business’s needs.
To combat the deluge of snappy marketing and overzealous sales pitch filled with promises to revolutionize all manual tasks, this series will look at RPA tools with a focus on IDP solutions and workflows. We’ll look at the complexities of document automation and provide suggestions on what to consider when making a selection.
Document Processing is Standard in RPA Platforms
Document automation is neither new nor trendy, but demand is growing rapidly. The oldest and most basic part of technology, OCR, has been identified by software giants as a product they can develop and market to other software developers.
This development has benefited RPA platforms as an easy-to-integrate feature add-on. They have snapped it up, plugged it in, and now they have a new feature to market!
All-in-one automation platforms position themselves to be an affordable option that reaches the masses and targets customers who otherwise need to source a fit-for-purpose software developer (aka a value-added reseller) to implement process automation. With a SaaS platform, they work to integrate across countless software products to be a bridge and deliver data to continue a process. Add in other features like cloud hosting, continually expanding capabilities, and a subscription-based price point, it becomes an enticing option that allows the customer to keep the work in-house.
So, how does this pertain to IDP?
IDP for RPA – Cheap OCR masquerading as IDP
Note – If these acronyms are confusing, stop by our knowledge base for a quick summary.
Let’s use an analogy to get from OCR to IDP by comparing it to how a brain works.
Consider how we read. Reading isn’t a single process, but a series of intricate steps orchestrated by the brain. Images are captured by the eye and delivered to the brain. The brain will process the image to symbols to extract meaning. It will use this interpreted information to complete a task. It might also store the information for future use. Each step is unique, quick, but vital to complete a task. Additionally, the interpretation must be accurate for a human to comprehend what was seen on the page.
IDP is no different. The foundation of IDP is optical character recognition (OCR) which identifies letters, numbers, and symbols. OCR as a technology has been around for decades. And much like a brain, an intelligent document process goes beyond symbol recognition and combines them into words. Those words are then interpreted with context in which those words appear to ‘understand’ the information in a way that it can be processed. When needed, human intervention will resolve data captured below a set threshold.
AI as it’s known is not a single algorithm. Countless unique data systems, algorithms, and large language models, make up a layered process that work together to tackle complex tasks much like a brain.
The problem is—IDP is hard and costly to develop. But not OCR!
Numerous open-source OCR solutions are available, many enabled with a plug-in for fast integration. With minimal work, these providers can offer IDP as part of their RPA platform!
If the Wizard of Oz was placed in the future, Dorothy the user would find this IDP to be a technology version of the Tin Man. No brain.
OCR captures basic data.
IDP will identify and classify a document before it extracts, validates, and cross-references data. It’s likely you’ll want some, if not all these capabilities as part of your workflow.
How SaaS companies offering RPA solutions spin this shortcoming
So how is this missing component resolved? It is on the user to build a set of directions to accomplish their exact task.
Platforms can assist users to build processes to mimic IDP’s way of ‘thinking’ and bridge the gap in understanding.
While this process could be marketed as an upsell (more control!), it means the users are responsible for designing additional steps to build out their process. It sounds like a benefit when phrased properly, but the reality is it introduces rigid rules that will cause countless exceptions and increase human review. An increase in human review brings with it increased cost and additional time and the opportunity to introduce errors.
IDP requires robust context processing to ensure automation functionality in the long-term. If a workflow has multiple document types, users must build a potentially complex process since the embedded ‘IDP’ doesn’t provide this feature.
While this sounds easy in theory, it is an incredibly complex task.
Simply put—the user is on the hook to design a complex workflow step using rudimentary tools like if/then commands. Unless a user or team knows how to build detailed commands that understand the nuances of language, the document identification process will require continual maintenance to function.
In conclusion, the chances of building a detailed, yet flexible set of directions will be difficult and it risks reducing the rate of straight through processing. [Definition link here}
Structuring a Detailed Proof of Concept Can Mitigate Issues
A thorough POC will bring these concerns to light. Beyond data capture and accuracy, asking questions about how a solution classifies documents and data will provide insight into capabilities or limitations. All document automations projects require this capability. If the IDP cannot accomplish these tasks, consider the steps needed to accomplish this task. Additionally, factor in the impacts when changes must be made for future automation.
Asses the system for the following capabilities:
- Detail the capabilities of the IDP thoroughly. Does it go beyond character recognition? Does it interpret and identify information fields to build key value pairs? (link to definition here)
- Understand the system’s document identification for classifications. What information does the system use to identify a document type?
- Understand the rigidity of locating field and data. What happens when the location changes? How about the name of the data? (Ex – If a field on one document says Invoice but the vendor changes its to ‘bill of sale’, what happens to your process?)
- Does the IDP prioritize data? What about determining what data is valuable and what is unnecessary?
- What changes in documents will impact automation? Consider field location, field names, fonts, table modifications, and handwriting.
- How hard is it to modify and add to a process? What if additional information must be captured?
Conclusion
Sourcing the right RPA to achieve straight-through processing of documents requires a critical evaluation to achieve the automation required for success. The promise of control and ease of implementation might look enticing, but it is important to evaluate the IDP portion thoroughly, especially when IDP is only a portion of the capabilities offered.
Ultimately, the effort invested when choosing a solution will save a team from frustration and additional work. Solutions failing to meet expectations add unnecessary complexity to an already lengthy and arduous process.