In Malcolm Gladwell’s book “Outliers”, he popularizes the concept that to be great at anything requires at least 10,000 hours of practice. He argues, citing various studies, that the more research conducted into acknowledged high-performing individuals across many fields, the less innate talent mattered and the more critical was a near-ritualistic devotion to preparation and practice. Turns out there is a practical reason why there aren’t more violin virtuosos; there are few people with both the talent and the drive to actually put in the work required. And this is true for just about anything.
Regardless of the accuracy of the “10,000-hour” claim, the real principle at the center is that the more you put into something, the more you get out of it. So what does this have to do with IDP you ask? Well, for starters, when it comes to IDP, the only purpose is to automate, as much as possible, the manual tasks of working with documents, and to do it at accuracy levels better than a human. Oh, and one more thing: implementing such capability has to be easier and less risky than sticking with the status quo.
Does that sound like a simple process of acquiring software, making a few configuration selections, and BOOM you’ll have a highly optimized system ready to go? Nope. Odds are, if what you need to automate involves documents that are anything like what we’ve seen, you’re likely going to spend closer to 10,000 hours to get where you want or need to be. That’s like employing 7 staff for a little over half a year.
But there are few organizations that really want to do that (see: violin virtuosos) or can afford to do that. So it remains a problem.
That shiny Web-based IDP software with all of those pre-built skills isn’t going to get you there either. Sorry to burst your bubble.
The reality is that anything prebuilt also has significant prebuilt limitations.
One is that both the level of automation along with the accuracy are often well below what most organizations expect, leaving organizations to, yeah you guessed it: spend more time and effort to achieve what they need. Do you really want 99% accuracy if you can only achieve 5% automation levels? Er… no.
On the opposite end of the spectrum are the highly optimized prebuilt capabilities. Those too have limitations. We know because Parascript supplies optimized APIs to some of the world’s largest organizations. Mainly you get what you get – if you have the need to address a requirement not already built-in, you don’t really have an option.
So you’re either limited with needing to spend time on pre-built capabilities that are functional but not optimized, or you get high optimization but limited capability. That’s a hard pill to swallow for many organizations that really want or need to achieve high levels of reliable automation.
“What about AI and machine learning?”, you ask. “Won’t that make things a lot easier?”. Pssst. I think you forgot something. You need loads of curated data for that. Got any lying around? Probably not. That’s why you’re so interested in prebuilt capabilities remember? There’s got to be a better way. There is. Stay tuned.