A few months ago, Cognilytica released a report on the market identified as the Intelligent Process Automation (IPA) market. In this report, we discussed the evolution of business process automation, modeling, management, and optimization, and some of the tentative steps vendors were making from non-cognitive, Robotic Process Automation (RPA) to what we called cognitive Intelligent Process Automation. Our insight in this report was that automating existing tasks with unintelligent robots still leaves way too many tasks that are dependent on humans to execute, identify, model, and manage. These bots are simply mechanizing the assembly line to take what humans repeatedly do and automate that so machines can then do them repeatedly. However, fundamentally, the processes are the same. Bots get stuck and humans have to help out. And these bots are not intelligent – they can’t handle changes to process flows, data, logic, or other issues.
In our IPA report, we identified four categories of Intelligent Process Automation, with increasing levels of cognitive ability:
|Level 0: Enhanced RPA (not AI)||Level 1: Language & Context Aware||Level 2: Intelligent Process Awareness||Level 3: Autonomous Process Optimization|
|Source: Cognilytica – Intelligent Process Automation Report (http://www.cognilytica.com)|
In some ways, this classification borrows from the autonomous vehicle and train industry. In those industries, level 0 represents the unintelligent state of technology, with increasing levels of autonomy requiring increasingly greater levels of cognitive capabilities and providing increasingly greater value to the human users. In the same way, we see moving up the ladder of cognitive ability of business process resulting in increasingly greater value to business organizations by tackling increasingly harder business problems of increasingly more strategic value.
First, Let’s get this out of the Way: RPA Is not AI
While automation is definitely part of the goals of artificial intelligence, and in particular automating things that require human cognitive capabilities, simply automating things doesn’t make them intelligent. Have a good conversation with your crock pot lately? We’ve discussed and written about this concept that automation is not intelligence quite thoroughly in our previous writings, so we won’t beat this horse again.
Specifically, we continue to be dismayed that the vendors selling their wares in the RPA market category continue to claim that they have AI capabilities, even though their products don’t seem to provide much evidence of that. Most of the leading vendors have added unstructured text, image, and in some cases, audio processing. These Natural Language Processing (NLP) capabilities are “table stakes” in the intelligence game. If your automation tool can’t process handwritten text or generate transcripts from audio, then you should get rid of that tool immediately. The capabilities for processing unstructured information are widely available from dozens of vendors. So what. Big deal. Unstructured information processing is barely Level 1 autonomous process capability.
The big idea of AI and cognitive technologies is to address and tackle problems that require human cognitive capabilities. Many vendors in the RPA market are simply using AI as a marketing vehicle and are not fundamentally changing or improving the way business process is done. This is why companies looking to pursue the path of Intelligent Process Automation need to have the desired end state in mind: full autonomous business process (ABP).
What does Fully Autonomous Business Process mean?
Let’s use the automobile industry example of autonomous vehicles as a not-perfect, but suitable parallel for what we’re trying to accomplish here. The vision of Level 5 autonomous vehicles is that fully autonomous vehicles will be able to drive human passengers anywhere in any weather condition – just tell the vehicle where to go, and it will take you there. No need to have steering wheels or control devices because you can’t intervene even if you wanted to. That’s the automobile industry’s vision of fully autonomous vehicles. We have a similar point of view on autonomous business process.
You can look at business processes from two perspectives: from the perspective of all the things that the business actually does in its day-to-day activities fulfilling all the necessary operations of the business, or from the perspective of the underlying technology that enables the business to do what it does. The first perspective is clearly the perspective we agree with because it’s all encompassing. Either the business is performing a business process or it’s not. The second perspective of business processes is more problematic.
First, not all business processes are encoded in technology – some are purely human-to-human. Second, some of the technology processes are not truly business processes but rather reflections of the way technology systems are setup to deal with various business requirements. In fact, we’ve had direct experience with line of business managers who hate the way the technology systems implement a business process, making the process much more complicated than it needs to be or adding additional cost or time to something that should be much simpler. Procurement systems: I’m talking about you! Since IT companies are good at handling IT-focused business processes, it makes sense then that the conversations we’ve been having about business process in the context of software tooling is from the technical perspective of business process. However, this is short-sighted, problematic, and backwards-looking.
What businesses want are systems that can autonomously understand the business processes as they actually exist in the organization and, without human intervention, provide augmentative assistance to get the business from one point to the other. That’s right. We’re talking systems that can autonomously discover business processes as they exist, and autonomously figure out how the various people, technology systems, data, and resources connect to those processes, and then figure out ways to optimize these business processes. This is about autonomous process discovery & modeling, autonomous process analytics, and autonomous process optimization.
Using some of the ideas from our Intelligent Process Automation report (which will now be called the Autonomous Business Process market), here’s how those various concepts map to capabilities:
- Autonomous Process Discovery & Modeling
- Automatically identify process flows in new systems (“process discovery”)
- Automatic process documentation
- Automatically discover explicit as well as implicit process flows by observing actual human, data, and system operation
- Automatically discover the entities which a business interacts with
- Autonomous Process Analytics & Management
- Automatically identify actual key performance indicators (KPIs) and metrics that determine process efficiency
- Automatically use overall business goals to measure how discovered processes are actually performing
- Automatically identify people, data, and system bottlenecks
- Autonomous Process Optimization
- Automatically anticipate and mitigate process flow exceptions
- Automatically understand system data and interface changes & make dynamic process changes
- Autonomously find and fix missing or incorrect data
Notice how at this level everything is automatic and autonomous?
That’s Crazy. Does Fully Autonomous Business Processes even make sense?
How could it possibly make sense for software systems to autonomously do all the things we outline above? Wouldn’t business managers scream that a system is taking control over their business without human intervention? Accidents will happen and people won’t have a way to control the system. Does this sound familiar? We’ve had the very same conversations about autonomous vehicles. People decry that fully autonomous vehicles won’t be safe and will lead to potentially disastrous outcomes. However, there’s just as many arguments to be made on the flip side that autonomous vehicles will lead to fewer accidents and greater efficiency.
Do you think businesses really love their process automation systems in place? In many cases these systems impede business capabilities. Humans often work around these systems or have to build yet more systems to deal with those systems. Processes that are automated can go haywire, taking down entire businesses, or cause havoc. So, let’s ask this question: do businesses really want unintelligent automatons executing the IT department’s vision of the business as it kinda exists, or do they want intelligent, autonomous systems that accurately and adequately define the business as it actually exists and handles the processes as they continuously change, and as humans execute them, rather than how their underlying systems approximate them? We think the latter.
Our Continued Coverage of the Autonomous Business Process Category
It’s important to note that Level 3 ABP is a goal. We may never get to truly autonomous business process, and that’s ok. The point is for businesses to take control of their processes and use cognitive capabilities the right way. Slapping OCR and text recognition on a dumb automation of technology process that might be inherently inefficient is not the way to do it. Just as no vehicle manufacturer is currently at Level 5 autonomous capability, so too there are no software vendors currently at Level 3 ABP, and we expect it will take a long time to get there. Any vendor claiming Level 3 ABP right now is flat out lying. Don’t believe their marketing hype.
One thing you might have noticed is that we’ve shifted our discussion of this market from Intelligent Process Automation (IPA) to Autonomous Business Process (ABP). This is for two reasons. First, we already said that automation is not intelligence, and as such the term IPA is a bit of an oxymoron like fresh frozen jumbo shrimp. Second, we think IPA and the unintelligent RPA market are too easily intertwined and confused, giving vendors ammunition in their marketing. We like the goals of ABP, even if they are not immediately available. And so, we’ll use the term ABP and continue to cover only those vendors and advances in the market that are pushing the industry forward with ABP and ignore everyone else.
Latest posts by Ronald Schmelzer
- Amazon Dives Deep into Reinforcement Learning - November 11, 2019
- Amazon advances conversational applications - November 6, 2019
- The Increasing Expansion Of AI In Business And Government - October 28, 2019