Intelligent Automation - Simplify please
Anand Ramakrishnan
Business Technology Leader | Process Excellence & Process Mining
What is intelligent automation? That’s the question I get asked often. I start by explaining that intelligent automation is a continuum that includes autonomics, rules engines, machine learning, natural language processing and RPA. Sure enough, minutes into my explanation, I know I messed up my answer. I was trying to explain one unknown of the questioner with many unknowns.
I always wondered if there is a better way to answer this question. I needed something that can explain multiple concepts of the continuum using things that are known and understood. I needed a story. Here is one shot at it.
Imagine user access management process for an enterprise application at a large firm that is currently manual. Tail end of the process is actual task of provisioning. Administrator logs into an application and creates a named user with specific roles extracted from an MS Excel list. This task has fairly structured inputs, standardized work instructions and predictable output. RPA tools can handle this task very well. Now, let’s move to a prior step of process.
Prior to approval of user’s role request, there is a need for separation of duties check (for SOX). Let’s assume, business process owners provide approval with help of cross-tabbed worksheets. These worksheets are essentially a set of embedded rules. It’s often difficult to keep these cross-tabs updated. Autonomics and machine learning based systems can help keep pace by automatically interpreting broad and stable policies, determine compliance, assess risk and provide approve/reject recommendations. While rules engine enforce rules, autonomics can formulate new rules based on higher level instructions (example: purchase order and goods inward can't be processed by one person). Now, let’s move another step back in the process.
Let’s assume the request was for change in access. In a conventional model, the request is created through email or ticketing system. Problem with this is: users often struggle to explain their access need in IT terms. For example, someone might need to modify a SAP sales order, but doesn’t know the specific roles in SAP that best suits him. Over/understated privilege request is common. In addition, requester might not provide all information (example - included/exclude organizational entities) requiring clarification/outbound calls, causing delays. What if this step is handled by a chatbot that takes in the request of user in natural language, disambiguate it (seek instantaneous clarifications) and guide users on requesting best suited specific application role? We are now expanding the process chain of automation.
We could extend this even further. What if we could leverage natural language processing to read job/role descriptions of the employee from HCM system and extract job/role objectives, entity boundaries and its relations within that document (IBM Watson Natural Language Understanding is a good fit here). This could allow us to automatically build and maintain an enterprise role profile for the user across all applications. Nice idea, right? Yes, in theory only. Quite often, the variation of language within JD's is high even within same family of roles (example: territory sales) – making HR JD standardization a pre-requisite. So, it will remain just a possibility for now, unless language in JD is consistent across the enterprise.
This brings us to the question – where do we stop in the continuum? The answer is economics. Not just the economics of adding a new automation capability, but also prerequisites for its effective use.
It’s therefore important to realize that even when its technically feasible to automate, economic feasibility is a must cross gate. Most automation implementation are struck in RPA stage (task or set of tasks) unable to cross these gates in the intelligent automation continuum. These are early days; we need to stay hungry but also be humble about art of possible vs economics of possible. Imagination is required for the former and scale helps the latter. Interestingly enough, imagination is one thing that requires humans, for now.
I am pretty sure, there are better stories and story tellers of this topic. Love to hear your story & comments.
Director of Global Delivery at Orion Innovation | Seasoned Program Management Professional | Trusted Development and Delivery Lead | Expertise in Governance | Leadership | Account Management | Client Relations
6 年Great post. Goes to show the combined value of interconnected systems that perform higher level functions like read (OCR), listen, interpret and augment our capacity for decision making. Imagination vs Economics is truly the bottom line. Also I have seen misguided expectations and lack of business case to stall projects in this realm. We need business leaders who can stay engaged and resist the compulsion to pull the plug!!!
Program Manager at IBM - PMI PMP , PMI ACP, SAFE RTE,POPM, AWS
6 年Very well explained with example ??