Digital Transformation
Apr 29, 2016

The Future of F&A: The Intelligent Digital Assistant

Genpact has built its business on supporting complex legacy back-office processes. We believe that the Intelligent Digital Assistant (IDA) is the future of the business process outsourcing (BPO) model as it cannibalizes our core business, by replacing FTEs with Artificial Intelligence (A.I.), leading to greater capture and retention of process knowledge, reduction in the number of FTEs to successfully manage a complex process, and simultaneously improves core process KPIs, such as productivity and efficiency, while decreasing the cost to manage these processes. CFOs would agree that future F&A operations will be lean, efficient, value-adding, and effective. We think that to make this a reality, the next stage of evolution of “the intelligent process" is to create “living processes" which are “self-aware", “self-correcting", and “self-adaptive". We believe that the IDA represents that future and are taking proactive steps to realize it.

Reimagining the back office
The Intelligent Digital Assistant will completely change the way back-office processes are managed by incorporating process intelligence, which we contextualize for traditional OTC, PTP, and RTR processes. The IDA addresses the challenges of balancing local and global constraints in a complex multi-objective process and reduces the cost of managing that process over the long run. An example of conflicting constraints includes the need for the global process owner to reduce cost, while the local process owner may simultaneously suffer from a downgrade in quality – the question being, what is the optimal trade-off between global and local objectives? 

In a world managed by KPIs, the complexity of this question increases as we ramp up the number of KPIs we simultaneously care about (e.g., cost, quality, production, cycle time, exceptions, etc.), where the corresponding process stakeholders (CXOs, Operations Leaders, Finance Leaders, etc.) offer up conflicting constraints and business objectives that are local to their domain. The question being: How do we balance these potentially conflicting objectives to arrive at the optimal solution for the business, and how do we implement that solution as a process change?

The IDA approaches this problem as one of nonlinear parameter optimization, and combines machine learning and the IoT to create the framework for determining the optimal solution, while providing the intelligence to deliver actionable process change recommendations that will effectively reduce process errors while improving process quality, execution, and cycle-time from beginning to end.

This is not merely Robotic Automation
Whereas robotic automation merely seeks to replicate human effort, the IDA takes innovation to the next level by actively managing end-to-end process execution while proactively identifying ways in which the process can be modified to achieve desirable outcomes. By exposing the total process to an intelligent continuous optimization framework, we can effectively achieve the equivalent of “Lean Six Sigma on steroids."

Consider a typical order-to-cash (OTC) process template:

Order Entry, Credit Check, Processing, Billing, Collections, Cash, Reconciliation & Reporting
Each process element in the IDA framework is continuously monitored through software and sensors (IoT), each of the inputs and outputs are measured (captured, digitized), each action is mimicked (automated), and local and global KPIs are balanced (optimized) against one another to yield improvements in process efficiency, quality, productivity, etc. The IDA simulates the process virtually, applies hypothetical process changes in the form of virtual lean experiments, and hardens those changes that yield improvements, while pruning solutions that degrade process performance.

“Process Self Awareness"
A self-aware process is self-monitoring with the help of the IoT. In the future all “Things" will be connected to the digital world and will be broadcasting their state, allowing us complete visibility into their workings on a continuous basis. This will form the basis for a continuous self-monitoring process. Couple this with machine learning and we can start to teach machines to understand how actions translate to process outcomes (KPIs).

“Process Self Correcting"
A self-correcting process is one that identifies problems leading to process exceptions, spiraling costs, or unmet productivity, and takes actions to remedy these scenarios, through an intermediary we call the “actuation layer" that actively makes process changes by adding and subtracting process steps. A process step can be as simple as moving an object from one point to another, filling in an entry in a web form, etc.

“Process Self Adapting"
A self-adaptive process is one that responds to changes in its environment. The world is a dynamic place, and processes must be adaptive to ensure that they do not become dislocated relative to their environment. This means enabling upstream and downstream visibility and embedding that into the definition of the process. Examples of real-world changes that can have process impact might include something as simple as the change in location of a client's primary business address. This can cause reconciliation issues which can create holes in the process whereby leaks will form, eventually leading to data ruptures which must be manually re-integrated at some point in the future at high cost. If the IDA was monitoring such changes in the external environment, it would be able to autolink “sunset" data to “sunshine" data, enabling process continuity amidst external (not in our control) changes.

Author: Pradyumna S. Upadrashta - Chief Science Officer, Analytics & Research