Making Finance ‘Intelligent’ through Automation

A robot is not necessarily a dark metallic contraption on two legs that goes beep beep and nurtures a malicious intent to enslave humanity. Their glowing red lights doubling up as eyes, renders the narrative most convincing too. For people of my generation who grew up watching Terminator, it may be hard to think otherwise.

The idea of a Bot, an abbreviated version, has been conceived, to give it an image makeover (debatable of course) and distinguish the fact that it needn’t be a piece of hardware device at all. But a few million lines of code, perhaps. In the digital age, the line between hardware and software is constantly blurring.     

Bots don’t fall sick, they don’t get tired and they don’t make “silly” errors. “Digital Workmates” are already a reality and now with the crisis, the adoption across functions will accelerate many times over. 

The finance function, particularly accounting, has a high element of repetitive, rule-based, consistent, and template-driven tasks. They take up a massive amount of man-hours which can be brought down substantially with the aid of intelligent automation. Our experience of having seen how BOT’s have operated in other companies and also in situations where we have implemented RPA indicates that a single bot in terms of productivity is equivalent to 8 – 9 FTEs. Considering a monthly average salary of $8,000 or INR 20,000, straightaway there’s an annual saving of close to $800,000 or INR 2,000,000. Even after throwing in the cost of the BOT, the platform, and the implementation, it is easy to see the ROI that intelligent automation can deliver immediately. 

This does not include ‘permanent savings’ from year 2 onwards and savings from utilization of the BOT (in most cases BOT’s aren’t fully utilized). When revenue streams dry up, profit margins & cashflows are under pressure, as is the case now for many companies, ‘optimizing’ costs & diverting people to more productive activities is a solution that stares them in the face. The beauty is this is controllable as compared to the external factors.  

To make this simpler and to put this in context, here is a listing of certain activities that occur frequently, are voluminous, standard and require little to moderate judgment (arranged in the order of increasing judgment):

·     Data input – extracting data from multiple sources such as applications, emails, files & folders.

·     Data output with predefined format across channels.

·     Reconciliation of data received from diverse applications.

·     Managing the quality of data and testing for integrity & consistency.

·     Implementation of dashboards and reports generation.

·     Application of accounting & business rules.   

In the case of activities where the judgment required is low, the cost reduction may be between 60 – 80% (Study by Ernst & Young) when bots/intelligent automation is deployed. In other cases, the savings may range between 20 – 40%.  

A very important consideration should be about process efficiency. If the process is saddled with a high degree of inefficiency, then you cannot simply wish it away through automation, which is why process re-engineering or re-imagination needs to be carried out before one embarks on the journey of automation. 

Let us look at a simple organization with three stakeholders – customers, suppliers, and vendors. The organization’s core applications sit in the middle and there’s real-time data exchange between the stakeholders and core applications. Quite obviously, some processes will comprise repeatable tasks and some won’t. We can’t be applying a one-size-fits-all-RPA based application to the entire value chain. There will be complex processes which will be outsourced and require Advanced Analytics, AI/ ML tools, intelligent voice recognition, etc. at the service provider’s end. In the final scheme of things, there has to be solid integration between disparate systems to enable the transfer of information in a secure manner between authorized users. To add to it, information silos will have to be dismantled. 

Admittedly, this is a simplistic example to help me drive home a point but in reality, it’s difficult considering the complex/real-life business environment. Let’s consider an organization that is multi-locational and spread across 50 countries. It’s a complex maze where the legal framework is a critical consideration too. For instance, the violation of GDPR (data protection) norms can attract punitive measures that can simply wipe out companies. This is a vital component that needs to be looked into when companies deploy intelligent automation in their processes. 

At a basic level, bots can handle single transactions and at an advanced level, a “pool of bots” can follow process maps, move structured data and run through processes in an automated data center. They can also be allocated to different processes through manual interventions (human in the loop) or AI-led controllers in real-time. The level of sophistication is a factor of the speed required, degree of complexity involved, and of course, the cost angle. 

It isn’t all hunky-dory of course, as it appears on paper. A lot of RPA projects fail and they do so massively. In these cases, there’s often a gap in understanding what the company wants and what the service provider thinks the company wants. The project has to be jointly envisioned to identify the problem statement and the service provider’s expertise. It has to be monitored with clearly defined and mutually agreed upon objectives coupled with evaluation scorecards. Enterprises typically use the CoE (Centre of Excellence) model to operationalize automation initiatives. The CoE is tasked with developing an enterprise-wide automation strategy, roadmap, and the governance model, including standardization and prioritization.   

The pandemic will likely change the way companies operate. There are fence-sitters who are undecided about their digital transformation journeys, just as there are early adopters of RPA. It may be argued that this crisis has violently pushed many to the point of inflection where they simply can’t maintain the status quo any longer and continue to be profitable. This is also an opportunity for all of us ‘humans’ to upskill ourselves and work on areas/processes that require us to increase our use of cognitive abilities, in individual and in team capacities.

The future will be about “where” do we keep humans in the loop but in an AI-led environment, progressively we may need to answer “WHY”.