How can the Effective Cost of RPA Implementation Lead you into a Technical Debt?

What is a Technical Debt and How Companies fall into that?

Technical debt is a concept that reflects the implied cost of additional rework caused by choosing an easy or limited solution now, over a better approach that would take longer. It is often compared to the monetary debt in the financial world. If technical debt is not repaid, it can accumulate ‘interest’, making it harder to implement changes later on. To put it simply, “technical debt” is like borrowing money for immediate needs that become a burden over time through having to repay the loan with interest.

While companies want to implement an RPA solution, they take into account of the short term benefits of automating repetitive tasks and cost effectiveness of RPA but RPA’s limitations could make it a liability in the long term, increasing the overall cost of adapting to the automated process that only solves a part of a problem instead of addressing end-to-end O2C processes. RPA is not a self-learning solution. So, it will continue to repeat the same tasks unless a manual change in the programming is made to adapt to evolving processes. This results in a reduced shelf life of the deployed solution.

RPA technology can fail to do the required task even if there is the slightest of variations in the predefined process. Consider an example of a paper-based invoice reconciliation process. A bot can analyze an invoice only if it is first digitized or scanned, which requires another digital technology or human intervention. Moreover, if the invoice format has fields placed in different areas or the cheque line items are interchanged compared to the set rule, the bot may interpret the information incorrectly. RPA can be used to pull in data and documents using web aggregation from various sources, such as TPM, carrier portals, and customer portals for deductions analysis, but it cannot validate a deduction. That is where native AI-workflows can be used to analyze the deduction and identify the invalid deductions upfront. RPA may require additional support from third-party software, such as, AI or ML, which adds to the cost of RPA, which can increase by 3 – 10 times.

So effectively companies might have allocated budget for an RPA implementation and exhausted it after implementation. They come to a predicament where they need to shell out more money to keep the process running along with new supporting software with AI. Thus they fall into a technical debt wherein they are neither able to scrap the implementation nor can afford to spend more. The time taken to reverse the technical debt can be huge and costly for organizations. 

“As per E&Y, 30 – 50% of the RPA projects fail due to unsatisfactory ROI arising from hidden costs of outsourcing or implementing additional technologies and non-compatibility with hanging industry practices”

Artificial Intelligence and Machine Learning Automation can help you

Artificial Intelligence and Machine Learning have opened new avenues of digital transformation and automation, especially for financial processes. AI and ML integrates with the existing systems to make them smarter, enabling them to interpret unstructured information, find patterns in a huge pool of data, and take decisions without human intervention. AI/ML-based automation results in better visibility and clarity across all O2C processes.

“AI/ML-based automation can deliver significant business outcomes, such as 30 – 40% DSO (days sales outstanding) improvements, 35% increase in efficiency, and over 60% faster processing”

In Credit Management, AI can be used to automate complex tasks, such as assessing the creditworthiness of customers, establishing new links between data and evaluating cases based on past trends, and enabling credit controllers or decision-makers with data-backed insights on credit risk.

AI-driven automation can measure credit risk across business units by identifying the common high-risk customers. As the risk increases, the consequent reduction in the exposure to business can be measured.

AI finds application in proactive credit reviews as it can process both large sets of structured and unstructured data. An AI-enabled system could consider both, internal and external, factors to get a micro and macro-economic view of a customer’s account. Then, it could automatically set the credit review period and trigger credit reviews for troubled accounts. This lets the credit team know about a stressed account even before an incident happens.

This way, the A/R, credit and sales team can work in tandem to drive recovery and collections from such identified at-risk accounts.

AI goes a step further to automate the entire credit review process and predicts with a certain level of confidence the probability of an order getting blocked. There are several machine learning algorithms to predict blocked orders but one of the most reliable ones is the learning algorithm, such as regression and classification, which rely on Decision Trees and Random Forest methods. These algorithms are specifically suited for the prediction of blocked orders as they can achieve high classification performance with a set of decision trees that grow using randomly selected subspaces of data. This way, the collections and cash application teams can proactively act to see that orders that should not be blocked are taken care of.

In Cash Posting, exception processes can be automated through machine learning that reads the existing patterns of manually handling exceptions, since ~80% of the exceptions are repetitive.

In Collections and Disputes,  AI-based deductions can proactively predict the validity of disputes and fast-track the recovery rate of invalid disputes by triggering an automated workflow to collectors, notifying them about the invalid deductions. This enables collections’ teams to focus specifically on valid deductions and spend time on reducing write-offs

In Billing, AI Automation can create an integrating billing process in which data from different processes can be leveraged, such as contractual discounts or historic payment patterns. Also, these patterns can be analyzed to establish credit terms.

In Reconciliations, Automation with machine learning capability can create an automated cash application process that can reconcile unstructured data and different payment formats. Further, automation can predict a customer’s preferred payment method to speed up the reconciliation process. 

Conclusion

AI/ML-based automation unlike RPA not only makes the O2C activities more streamlined and efficient but also generates real-time insights that the Leaders  can use to embrace an Agile Operating model across their Global Business Services and make their GBS future-ready.  For example, credit risk insights help a sales representative specifically focus on customers at risk of order blocks and similarly, a collections representative gets a better clarity about prioritizing customers for payment collection.

Most of the customers are unaware of the weakness of RPA and fall into a technical debt. Then they realise the potential benefits of AI and ML-based solutions and by the time they realise the cost incurred becomes high to revert. To differentiate from simple RPA workflows that use simple “if, then” logic, finance and technology leaders must highlight and focus on AI- ML models that improve business outcomes as they learn more about the customer.