ICICI Bank moving into complex automation, combining RPA, OCR and AI, ML

ICICI Bank was one of the first in the country and among few, globally, to deploy ‘Software Robotics’ that emulates human actions to automate and perform repetitive, high volume and time consuming business tasks cutting across multiple applications. The software robots at ICICI Bank are configured to capture and interpret information from systems, recognise patterns and run business processes across multiple applications to execute activities including data entry and validation, automated formatting, multi-format message creation, text mining, workflow acceleration, reconciliations and currency exchange rate processing among others.

RPA journey
After three years, the bank is running 1350 business processes on RPA, which is currently being used in customer on-boarding, transaction processing, post transaction servicing, reconciliation, loan processing, where it involves moving data between systems. “RPA is used in a big way at ICICI Bank. In the area of operations, the bank, on a daily basis does 6.5mn transactions. Mammoth organisations having a baggage of legacy systems, have high use cases for RPA. The reason being, legacy setups are not tuned to work with API in many cases and thus RPA can be leveraged; they don’t have good data upload functionalities and even if it’s doable, conducting validations is an issue,” says Anubhuti Sanghai, Head – Operations and Customer Service, ICICI Bank.

The actual building of the RPA tool began with the inhouse team working on it since 2012. The official announcement was made in 2016. ‘Corporate limits setup’ was the first process, that RPA was enabled for. The sanction for the corporate limit for fund and non fund based products is complex. Once the sanction is received, the disbursement is done. Subsequently, the data is stored in the core banking solution (CBS).

Hitherto, post the sanction, the processes were manual i.e the data was posted manually in the CBS after doing the necessary validations. With an investment of just two subject matter experts, a robot was built in a week to automate the process.

Currently, close to 1350 business processes are running on RPA. In technology operations, there are thousands of processes that are RPA enabled. It involves, data being transferred from and to unconnected IT systems. In 2016, ICICI Bank announced automating over 200 processes.

“However, they continued to be RPA enabled until they were fully dissolved into an automated system. So, the 200 processes were reduced to, for example 100 processes. They have now gone upto 1350. The function of the number of processes being run through RPA is dependant on the amount of digitisation of the overall systems in the bank, for example, the case of corporate limits mentioned earlier, is now being API enabled. The sanctions system is now connected with the CBS via an API and thus no longer is there a need for it to be run through RPA,” says Balaji VV, Head – Business Technology Group, ICICI Bank.

The more deep rooted an organisation moves towards digitisation, the more nuanced, the need for RPA. Currently the systems are partially digital and partially physical requiring the information take off and hand off. This is where the need of an RPA arises. However, the more business processes are digitised, removing the physical systems, it will take away the need for RPA kind of systems. “ICICI Bank is moving towards a service based architecture, which is primarily API based. Wherever API comes in, there is no need for RPA,” says Balaji.

Major RPA use cases
Bank reconciliation: Faults in bank reconciliation can lead to fraud. “ICICI Bank has acted in this regard. The bank is using RPA to reduce manual reconciliation and at times completely eliminate it. We have been using RPA over and above an automatic reconciliation team in the operations department. We use RPA to further bring down the false matches, which are coming in,” informs Sanghai.

Form 15G / 15H : Software robotics has been deployed to perform end to end process involving resolution of queries related to Form 15G / 15H. This form is a self-declaration form submitted by an individual for non-deduction of tax on interest income. There are lakhs of such forms that are processed annually in Q1 of the financial year. The activity would take around three days, but with software robotics, the turn-around-time has reduced from three days to same day closure with 100 per cent accuracy.

ATM query resolution: This is the process of handling ICICI Bank customer queries who have encountered unsuccessful cash withdrawal while using an ICICI Bank ATM. Earlier, the turn-around-time for resolving such queries was seven days. Now, with software robotics, the resolution time has come down to four working hours.

Similarly, there are significant time efficiencies that have taken place for many business processes like, mobile number updation for internet banking alerts, ECS presentation for loan repayments, Aadhaar details validation importer-exporter code validations, debit card re-issuance to name a few among many others.

“We have a process which involves fetching vehicle registration details field by field from the central motor registration repository (Vahan) and updating manually in our core system. Each record in a manual process used to take 10 minutes. With the deployment of Software Robotics, the same activity is being performed in less than a minute,” states Balaji.

RPA team
The automation team at ICICI Bank has 60 developers (30-35 fulltime with the bank and the rest are at vendor locations), who work on RPA and other kinds of automation like bots, AI, ML and physical bots. The same set of developers work on RPA for across functions like treasury, trade, retail, corporate, international operations, etc. Apart from developers, there are product, process experts, business analysts and a testing team.

The testing cycle involves, IT, users (customers and employees), security testing done through a change management software. The RPA is put into production only after it passes through these cycles.

The bank is working with 6-7 vendors for RPA. “We work on two models. Firstly, the development is given out to the vendor and secondly, the operations and the technology team work together,” says Sanghai. There are four levels of automation operational at ICICI Bank: RPA, bots, AI and ML and physical bots. “RPA is handled by three-four teams working under the operations team and one team under IT,” mentions Balaji. Technology and the operations team has graduated to a level where the IT team provides the front end access and the operations team has the people with the required skill sets to run the automation on their own without any help from IT.

Future roadmap and learnings
ICICI Bank is moving into more complex automation, which is RPA + OCR + AI, ML. “It involves, going through multiple processes, picking data through OCR, running AI, ML on it and also validating the same data and updating an altogether different system,” says Sanghai.

Companies should ultimately go for a services based architecture. Moving towards an end to end automated platform, which eliminates RPA over a period of time.