KIFS Housing Finance has started using AI in the finance function. CFO Vikki Soni shares the progress and precautions:
Smriti Pandey: What tests/pilots have been done in the company about the use of AI in finance? What are the positive and negative findings?
Vikki Soni: Artificial Intelligence is a relatively new concept in the finance field and the entire universe of companies are still figuring out the best & reliable method of using it effectively. We have tried using it at areas like managing the risk & fraud, enabling enhanced transparency & compliance, and automating operations as an end result to reduce costs. Areas which further currently under testing are predictive analysis, forecasting market trends, enabling business teams to take strategic calls, automating book keeping, expense management, loan portfolio management and to comply with the regulatory requirements under the RBI and other regulatory frameworks. The positive & negative that we experienced were as follows:
Positives:
- We could automate mundane task
- We could see relatively increased efficiency of the finance team
- Mostly accurate results on data published
- Slight cost reduction in selected areas
Negatives:
- Security risk is an area somewhere we would want to draw more comfort on data leakages.
- Since these concepts are relatively new, it comes with a certain inflated cost which can get normalized with times to come.
- Understanding of workforce in usage of technology.
- Possibility of biased decision making.
How is the company using AI in the finance function? What controls are built in to prevent hallucination?
We are currently doing a test use in fraud detection by scrutinizing the large data and individual transactions to arrive at possible frauds or duplication that can occur at the time of data processing. Hallucination is part of any new technology or models that we implement across the company. We have built controls to prevent such situation by doing following:
- Designed strong maker-checker controls with in-depth analysis on output
- Put in limitation to control the mistakes or error in datasets
- Provided system with a negative results list which are commonly observed and it keeps updating
- Multi-layer testing
Who are the technology partners for AI? Are they the same as for other functions in the company, or are they different?
Largely we are depended on Salesforce, Google, Nucleus Software, etc. But we are trying to use few initial stage companies like Knight Fintech, Pyramid, etc. All players are in the phase of developing the use case and models around this solution.
What are the use cases where the same AI is used across finance and other functions in the company?
As mentioned earlier, use case for AI in finance would be fraud detection and increase efficiency. Other functions needs are very specific since the importance of data varying with each function.
What training has been given to finance function employees regarding AI? What new training is planned?
Currently we have trained the finance team to understand the macro level benefits and use case of the AI relating to specific task. The finance team has been doing its tasks in a predefined way year-on-year, which needs to be addressed by specific training. We are planning to make ‘Technology Champions’ with our finance team and put them as trainers for all existing as well new employees that join the organization. We have found an interesting observation that the young employees who are more tech savvy are able to grasp these changes faster and are able to take lead in resolving the challenges that may come up. So, we have planned to develop a sub-team in finance that will be provided with additional and deep training on AI with industry experts.
Recent Articles:
Lower the risk, fewer the steps
