AI will replace humans decision-making and do it better: HDFC Life COO

AI has matured and is now embedded in all use-cases. Hence, a few years from now, it will be par for the course and will replace human decision-making in many processes. Automation has replaced human repetitive tasks but AI will replace decisions (by humans) and will do it better in most cases.

At HDFC Life, we have always believed that disruption can be mitigated not by the latest technology but by going closer to the customer and solving for the customer voice. We have and will continually invest in digital in making life simple for the customer in the areas of buying and servicing insurance.

We have created multiple tech teams (platform teams, data tech team, digital team, core legacy team) which cater to different needs of the customer and solve different problems using latest and emerging technologies. Our AI usage is divided in five key areas:

Text AI: WhatsApp bot ‘Etty’, Customer Avatar ‘Zoey’

Vision AI: ‘FaceSense’, ‘Bodmeter’, ‘Age Tymer’

Voice AI: ‘Ezra’ (Google Assistant) and ‘Elsa’ (Alexa)

Machine learning: Propensity, Risk and Customer Retention models

Cognitive bots: 250 bots live across the organisation

Cloud storage/computing has helped us increase scalability and enable customisation as per changing business needs. We currently have 25 cloud-native products with most of our applications migrated to the cloud platform. AI/ML has aided us in multiple areas such as underwriting engine, persistency modelling, risk mitigation at payouts; sentiment analysis at service touchpoints; hyper-personalisation of sales incentives, amongst others.

Cognitive bots have enabled us to automate manual processes across different functions to improve operational excellence. Some of our prominent customer-facing bots includes email bot ‘SPOK’, WhatsApp bot ‘Etty’ and Twitter bot ‘Neo’, which ensure a 24x7 service experience for our customers.

‘InstA’, a virtual assistant, leverages AI/ML and NLP technologies to answer product- and process-related questions of our sales and operations teams, enabling them to better serve our customers. It currently handles over 20 lakh queries per month across almost 1,000 query types, with an accuracy of around 99 per cent.

At HDFC Life, we measure the success of AI in terms of adoption. Any measure of success has to be customer adoption or sales adoption to benefit customers.

‘InstA’, for instance, has 14,500 active sales users. It is being used across all our branches and 17 call centres, addressing over 17 lakh queries.

‘Etty’ supports more than 650 service queries has served 6.5 lakh users and has handled more than 35 lakh queries with a 94 per cent accuracy

‘Ezra’ and ‘Elsa’ have seen over 1,000 inquiries in the last few months.

Our ‘FaceSense’ application, which is used for facial authentication of the customer, has processed over 42,000 cases year to date and processes almost 600 cases daily at our branches. More than 5,000 cases at the pre-conversion verification check and over 25,000 verifications at our medical centres

Our voice bot ‘SVAR’ is available in 14 Indian languages. It reaches out to 4.5 lakh customers every month for payment collections. Approximately ₹3.5 crore of payment per month is attributed to this bot. In HDFC Life, we use AI extensively across multiple products. FaceSense, for instance, is used to mitigate risk of incorrect payouts at branches. Customers walking in for payouts are asked to take a simple picture much like the ones we see at the reception of offices today. This is compared with the image at the policy inception, thus ensuring it is the same customer. FaceSense is also being used at medical diagnostic centres to mitigate the risk of misrepresentation at medical centres.

We use natural language processing to power our conversation engine and ‘Etty’. ‘SVar’ is used to automate routine calls.

All the above technologies are amenable to the insurance sector and fit in very well with the user requirements. AI adoption is very quick if the use-case is solved from the customer point of view; the solution is frictionless and is backed by enough data to make the decision-making more close to reality.

In my view, quality and availability of data — both structured and unstructured — is a larger barrier to AI adoption. While text data is sketchy, good-quality photos and voice samples of the customer are almost non-existent. Data consciousness has pervaded into companies in the last five years. Most social data is concentrated with a few companies. Storing relevant data and retrieving it, cataloguing it, governing it, collecting it, cleaning it are the foundation requirements for AI. Other barriers include measurement of RoI and clarity of expectations of the project.

I believe insurers may be at par if not ahead of the curve. Banks have traditionally used a lot of analytics and modelling for risk management given the nature of the business, hence the perception of higher adoption. But the use of new technology like cloud, voice AI and vision AI is higher in insurance. We launched our cloud data lake before banks; we have vision-based face sense application live in our branches for customer authentication. We have vision-based ageing visualisation ‘Age Tymer’ and BMI calculation ‘Bodmeter’, which isn’t there with any bank.

COMMENT NOW