(US & Canada) VIDEO | We Adopt GenAI Uses that Benefit Without Compromising Privacy — EMD Serono Head of Data-Driven Innovations and Analytics

Suresh Martha, Head of Data-Driven Innovations and Analytics at EMD Serono speaks with Lana Feng, Co-Founder and CEO of Huma.AI in a video interview, about adding value with generative AI, Tableau deploying its own GPT through an OpenAI partnership, build versus buy, transitioning to agile, picking the right use cases, bringing in transparent AI solutions, and the industry’s approach to GenAI while ensuring guardrails to protect data integrity and privacy.

Merck Serono (EMD Serono in the U.S. and Canada) is a German pharmaceutical company and a biopharmaceutical brand and division of Merck.

Speaking about adding organizational value with generative AI, Martha mentions automating repetitive tasks and evaluating chatbot functionality. Apart from data privacy being the only concern, he states that the organization has an enterprise-wide data analytics system that combines most of the internal and external data.

Additionally, Martha affirms trying to put a language model on top of the system to make it easier for people to get answers. He mentions exploring the tableau visualization and shares how Tableau will be deploying its own GPT in partnership with OpenAI.

This will allow people to get insights right at their fingertips instead of asking an IP or a data analyst to build a dashboard, says Martha. He further adds that Tableau has agreements to put guardrails wherein OpenAI cannot use data to train the model or keep history, thus easing security concerns.

When asked about building versus buying, Martha points towards the need to have a hybrid approach. He maintains that sometimes, depending on the availability of resources, it is not feasible from a financial standpoint to build from scratch.

Therefore, Martha suggests using something that has been built already and has passed through the testing process. He refers to what Tableau is doing by tapping into OpenAI’s language model.

At the same time, Martha suggests organizations have enough internal resources to fine-tune the model and customize based on organizational needs, rather than making a massive investment to build from scratch.

Commenting on working with start-ups in the generative AI space, Martha states that it all depends on what the use case is and who has the solution for the use case with less investment.

When asked about reconciling traditional business models with innovation, he states that having a big team makes it easier to allocate time to explore and fail fast. Whereas, with a smaller team, it is challenging to get innovative.

Martha mentions allocating some time per week to see what can be done outside of the normal workflow. He also advises team members to reach out to other CDOs, to learn and bring the insights back to work.

Adding on, Martha shares that the organization has been transitioning to an agile model for a couple of years, and with generative AI, he sees the hunger in the team to learn new things.

About the challenges in adopting AI technology, Martha asserts that while evaluating whether to buy something from outside, the organization first tries to implement it within a private network for data protection.

Moving forward, he says that the organization cautiously adopts the use case that benefits without compromising data privacy. He mentions the corporate organization in Germany that bought an OpenAI model and customized it with all the ensured guardrails.

When asked about hallucination problems, Martha shares that it is still not a big problem for the organization as it is waiting for Tableau to come up with its language model where everything will be taken care of.

He further mentions hearing about a lot of the companies that are coming up with large language models, which can be deployed in the cloud environment so that it is protected and beneficial.

To bring more transparency to the AI solutions, Martha suggests not exposing AI too much, other than the variables used in the model, what it can predict, and not technicalities. He notes that showing the impact of AI will help get it through the leaders.

Furthermore, reiterating the pharmaceutical industry’s approach to generative AI, Martha mentions adaptations on the R&D side to reduce time in bringing new drugs to market. He confirms that generative AI has also been adopted in clinical trials and licensing as well.

In conclusion, Martha says that the concerns include ensuring guardrails, and assessing how to solve problems with it without losing data integrity and privacy.

CDO Magazine appreciates Suresh Martha for sharing his insights with our global community.