Companies have a lot of options for AI models to use. Ancestry says it's 'agnostic' and uses many of them.

In the competitive landscape of Silicon Valley, artificial intelligence (AI) has emerged as a pivotal technology, particularly following the launch of OpenAI's ChatGPT in 2022, which popularized large language models (LLMs). This innovation prompted a rush among companies to develop their own AI chatbots, complicating the decision-making process on which models to adopt. Sriram Thiagarajan, Ancestry's Chief Technology Officer, emphasized the company's flexible strategy regarding AI models. "We are agnostic to LLM models," he stated, noting that Ancestry leverages a variety of AI technologies, including those from Azure, OpenAI, Meta's Llama, and Amazon Bedrock. This approach allows Ancestry, a genealogy firm based in Utah, to prioritize outcomes over specific brand affiliations. To optimize this integration, Ancestry has created an AI gateway, which acts as an abstraction layer that selects the most suitable models for their needs. Thiagarajan explained, "We have built our own agentic framework to deliver unique family stories and personalized experiences for our users." When Thiagarajan joined Ancestry in 2017, the company was just starting to explore AI and machine learning solutions to digitize extensive records efficiently. Ancestry's collection includes over 65 billion records from various sources such as birth certificates, military documents, and newspapers, amounting to about 10,000 terabytes of data. Historically, digitizing such records was a labor-intensive process. For instance, converting the 1940 census data took nearly nine months, which was significantly costly. However, with the advent of AI, Ancestry shifted its strategy. "We asked, 'Why not apply computer vision AI techniques to automate digitization?'" Thiagarajan recounted. By 2021, the company utilized proprietary handwriting recognition technology, reducing the digitization time from months to just nine days and at a fraction of the original cost. Despite these advancements, Ancestry maintains a commitment to accuracy. Thiagarajan acknowledged the importance of human oversight, stating that the company has implemented automated systems to streamline fact-checking while ensuring that their AI outputs remain rooted in factual accuracy. Ancestry is also fostering a culture of AI education among its employees through initiatives such as brown bag lunches and internal forums. Hackathons are regularly held to promote innovation and collaboration across different departments, reinforcing the integration of AI into everyday workflows rather than isolating it within specific teams.