A powerful struggle for technological dominance is unfolding in Silicon Valley, and at the very core of this competition lies one transformative innovation: artificial intelligence. More specifically, the conversation often circles around advanced AI systems powered by large language models, commonly referred to as LLMs. The release of one such model in late 2022—OpenAI’s conversational agent, ChatGPT—marked a monumental moment, as it presented the remarkable potential of these models to the general public in an accessible form. What had previously been discussed primarily among researchers and niche technology communities instantly became a mainstream phenomenon. Business leaders from diverse industries began to recognize not only the implications but the opportunities that AI presented for enhancing decision-making, improving efficiency, and redefining customer experiences. At the same time, the immense success of ChatGPT placed every competitor under intense scrutiny, sparking a frenzied race to develop and release rival systems that could capture both market relevance and consumer fascination.
As the number of available LLMs continues to grow at a rapid pace, enterprises now face a complex question: which model, among the expanding list of contenders, should they deploy for their specific needs? For some businesses, this decision can be paralyzing, since each system has distinctive strengths, limitations, and integration challenges. However, for Ancestry—the widely known Utah-based genealogy company specializing in DNA analysis and massive historical records databases—the answer has been refreshingly pragmatic. Sriram Thiagarajan, who serves as the organization’s chief technology officer and executive vice president of product and technology, explained to Business Insider that Ancestry has decided not to narrowly align itself with any single solution. Instead, the company has adopted what he described as a ‘more-the-merrier’ approach, in which different LLMs are deployed as needed, depending on the use case. In his words, their strategy is explicitly model-agnostic: they employ offerings from a range of providers, whether Microsoft’s Azure and OpenAI’s products, Meta’s LLaMA model, or options available through Amazon’s Bedrock framework.
For Ancestry, the essence of effective AI integration lies less in brand loyalty and more in practical results. Rather than becoming tethered to the advantages or constraints of one vendor, the company has invested in a sophisticated infrastructure of its own, designed specifically to maximize flexibility. At the heart of this system is what Thiagarajan terms an ‘AI gateway’—an abstraction layer that allows their teams to seamlessly leverage whichever models are best suited for a given task. This gateway acts like a universal connector, enabling the platform to use diverse algorithms without forcing costly reintegration each time. Moreover, Ancestry has created an entirely bespoke agentic framework that enables their services to stitch together data, context, and personalization in ways that resonate with their customers. The central goal is to provide people with not merely raw genealogical data, but vivid narratives that help them uncover unique family stories and deeply personal connections across generations.
Reflecting on the company’s AI journey, Thiagarajan noted that Ancestry had only begun experimenting seriously with machine learning and artificial intelligence around the time he joined the organization in 2017. At that point, their primary ambition was to address one of their most pressing challenges: digitizing massive amounts of historical content. This was no small task. Ancestry’s treasure trove includes a vast spectrum of documents—ranging from birth and death certificates, military enlistment data, land and immigration papers, and census records, to older newspaper archives. To convey the scale, Thiagarajan revealed that Ancestry has amassed more than 65 billion distinct records spanning over 80 countries, resulting in an extraordinary data volume representing approximately 10,000 terabytes. These records form the backbone of Ancestry’s promise: to surface groundbreaking discoveries about users’ family histories.
Traditionally, handling such immense data collections was a dauntingly painstaking process. For example, digitizing the 1940 U.S. census—an expansive set of paper forms—took roughly nine months to complete manually some 15 to 20 years ago, and at a cost nearly an order of magnitude higher than what today’s technologies can deliver. The limitations of purely manual processing were clear: it was slow, prohibitively expensive, and inevitably error-prone. Recognizing this, Ancestry’s leadership embraced a paradigm shift. By applying advanced computer vision techniques, they began to explore the possibility of automatically transcribing and digitizing handwritten content without the need for exhaustive human intervention. This exploration blossomed into innovation. By 2021, the company had successfully implemented proprietary handwriting-recognition tools, merging pattern recognition, image analysis, and AI learning models. As a result, processes that once demanded many months could now be accomplished in fewer than ten days, at only a fraction of the prior cost. The efficiency gains and enhanced accuracy demonstrated the transformative power of AI when strategically integrated into workflows at scale.
Nevertheless, Thiagarajan stressed an important caveat: human oversight remains a cornerstone of their approach. Despite automation’s advantages, Ancestry acknowledges the necessity of verifying outputs for factual reliability. AI systems, while powerful, are not infallible. To guard against inaccuracies, the company has engineered automated control mechanisms and validation systems that dramatically reduce the manual workload but still ensure that everything produced is firmly grounded in truth. In genealogical research, where accuracy carries personal and historical significance, this commitment to factual rigor is indispensable.
Today, the organization’s enthusiasm for AI extends well beyond product applications and runs deeply into its internal culture. Within Ancestry, employees are actively encouraged to learn about, experiment with, and engage in discussions surrounding AI. Thiagarajan described a range of internal initiatives: from casual educational formats such as lunchtime forums and ‘brown bag’ sessions, to cross-departmental spaces where employees are free to test new models or propose innovations. Hackathons are held regularly, fostering collaboration across otherwise separate teams and reinforcing the idea that innovation thrives when embedded in the regular rhythm of work, rather than sequestered into isolated silos. Through these deliberate efforts, Ancestry has effectively created a company-wide ecosystem where AI is not just a tool to be used, but a shared discipline, a collective pursuit, and a driver of continual progress.
Sourse: https://www.businessinsider.com/ancestry-ai-models-cto-sriram-thiagarajan-2025-9
