Sanofi’s strategic shift to in-house AI for enterprise control and cost savings

Sanofi’s digital leadership has decided that the most useful AI tools are not necessarily the most polished ones sold by the biggest software brands.

At the start of the generative AI surge, Emmanuel Frenehard, the French drugmaker’s chief digital officer, took a sceptical view of the enterprise chatbots being marketed to large companies. Rather than roll out an internal version of ChatGPT or broadly deploy Microsoft’s Copilot after a pilot, he concluded that the main att...

raction was security rather than materially better performance. That view helped shape Concierge, Sanofi’s in-house AI assistant, which went live in October 2024 and is now used by about 60,000 employees worldwide, or roughly 80% of the workforce.

Sanofi says the assistant draws on internal policies, organisational charts and data from systems including ServiceNow, Workday and SAP, while also connecting outside services. The company describes a broader AI strategy that spans research, development, manufacturing and employee productivity, with separate efforts ranging from discovery tools to everyday workflow support.

Frenehard said the design was influenced less by software vendors than by the hospitality industry, where a well-trained concierge can guide visitors, make recommendations and solve problems quickly. In his telling, the goal was to build an assistant that understands how Sanofi actually works, rather than a generic bot that merely sits on top of enterprise software.

That logic is now extending beyond simple employee queries. Frenehard is increasingly focused on agentic AI, but he is not eager to buy multiple vendor-specific agents that talk to one another across platforms. Instead, he wants Sanofi to run more of these workflows directly on its own central data lake, working with Snowflake and Elementum AI, whose platform operates inside Snowflake’s data cloud. Elementum says its approach is designed to let companies keep control of their own data and avoid being locked into a vendor’s database.

SAP has also been part of the wider architecture, with its Snowflake integration aimed at harmonising SAP and non-SAP data for enterprise AI. For Sanofi, the point appears to be consolidation rather than multiplication: one data foundation, fewer disconnected tools and more control over where the intelligence sits.

The pay-off Sanofi is targeting is substantial. Frenehard said he wants AI agents to resolve as much as 80% of employee requests automatically, a move he believes could save about 10 million euros a year in IT alone. He also expects procurement automation to deliver tens of millions more in savings by reducing manual purchasing work.

The company is now applying the same thinking to sales. Earlier this month, Sanofi began rolling Concierge out to 60 representatives, using it to prepare for meetings with doctors by surfacing relevant patient history, marketing information and lessons from previous conversations. Frenehard estimates that this could save each sales representative about a day a week.

Sanofi is not rejecting outside AI entirely. It continues to work with major hyperscalers and specialist providers, including Anthropic’s Claude for Healthcare, and has also partnered with OpenAI and Formation Bio on drug discovery software. But the company’s stance on agentic enterprise software suggests a broader shift in how large organisations may think about AI: not as a layer they simply buy, but as something they increasingly want to build around their own data and workflows.

That view comes at a moment when the software-as-a-service industry is facing mounting pressure from investors and customers alike. Many vendors have made AI central to their offering, yet the rise of agentic systems is also encouraging some buyers to bypass standard per-seat products in favour of more tailored internal platforms. Frenehard argues that Sanofi’s software should not resemble that of rivals such as Novartis or AstraZeneca simply because they all operate in the same sector.

Even so, he is clear that this shift is not about replacing staff. Sanofi says the savings from IT and procurement automation will come largely from reduced reliance on outside suppliers. Frenehard has warned that framing AI as a way to cut employees directly creates resistance rather than momentum, and he says the company has learned that workflow change is rarely smooth, even inside his own IT team.

In April, Sanofi held what Frenehard called a “digital month”, centred on AI training and on teaching employees how to use Concierge properly. The company’s broader message is that adoption will only matter if AI becomes embedded in the core of the business, rather than confined to pilots and productivity demos.

That approach reflects a wider pattern across enterprise AI. Research shared by General Assembly suggests that leaders are increasingly encouraging AI use, but most workers still confine it to low-risk tasks such as searching, summarising and drafting. Sanofi’s bet is that the real value comes later, when AI is wired into the systems that move work, money and decisions through the company.

Source: Noah Wire Services