Roche’s AI playbook puts governance and work redesign first
Roche is using AI across diagnostics, drug discovery, manufacturing, and management support, but its technology chief says the harder work is organizational. The company’s approach offers a practical lesson for leaders: data, governance, skills, and workflow redesign have to move together.
For employers trying to move AI from experiments into everyday work, Roche offers a useful signal: the decisive question is not whether the technology is powerful enough. It is whether the organization has the data, governance, talent, and operating model to use it safely and repeatedly.
In an interview with CIO, Roche chief digital and technology officer Wafaa Mamilli described AI as a leadership and change management challenge as much as a technical one. That framing matters in healthcare, where faster analysis can affect diagnosis, drug development, manufacturing, and patient outcomes, but where errors and weak oversight carry real consequences.
Roche’s stated mission is to prevent and cure disease, improve patient outcomes, and reduce structural costs in healthcare systems. Mamilli said data and AI are central to that strategy, supported by more than 125 years of data accumulated by the company. The task now is to use AI to interpret large and varied information sets, including genomic sequences, medical images, clinical records, and complex biology.
AI use cases are spreading across the value chain
Roche is not treating AI as a single corporate tool. Mamilli pointed to a portfolio of applications that runs from internal productivity to clinical and industrial uses.
The company has built an enterprise platform called Galileo, which includes internal generative AI tools, search, and controlled environments where employees can experiment while staying within compliance rules. It has also developed AI coaching tools intended to help managers prepare for employee conversations, a reminder that AI adoption is reaching into people leadership as well as scientific work.
In diagnostics, Roche has implemented an AI algorithm called Kidney Klinrisk. Its purpose is to help physicians predict the risk that chronic kidney disease will progress before permanent damage occurs. In digital pathology, AI image analysis is being used to help pathologists identify cancer cells more quickly.
Roche is also applying AI in research and development. Mamilli said the company uses AI-powered simulations to model clinical trial scenarios. In drug discovery, scientists can model more than 100 experiments, compared with five in a physical plant. In manufacturing, Roche has deployed digital twins of production lines. According to Mamilli, those virtual models have reduced development cycles by two to four times and cut technology transfer times by 50%.
For business leaders outside pharmaceuticals, the pattern is more important than the specific tools. Roche is looking for places where AI can support expert judgment, compress testing cycles, and improve decisions inside regulated workflows. That is different from scattering chatbots across the company and hoping value appears.
Governance as a speed mechanism
Mamilli’s comments also challenge a common assumption that governance simply slows innovation down. At Roche, she described governance as part of the company’s ability to move quickly without losing control.
The company has created a horizontal framework focused on AI risk and ethics, along with an AI strategy and governance function designed to ensure models are used responsibly. In a business that develops medicines and diagnostics, that oversight is not optional. But Mamilli’s point was broader: leaders need enough structure to make decisions about which experiments can scale, which need more work, and which should stop.
That is a practical lesson for any management team. AI governance should not be a document that sits outside the work. It needs to be connected to investment decisions, operational risk, data quality, compliance, and frontline adoption. When those pieces are separated, teams either move too slowly or scale tools without understanding the consequences.
Mamilli also emphasized executive attention. She said she personally spends time understanding what is at stake because the environment is changing so quickly. That is a useful warning for boards and leadership teams. AI oversight cannot be delegated entirely to technical specialists. Senior leaders need enough fluency to ask better questions and make trade-offs deliberately.
The operating model is changing too
Roche has also reorganized its technology leadership for the AI era. Mamilli renamed her organization Roche Digital Technology, or RDT, to reflect a broader role in powering the business through digital technology and AI in a federated model.
She created dedicated AI and data teams reporting to two separate roles, a chief AI officer and a chief data officer. The split is intentional. Her view is that AI depends on data that is structured and ready for use, and that the data foundation still requires focused work.
The AI team, as Mamilli described it, resembles what the market often calls forward deployed engineers. In plain terms, these are technical specialists who work close to business teams, helping redesign processes and deploy AI in real workflows. They also feed lessons back to platform and tools teams so enterprise systems improve over time.
Her profile for the chief AI officer is not just a technologist. The role requires business outcome focus, technical understanding, learning agility, willingness to take risks, delivery discipline, and deep process knowledge. Mamilli linked the work to business process re-engineering, but with a sharper focus on what people actually do between systems, not only what software records.
That may be the most transferable insight. AI value usually depends on redesigning work, not simply adding a model to an existing process. Managers need to know where decisions happen, where handoffs break down, where employees use informal workarounds, and where automation should support rather than bypass human expertise.
For tomorrow’s CIOs, Mamilli’s advice was to be both business leader and technologist. She argued that business fluency is now expected, but deep technology understanding is also essential because AI is changing how companies operate, make decisions, and create value.
The leadership challenge is to move with confidence without pretending the path is fully known. Roche’s example suggests that the organizations best positioned for AI will be those that learn quickly, govern thoughtfully, and bring employees with them as the work changes.
Reported by Hybrion Insights with reference to CIO.
hybrion
Better together. Built for what's next.
Stay updated
Get insights on AI, teams, and the future of work.