Cactus Life Sciences (CLS) has operated since 2008, helping pharmaceutical and biotechnology organisations translate complex clinical and scientific data into actionable insights for healthcare professionals, payers, and patients — across the United States, Europe, the UK, India, and Japan.

The core challenge was scale. Scientific writers were managing enormous volumes of abstracts and clinical documents through manual, time-intensive processes — search string creation, document review, and structured data extraction. Project managers, meanwhile, were navigating overflowing inboxes coordinating remote teams across multiple geographies. The work was governed well, but it was unsustainable at scale without compromising quality or security.

CLS selected Microsoft 365 Copilot as the foundation for its AI transformation, prioritising its enterprise-grade security and seamless integration with tools employees already used daily. Security was non-negotiable: all client data had to remain protected under stringent pharmaceutical-industry standards.

Using Microsoft's agent builder framework, CLS built more than 30 custom automation agents tailored to key workflows. These agents now handle the retrieval and structuring of information from scientific literature, allowing writers to focus on higher-value analysis. Additional agents automate abbreviation checks, formatting consistency, and alignment with regulatory and publication standards. Project managers use dedicated agents to summarise lengthy email threads and auto-generate task lists — dramatically cutting coordination overhead.

As Odity Mukherjee, Lead AI Transformation at Cactus Life Sciences, described the philosophy: "We decomposed our workflows and built agents to accelerate each step, so people have time to review and raise quality. AI supports the labour-intensive parts; it doesn't own the task."

Since rolling out Copilot and its custom agents, CLS reports that structured data extraction is now 35% to 50% faster compared to previous manual workflows — even as every output continues to pass through the company's established human review and quality governance processes. This is the key distinction: speed has improved dramatically, but oversight has not been reduced.

Scientific writers are now processing larger volumes of articles within shorter timeframes. Project managers report significantly less time spent on inbox management and task coordination. Across the organisation, employees are spending more time on the work that truly requires their expertise — clinical insight, scientific judgment, and client strategy.

The CLS story is a model worth studying — not because of the headline efficiency numbers, but because of the architecture of trust built around them. In industries like pharmaceuticals, healthcare communications, and regulatory affairs, AI cannot simply be dropped into existing workflows. It must be embedded thoughtfully, with governance structures that match the gravity of the work.

Microsoft 365 Copilot's enterprise-grade security, combined with its deep integration into familiar tools — Word, Excel, PowerPoint, Teams — gave CLS the confidence to move quickly without rebuilding their technology stack. The phased rollout, starting with more than 30 targeted automation agents, allowed the organisation to test, refine, and expand without disrupting client-facing quality.

As Mukherjee summarised: "AI is powerful, but it requires careful training and oversight." CLS will continue exploring how agents can complement — not replace — human effort, with all AI-enabled outputs remaining subject to established review, quality control, and governance processes.

Key Takeaways

• →Cactus Life Sciences deployed over 30 custom Microsoft 365 Copilot agents, achieving structured data extraction speeds 35–50% faster than previous manual workflows.
• →All AI-generated outputs remain subject to human review and quality governance — CLS's "Human Anchored AI" principle is central to the strategy, not optional.
• →A Copilot Champions programme and centralised knowledge repository drove peer-led, bottom-up adoption rather than top-down mandates — critical for buy-in in a scientific culture.
• →Enterprise-grade security was a non-negotiable requirement; Microsoft 365 Copilot's built-in protections and dedicated project environments enabled CLS to work safely with sensitive pharmaceutical data.
• →Agents were designed to accelerate individual workflow steps — not to replace scientific judgment — freeing writers to focus on deeper analysis, insight, and client strategy.
• →The CLS deployment demonstrates that AI in highly regulated industries requires phased rollout, modular training on both capabilities and limitations, and a culture of continuous, governed experimentation.