Deloitte’s 2026 Life Sciences Outlook describes how Life Sciences leaders are entering 2026 with both optimism and a clear need for resilience. The report highlights accelerated digital transformation, generative AI, agentic AI, regulatory change, pricing pressure and cost control as important themes shaping the industry. In Deloitte’s survey, 48 percent of executives identified accelerated digital transformation as a major trend, 41 percent pointed to generative AI and 30 percent to agentic AI. At the same time, only 22 percent said they had successfully scaled AI, and just 9 percent reported significant returns from their AI initiatives. [deloitte.com]

For CFOs, these trends translate into a more practical challenge: how do we make AI useful, measurable and safe in a regulated business?

For a Life Sciences company with 25, 100 or 500 employees, the answer is rarely to start with the most advanced AI model. More often, the first step is to look at the systems, reports and data the company already depends on: ERP, finance, inventory, quality, purchasing, sales and management reporting.

1. How do we calculate the return on AI, not just test the technology?

Many AI initiatives start with curiosity. A department tests a tool, a team explores an assistant, or management asks where AI could increase efficiency. That is understandable, but it is not enough for a CFO.

In its 2026 outlook for pharma and biotech, ZS describes a shift from asking “where can AI work?” to asking “where will AI drive the growth that matters?”. The same report states that 67 percent of surveyed leaders warn that launching AI initiatives without clear goals and success metrics is a mistake. [zs.com]

For CFOs, this makes the first step clear: drive the data foundation discussion.

That does not mean finance should own the technology. But finance should help define what value means. Which KPIs matter? Which reports are unreliable or too manual? Where do we lack visibility? Which numbers would prove that AI has actually improved the business?

A good starting point is to ask:

  • Which reports are still built manually in Excel?
  • Which forecasts are based on assumptions rather than structured data?
  • Where do we lack visibility into cash flow, margins, stock levels or lead times?
  • Which quality-related costs are difficult to track?
  • Which processes create repeated manual work every month?
  • Which KPIs would show that AI has created measurable value?

For many SMEs, this is where AI becomes a data and reporting question before it becomes a technology question. If the company cannot trust its master data, inventory data, supplier data, customer data or financial data, AI will not magically solve the problem. It may simply make poor data move faster.

A CFO’s role is therefore to push the organization from “let’s test AI” to “let’s define the business cases where better data and automation can improve decisions”.

2. Where should a Life Sciences company start with AI?

The best starting point is usually not the most advanced or most regulated process. It is the area where the business value is visible, the data already exists and the compliance risk is manageable.

ZS describes two investment paths for AI in pharma and biotech: areas where value is easier to prove in shorter cycles, such as enterprise technology, data operations and commercial processes, and longer-cycle areas such as R&D and clinical development, where the potential is significant but the return can be harder to measure. The report also notes that supply chain and manufacturing currently show lower proven impact, but high expectations for future value. [zs.com]

For a small or mid-sized Life Sciences company, a practical starting point could be:

  • Financial reporting
  • Forecasting
  • Inventory analysis
  • Demand planning
  • Cost follow-up
  • Margin analysis
  • Cash flow visibility
  • Invoice flows
  • Purchasing patterns
  • Slow-moving stock
  • Sales and operations planning
  • Deviation trends
  • Management dashboards

Many of these areas can be reviewed using existing ERP data and standard Power BI reports. If the company does not need heavily customized fields from the start, standard reporting can be a fast and practical way to see what data is already available, where the gaps are and which KPIs are worth improving.

For example, a CFO might begin by reviewing:

  • Inventory value and stock turnover
  • Slow-moving or obsolete items
  • Forecast accuracy
  • Supplier performance
  • Overdue receivables
  • Purchase price development
  • Gross margin by product group or market
  • Cost deviations
  • Manual reporting effort per month

This is where AI can become practical. Once reporting is structured and data quality is understood, AI can support recurring analysis, identify deviations, summarize trends, or help users ask better questions about the data.

The key is to start with use cases where the organization can learn safely. A company does not need to begin with GxP-critical processes, clinical workflows or patient-related data. Those areas may come later, but only when governance, validation and data quality are mature enough.

3. How do we govern AI in a regulated environment?

In Life Sciences, speed is not enough. A fast answer still needs to be correct, traceable, and appropriate for the process it supports.

ZS highlights that responsible scaling of AI requires trust practices throughout the AI life cycle. The report also states that different AI use cases require different levels of rigor. It gives adverse event detection in clinical trials as an example of a GxP-regulated process where errors could affect patient safety or the study itself. [zs.com]

For CFOs, this means AI governance should be treated as part of risk management and internal control, not simply as an IT topic.

Before scaling AI, companies should ask:

  • Which data can the AI access?
  • Does access follow existing roles and permissions?
  • Is sensitive company information protected?
  • Can AI-supported decisions be documented?
  • Which outputs require human review?
  • Which use cases are low risk, and which are GxP-relevant?
  • Who is responsible if an AI-supported process creates an error?
  • How do we avoid uncontrolled use of external AI tools?

For companies already using Microsoft 365 and Business Central SaaS, Copilot can often be a practical and controlled starting point for generative AI in the Microsoft business application landscape. Microsoft states that Microsoft 365 Copilot and Copilot Chat offer enterprise data protection, that prompts and responses are protected under Microsoft 365 commercial terms, that access controls and policies apply, and that prompts, responses and data accessed through Microsoft Graph are not used to train foundation models. [learn.microsoft.com]

That does not remove the need for governance. It simply means that companies already working within the Microsoft cloud ecosystem may have a more controlled starting point than if employees use disconnected public AI tools with unclear data handling.

For a regulated SME, the most important principle is clear: AI should only work with data that the user is allowed to access, and company data must remain protected within a governed environment.

This requires collaboration between Finance, IT, Quality and the business. Finance defines value, control and follow-up. IT defines architecture and security. Quality defines validation and compliance requirements. The business defines where AI can make daily work easier.

4. Do AI and agents work on-premises?

This is one of the most important questions for Life Sciences companies that still run legacy, validated or hybrid IT landscapes.

For CFOs, the practical takeaway is this: on-premises AI is not necessarily impossible, but it is usually more complex. It may require connectors, integrations, private networking, additional governance, and more technical expertise.

That is why the ERP roadmap matters. A cloud-based ERP environment can often provide a simpler foundation for AI, automation and future agent-based workflows, depending on architecture, integrations, validation requirements, and governance.

If moving fully to SaaS is not realistic immediately, companies should at least consider a hybrid path that reduces complexity and prepares the business for secure AI adoption.

For Life Sciences SMEs still running ERP on premises, the relevant question is not only “can AI work with our current setup?”. It is also: “how much complexity are we willing to carry, and does our current architecture support where we want the business to go?”

5. When should we use Copilot, BI or specialized AI?

Not every AI solution should be used for every problem.

Benchling’s 2026 Biotech AI Report is based on a November 2025 survey of around 100 biotech and pharmaceutical organizations already actively using AI. The report shows that early AI wins among these organizations are often found in areas such as literature review, protein structure prediction, scientific reporting and target identification. It also notes that AI adoption becomes more difficult in complex, regulated scientific areas where data is fragmented, incomplete or hard to validate. [benchling.com]

For CFOs, this creates a useful distinction.

Use Copilot or similar governed productivity tools for:

  • Meeting summaries
  • Drafting and reviewing documents
  • Internal communication
  • Finding information
  • Supporting daily knowledge work
  • Summarizing policies, notes or reports

Use BI and ERP data for:

  • Management reporting
  • Forecasting
  • Inventory analysis
  • Cost control
  • KPI follow-up
  • Margin analysis
  • Cash flow visibility
  • Operational performance reviews

Use specialized AI for:

  • Scientific analysis
  • R&D workflows
  • Clinical processes
  • Advanced modelling
  • Highly domain-specific or validated use cases

For many Life Sciences SMEs, the right path is not to jump directly into advanced AI models. A more realistic approach is to start with secure productivity gains, improve reporting and data quality, and then move toward more advanced AI scenarios when the organization has the governance and data foundation in place.

What CFOs should avoid

AI can create real value, but it can also create distraction. CFOs should be cautious of initiatives that sound innovative but lack clear ownership or measurable business logic.

Avoid:

  • Buying AI tools before defining the business problem
  • Starting with high-risk GxP-critical processes
  • Assuming AI can compensate for poor master data
  • Letting every department test AI separately
  • Building business cases without measurable KPIs
  • Ignoring role-based access and data security
  • Treating Copilot or any AI tool as a replacement for governance
  • Underestimating the complexity of on-premises integrations

For SMEs, focus is critical. The best AI strategy is often a narrow one: start with a few use cases, measure the outcome and scale only when the value is clear.

The CFO’s AI checklist for 2026

Before launching the next AI initiative, Life Sciences CFOs should ask:

  1. What business problems are we trying to solve?
  2. Which KPI will prove whether AI creates value?
  3. Is the data accurate, structured, and accessible?
  4. Are we still relying on manual Excel-heavy reporting?
  5. Can we start with reporting, forecasting, or inventory analysis?
  6. Does access follow roles and permissions?
  7. Is the use case low risk, business-critical or GxP-relevant?
  8. Do we need a Copilot, BI or a specialized AI model?
  9. Does our ERP architecture support secure AI adoption?
  10. If we are on premises, what is our cloud or hybrid roadmap?
  11. Who owns AI governance across Finance, IT, Quality and the business?

These questions help move AI from experimentation to business value.

From hype to measurable value

AI in Life Sciences will not be won by the companies that test the most tools. It will be won by the companies that connect AI to measurable business value, trusted data, and controlled processes.

For CFOs, this means taking a more active role in the AI journey. Not by owning the technology, but by asking questions that determine whether AI becomes a business accelerator or another isolated pilot.

For small and mid-sized MedTech, Biotech and Pharma companies, the first step does not need to be complicated. Start with the data. Start with the reports. Start with the manual work that already slows the business down.

Then build from there.

In Microsoft-based environments, this often means reviewing how ERP, reporting, role-based access, and productivity tools can work together. Solutions such as Microsoft Dynamics 365 Business Central, Power BI and Microsoft 365 Copilot can form part of that foundation, especially when combined with industry-specific processes and a clear governance model.

The real opportunity is not AI for its own sake. It is better visibility, better decisions, and better control in a business environment where resilience, compliance, and cost discipline matter more than ever.

Want to go deeper?

Join COSMO CONSULT’s upcoming webinar (in Swedish) on 15 September 2026, Framtidens ERP i Life Science: Cloud är sekundärt – rätt branschstöd är avgörande. The webinar focuses on why Life Sciences and MedTech companies need more than a standard ERP, and how Microsoft Dynamics 365 Business Central together with COSMO CONSULT’s industry solutions can support compliance, quality assurance and traceability. [cosmoconsult.com]

register here

Sources

Author


By Per Bay, Solution Architect at COSMO CONSULT in Sweden