Governing AI Effectively in Clinical Development
By Jon Newton, NewtonBio Consulting
For the past two years, organizations have encouraged the use of AI to enhance productivity and efficiency. Teams adapted their workflows around AI tools that simplified various tasks. However, a shift occurred in 2026, where leadership began to enforce tighter spending restrictions on these technologies. This abrupt change left many employees struggling as their AI-access allowances were capped, leading to stalled processes and increased complaints. The real challenge lies in managing AI utilization effectively, especially after encountering significant budget impacts.
Understanding the Cost Implications
Starting in 2026, AI services moved towards token-based billing, which depends on the volume of text processed rather than a flat subscription fee. This change exposed organizations to unexpected costs, highlighted by instances such as Uber, which rapidly exhausted its AI budget, prompting urgent reassessments.
Agentic AI—capable of planning tasks and calling tools—can drain resources exponentially more than basic chatbots. Companies need to recognize that initial budgeting for AI does not align with the actual consumption seen in operational realities.
Rationing Issues
The pattern of providing unlimited access to AI before later restricting it leads to significant operational disruptions. Employees, who have integrated AI tools into their daily workflows, find their processes halted when access is limited. This creates unforeseen challenges that can cripple ongoing projects and workflows.
Consequences Beyond Budgeting
Limiting access to AI can lead to unintended behaviors, such as employees resorting to personal tools or devices to continue their work. This “shadow AI” not only poses data security risks but has also been linked to increased incidences of data breaches involving sensitive information.
For industries like clinical research, this can be particularly detrimental as it involves handling sensitive patient data, confidential strategies, and other proprietary information that should not be processed through consumer-grade tools.
Strategies for Effective AI Governance
To avoid the pitfalls associated with AI spending and usage, organizations should implement thoughtful governance strategies:
-
Budget AI as a Meter: Design budgets based on actual predicted usage and implement controls early to prevent overspending.
-
Avoid Critical Dependence on Unmetered Tools: Identify workflows reliant on AI and ensure they have fallbacks that don’t solely depend on available token supply.
-
Model Appropriately for Tasks: Route routine tasks to lower-cost models while reserving high-end options for more involved work.
-
Provide Approved Tools: Instead of prohibiting the use of AI, present sanctioned tools that meet employee needs and establish clear data classification guidelines.
-
Focus on Outcomes Over Usage: Shift focus from metrics around token consumption to actual productivity and results achieved through AI.
-
Draft Policies Before Issues Arise: Establish acceptable use and data governance policies proactively to cover all employees and contractors dealing with sensitive data.
Conclusion
Leveraging AI in clinical research is not solely a cost issue; it’s about effective governance and adaptability. Organizations that approach AI as a critical utility rather than a fixed expense, and that prepare for its strategic management from the outset, will be better positioned to harness its potential while mitigating risks related to data security and operational disruptions.
About Jon Newton:
Jon Newton of NewtonBio Consulting has over 25 years of experience in clinical development, specializing in clinical operations, corporate development, and innovative partnerships. He has held leadership roles in global CROs and advises digital health ventures.
