Going Beyond the Prompt: Making AI Useful in Medical Affairs
Scott Thompson shows why harnesses, connectors, and skills shape AI output more than prompt wording, building an MSL pre-call plan step by step in Copilot.
Many Medical Affairs teams have answered disappointing AI output with more prompt training, and the gains level off quickly. The gap between using AI and getting value from it is broader than pharma: McKinsey's 2025 State of AI survey found that 88% of organizations use AI regularly in at least one business function, yet no more than 10% report scaling AI agents in any single function.
In this webinar, Scott Thompson (CEO, Acceleration Point) explains why three elements around the prompt matter more than the prompt itself: the harness, connectors, and skills. A harness is the software that tells an AI model what to do and where it does its work, while the model is the engine inside it. Connectors give the harness structured access to sources such as PubMed, ClinicalTrials.gov, the NPI Registry, a CRM, or SharePoint. A skill is a reusable playbook of steps, standards, templates, and guardrails the AI applies every time a task is requested. Scott Thompson demonstrates all three by building an MSL pre-engagement plan for a physician meeting, using a fictional HCP and a fictional product. Each layer made the plan more structured and specific, while the prompt itself barely changed. As Scott Thompson put it, "It's the fact that we built things around it." If your team has finished prompt training and the output still feels generic, this session shows what to build next.
What You'll Learn
- A harness and a model are different things. The same harness, such as Copilot, can run models from several providers, and the harness determines whether AI only returns guidance or can do multi-step work for you.
- Connectors improve accuracy because they teach the AI how a source is structured. With a PubMed connector, the AI reads a record as a title, author, and abstract instead of guessing from a web page or hitting a paywall.
- When a direct CRM connection is not permitted, a scheduled daily CRM report sent to your email and saved to OneDrive by a rule gives the AI interaction history to work from.
- A short text file describing your role, therapeutic area, region, strategic priorities, and key insight questions, paired with an exported interaction history, is enough context to produce a noticeably more personalized pre-engagement plan.
- Building a skill does not require code. Scott Thompson asks the AI to critique its first output, ask him clarifying questions, and then "Build the skill and save it." Because "agents exist to make you happy," he adds a second, independent evidence-reviewer agent that checks citations before the plan comes back.
- The best AI projects shrink an end-to-end process timeline, not only individual task time. Saving a medical director an hour and a legal reviewer two hours adds little if an HCP contract still takes 48 days to process.
FAQ
What is an AI harness, and how is it different from an AI model?
An AI harness is the software that tells an AI model what to do and where it does its work. Copilot, Claude, ChatGPT, and Gemini are examples of harnesses. The model is the engine inside the harness, and one harness can offer several models. In the Acceleration Point webinar Going Beyond the Prompt, Scott Thompson recommends that Medical Affairs teams look for a harness that can do multi-step work for them, not only answer questions.
Why use an AI connector instead of letting AI search the web?
An AI connector teaches the AI how a source is structured and how to use it. Without a connector, the AI reads the open web, may hit paywalls, and has to guess which part of a page is a title, an author, or an abstract. In Scott Thompson's demonstration, connecting Copilot to PubMed, ClinicalTrials.gov, and the NPI Registry produced a more structured, better-cited pre-engagement brief for an MSL.
What is an AI skill, and how is it different from an AI agent?
An AI skill is a reusable set of steps, standards, templates, and guardrails that the AI applies each time a specific task is requested, such as MSL call preparation. An AI agent is given a goal and decides how to reach it, and it can call one or more skills along the way. A single skill can be used by multiple agents, and most platforms let teams share a skill so updates reach every user.
Do Medical Affairs teams need IT support to start building with AI?
Not for every step. Individual users can create context files and build skills on their own. Connector access depends on how the organization has configured its AI platform: some teams can add connectors themselves, while others must request them from IT. Before connecting AI to CRM data or approved medical materials, Medical Affairs teams should confirm the connection fits their organization's IT, compliance, and data governance policies.
When should a Medical Affairs team use something other than a general AI platform?
Scott Thompson describes two boundaries. First, AI is best reserved for judgment-heavy tasks that require reasoning; a simple rule, such as filing every PubMed alert email into one folder, is better handled with plain automation. Second, custom software becomes the better fit when a team needs a dedicated interface, data that is stored and updated over time, or many users working in the same tool.
About Our Speaker
Scott Thompson is CEO at Acceleration Point, which he founded about 16 years ago. Scott Thompson teaches AI to Medical Affairs teams and builds AI-enabled solutions for Medical Affairs. He and his team work side by side with Medical Affairs organizations to document processes from beginning to end and reduce the time lost between steps.
Connect with Scott on LinkedIn: https://www.linkedin.com/in/medical-excellence/
For more on AI in Medical Affairs, watch The AI Reckoning in Field Medical and AI for Medical Affairs: From Medical Leadership to MSLs.
Curious how this applies to your organization? Let's talk.
About Acceleration Point Webinars
Acceleration Point webinars bring Medical Affairs and Field Medical leaders together to share practical insights, emerging strategies, and real-world perspectives on the challenges shaping the industry. Each session is designed to support the professionals driving scientific exchange, field medical excellence, and Medical Affairs transformation.
Explore More Resources
Discover more insights, case studies, and podcasts from our team.
Browse All Resources