GPTs for Science Teams: Smarter Workflows, Clearer Communication

Welcome, reader!

In every lab, clinic, and startup, time is short—and clarity is everything. This week, we’re diving into how GPTs are becoming practical teammates for biotech, medtech, and healthcare professionals. Whether you're translating data, designing slides, or simplifying patient language, these tools are changing the game.

Let’s explore how to use them wisely, creatively, and with scientific integrity.

Featured Insight

How Custom GPTs Are Reshaping Scientific Workflows

From protocol drafting to patient education, custom GPTs are quietly transforming how scientific teams work. When tuned right, they don’t just answer questions—they amplify expertise. We explore how innovators are using GPTs to clarify complexity, support design teams, and boost credibility.

Innovation Showcase

GPTs Designed for Biotech and Medtech

We’ve curated a short list of domain-specific GPTs helping scientific teams tackle different aspects of design, analysis and documentation:

• Protein design GPTs: Generate novel proteins/peptides from language prompts with structural and functional accuracy.

• Research assistant GPTs: Provide domain-specific insights and data support for drug discovery and life sciences.

• Clinical documentation GPTs: Automate medical note-taking and administrative tasks, integrated with Electronic Health Records (EHRs).

Pinal is a 16-billion-parameter AI model that designs novel proteins from natural language prompts by generating structures first, then sequences. Trained on 1.7 billion protein-text pairs, it outperforms prior models in protein design. Experimental tests confirmed functional activity in several designed enzymes, showcasing its effectiveness for de novo protein engineering.

Peptide-GPT is a generative protein language model that designs peptides with specific properties like hemolytic activity and solubility. It uses a multi-step filtering and classification pipeline to ensure structural validity and desired biological functions, demonstrating strong performance in targeted peptide design.

Hiro LS by PatSnap is a PharmaGPT-powered AI assistant for drug research, offering instant insights across discovery, clinical pipelines, patents, and regulations. It features a chatbot interface, data visualization, and secure integration with scientific databases, helping biopharma teams accelerate R&D and improve decision-making.

Sunoh.ai is an AI-powered medical scribe integrated with eClinicalWorks that converts patient-provider conversations into detailed clinical notes in real time. It supports multiple devices and languages, reduces documentation time by up to two hours daily, and improves accuracy and workflow efficiency.

Doximity GPT is a free, HIPAA-compliant AI assistant for U.S. clinicians and medical students that streamlines clinical documentation and administrative tasks. It generates comprehensive patient notes, letters, insurance appeals, and patient education materials, supports multiple specialties with tailored prompts, and integrates easily with EHRs. Doximity GPT saves clinicians significant time, improves workflow efficiency, and ensures patient data privacy with zero data retention.

Practical Tools

Plug-and-Play GPTs You Can Use Today

Each tool below is ready to use and science-aligned:

1. BastionGPT: Simplifies FDA documentation for internal use

2. SciSpace AI Writer: Helps design and write scientific posters for conferences

3. Writefull: Drafts sections for grant applications using NIH/NSF style

4. SiteGPT AI FAQ Generator: Converts technical docs into FAQs for lay audiences

5. DiagramGPT by Eraser: Turns complex processes into clear diagrams

From the Field

Case Study: AI-Powered Personalization in Wellness: How WOO and 10Clouds Built a Custom GPT Chatbot for Tailored Health Guidance

WOO (Well Over One), a Polish health and wellness company, partnered with 10Clouds to create a custom GPT-powered chatbot for their mobile app. This collaboration focused on delivering highly personalized, AI-driven health guidance, behavior-change strategies, and wellness monitoring. The custom GPT was instrumental in translating complex, multidimensional health data into actionable, real-time insights, enhancing user engagement and trust. 10Clouds’ expertise enabled seamless integration, ensuring the platform adapts and evolves with each user’s needs.

Behind the Scenes

How to Build a Custom GPT for Science Teams

Our Content Strategist Aditi Gangal shares her experience in setting up a custom GPT in this section.

“At SciRio, I use personal GPTs to help me streamline project understanding, synthesise complex information to build client’s communication strategy. It is never about generating plug-and-play content; it’s about building intelligent assistants that support my thinking process.

Prompt Strategy: I define the assistant’s role as a research aide or editorial collaborator—depending on the project. For example, I might create a GPT that helps me map out stakeholder concerns in a policy brief or summarise the latest regulatory shifts in digital health. The prompts include explicit instructions to cite sources, ask clarifying questions, and avoid assumptions.

Sample prompts:

Data Input: I use only publicly available documents, like project reports, press releases, and brand manifestos, and use them to scope the work. I avoid feeding in private data or client drafts, keeping confidentiality and consent at the centre of GPT use.

Sample prompts:

Visual UI: When I explore message framing or story angles, I use structure-based cues (e.g., compare-and-contrast tables, narrative frameworks, or tone checklists) so that the assistant offers me editorial paths to consider, not final outputs. This helps me evaluate narrative directions while staying grounded in my own judgment.

Sample prompts:

Testing & Guardrails: I review outputs critically, cross-check with original sources, and ensure the GPT is not hallucinating or skewing interpretations. I also add reminders within prompts to "default to saying you don’t know rather than guessing."

Sample prompts:

While we build our communication strategy from scratch, custom GPTs have expedited our processes.”

Community Corner

Reader Picks: GPTs They Can’t Work Without

📩 Want to share your GPT use case? Reply and we may feature it!

Resource: Here’s how to set up your own agent from Open AI.

Missed our last edition? Read it here.

Subscriber Bonus

🧪 Build Your Own Scientific Assistant GPT

Step-by-step guide using OpenAI’s Custom GPTs

1. Choose the Right GPT Model

Use: GPT-4-turbo (via OpenAI’s ChatGPT Custom GPTs platform)

• Why GPT-4-turbo?

  • Lower cost and faster responses than GPT-4

  • Retains advanced reasoning, code support, and multilingual capabilities

  • Handles long prompts (128K context window)

• Where: https://chat.openai.com/gpts

  (Click “Explore GPTs” > “Create” > Use the no-code setup flow)

2. Define the Assistant’s Purpose

Be specific. A vague assistant will give vague answers. Start with:

“This GPT helps [WHO] do [WHAT] by [HOW].”

Examples:

• “Helps regulatory teams summarize FDA guidance and compare against past filings.”

• “Helps lab scientists convert protocols into visual flowcharts.”

• “Helps medtech companies convert device manuals into patient-friendly guides.”

3. Provide Clear Instructions (System Message)

Set behavioral boundaries. Include:

• Acceptable sources (e.g., PubMed, FDA.gov, internal documents)

• Format of responses (e.g., “Respond in checklist form unless told otherwise.”)

• Voice/tone (e.g., neutral, accessible, or regulatory-compliant)

4. Upload Reference Materials (Optional but Powerful)

You can upload:

• Protocols

• Guidance documents

• Internal glossaries

• Visual style guides

GPT-4-turbo can “read” these during the session and answer accordingly.

5. Know the Limitations (and Workarounds)

6. Test with Scenarios

Use realistic test cases:

• Can it summarize a trial protocol accurately?

• Can it convert SOPs into step-by-step visual instructions?

• Does it avoid making up claims or references?

Tip: Create a simple rubric (e.g., “Clarity,” “Accuracy,” “Format”) and review answers accordingly.

7. Add Guardrails for Ethics and Safety

Include:

• Disclosure messages (e.g., “This is an AI assistant. Validate with experts.”)

• Warnings not to handle patient-specific questions

• A fallback for “I don’t know” when data is unclear

8. Iterate and Improve

Each use uncovers edge cases. Periodically:

• Refine instructions

• Replace outdated uploads

• Add examples or banned phrases to avoid

Final Word

AI tools are only as good as the science behind them. When used thoughtfully, GPTs don’t replace expertise—they scale it, shape it, and make it more shareable. Let’s build tools that respect nuance, value transparency, and tell science stories well.

Have you read our latest article yet?