AI Assistant
The AI Assistant is a tool designed to help you build and fine-tune data tables quicker and more easily, without needing a deep understanding of how our data is structured or how the Table Builder operates. You can simply describe your area of interest or research question, and the AI Assistant will do its best to generate a table that fits. It's a faster, more intuitive way to explore DrugBank data, especially if you're not sure where to start.
It is currently available in beta, which means development is ongoing so you may occasionally encounter slowdowns, formatting quirks, or inaccurate results. It’s still learning how to interpret queries effectively, so we recommend reviewing its output with a critical eye.
Our goal is to build a reliable, intuitive tool that speeds up your research and helps uncover powerful insights.
What the AI Assistant Can Do
While the AI Assistant can help with a wide range of table-building tasks, there are a few things it’s especially good at right now. Whether you’re starting from scratch, refining an existing table, or exploring specific areas like the clinical trial landscape, it can save you time and effort. Here’s a look at what it is currently doing best.
Eliminates the learning curve
The AI Assistant gets you started in a table with just one question. No need to learn our tools or data structure, simply describe what you need in your own words. It interprets your request and builds a table for you, helping you explore results right away. You don’t have to figure out which starting point or filters to use, it’ll set you up.
Helps you build and refine tables
It can generate complete tables from scratch or improve ones you’ve already created. It will select a starting point, relevant columns, and filters based on your prompts. You can then adjust the table manually or continue refining by asking follow-up questions.
Connects drug discovery concepts through clinical trials
The AI Assistant is especially effective at answering clinical trial questions. It’s built to handle the many ways you might naturally phrase these queries, whether by drug, target, condition, or outcome. You don’t need perfect terminology. Clinical trials also serve as a strong entry point to other key data, such as drugs and related targets, making it easier to explore how everything fits into the broader drug discovery landscape.
Helps you navigate your options
While the AI Assistant can’t see or interpret your table, it can tell you what filters and columns are available, explain key terms, and suggest ways you might refine your results. Whether you’re trying to get more specific, change direction, or just see what’s possible, the AI Assistant can help you move forward with confidence.
Current Limitations
We are working to fully optimize the AI Assistant’s capabilities and functionality. As we make improvements you will notice changes that result in greater flexibility and accuracy. For the time being we are aware that there will be some gaps and rough edges. Here are a few of the current limitations we’ll be addressing:
It doesn’t analyze the data in your table
The AI Assistant currently has no ability to see the specific data you are working with so it can’t help you interpret, analyze, or explain the actual data shown. As a result, it can’t offer you contextual help or guidance about what other relevant data or features you might want to add to your table. It can provide general information about what types of data or filters are available to you.
It can’t make certain table adjustments
It won’t intuit the best order of your columns or customize the layout of your table based on your question. Currently, it can’t successfully sort the data within a column into ascending or descending order or reorder your columns. And, it does not currently control any Table Setting toggles or the Journal Reader data. If asked, it can guide you through making these changes manually.
It won’t save your table or chat history
Every table you build, whether with the AI Assistant or by manually selecting columns and filters, needs to be saved if you want to return to it. If you don’t save your table, your chat history will also be lost. However, if you have used the AI Assistant and you save your table, your chat can be reopened when you reopen that table.
Tips for Best Results
As a beta product the AI Assistant is still learning a lot about how to understand and best respond to queries. Here are some tips that we’ve found deliver the best results:
Use it alongside manual table tools
For the best results, use the AI Assistant alongside manual adjustments. You can build a table on your own first, then ask it to refine or walk you through your options. Or, start with AI and switch between manual and AI Assistant refinements as you explore. We have found this to be a very useful approach.
Save your work
Want to keep your table and conversation? Save it. Otherwise, you’ll lose your chat history when the session ends.
Give it a second try
It’s okay if your first prompt doesn’t work. Clarify, simplify, or restructure your request. The AI Assistant is still learning how you ask questions and how to most effectively answer them. Asking follow up questions or providing more directions can prove extremely helpful.
AI Assistant and Your Data Privacy
AI Assistant is currently in beta. It has been designed to help you build, refine, and explore DrugBank tables more efficiently. To continue improving it, we analyze how it’s used - but always with your privacy in mind.
Here’s what you need to know:
No retraining on your queries: We never use your individual prompts to re-train our models, and we don’t share your queries or data across users. What you type stays private to your session.
Prompt tracking for improvement: Similar to other searches on the platform, we do log prompts internally to understand usage patterns and improve performance over time. This helps us fix issues, expand capabilities, and ensure the tool continues to evolve in ways that benefit you.
No table data access: Currently, the AI Assistant does not view or interpret your actual table data. It operates based on your prompts and the structure of DrugBank’s data, not on the values within your saved tables.
Human-in-the-loop development: During this beta phase, members of our team may manually review prompts to identify failure points or improvement areas. These reviews are focused solely on improving system accuracy and reliability. They’re never used for marketing or external profiling.
These practices ensure we fully support you while respecting your privacy and ownership over your work. We’re committed to building a trustworthy, transparent, and effective AI experience, one that enhances your research without compromising your control or confidentiality.
If you have any concerns or questions, please contact us, we’re here to listen and improve.
We Want Your Feedback
We’re actively developing the AI Assistant, and your feedback is crucial. If you run into issues, spot errors, or have any ideas for improvement, let us know directly.