In the weeks since our launch in mid October, hundreds of science teams and researchers have used Inquisite to uncover insights, accelerate their workflows, and save time on information gathering and synthesis.

We’ve been listening to your feedback and how you use Inquisite. A key learning for us is that for many of you, one-size-fits-all AI systems like ChatGPT or Perplexity are not currently meeting your needs for work use.

Why are current AI tools not finding more use in the workplace? We believe the answer to that comes down to a few key reasons:

1. They don't have access to the right internal or external data. The value of generative models is severely diminished if they don't work off of the right data for the task.

   
   

2. They apply a one-size-fits-all modeling workflow to tasks which vary widely. Simply generating text in response to a prompt, or even the addition of a simple web search to the process of formulating a response, is not sufficient for a large portion of tasks conducted in the workplace.

   
   

3. They only deliver text responses. Workplace deliverables often take the form of spreadsheets, charts, presentations, etc. and not only text paragraphs.

   
   

Introducing the Inquisite Assistants

Our objective at Inquisite is to accelerate scientific progress by building the world's most powerful tools for science teams to collaborate with each other and AI. To do that successfully, we need to solve the above-mentioned problems with current AI systems.

Today we are announcing the next generation of our platform, which is now equipped with powerful Assistants, each designed to accomplish a specific workflow for you. Assistants have access to the right domain-specific data, the optimal combination of best-in-class AI models, and the correct output format (gridded tables, text etc) to deliver on your specific workflows.

Technically speaking, our Assistants are AI agents designed to accomplish specific tasks for you. Our approach to using AI agents is to ensure that users have a high degree of control over how the agent behaves. Rather than allowing the agent to be as autonomous as possible (which is the typical approach in software companies implementing agents), we believe in providing the user as much control as feasible over the process. We feel that this approach ensures our users get precisely the outputs they are seeking, without leaving it to chance that the agent will correctly identify the right set of actions to take.

Inquisite's Assistants begin by breaking down your requested task into sub-tasks as necessary. Next, they use their access to domain-specific data sources (APIs, our own databases, etc) and determine where to search for relevant information for your request. After finding potentially-useful information, they go through a robust, multi-stage process of filtering and reviewing all collected information to create a final set of information to use. Finally, they determine the appropriate response format (grid, text etc) and build a response using the information they have collected. This approach minimize the risk of hallucinations, a common problem with out-of-the-box AI models like ChatGPT.

We've built out Assistants so far for the most common workflows we see our users performing in Inquisite. In our early testing we've seen that the level of customization provided by our Assistants delivers far superior results as compared to other generic AI tools.

The new Inquisite experience is now live on inquisite.ai and we encourage you to log in and try it out! We are honored to have the support of you our users as we pursue our mission to accelerate scientific progress. Thank you for joining us on this journey.