Deep Research lets the SDLabs AI assistant run a literature review for you, straight from your experiment. Ask a question in the chat, approve exactly what will be searched, and a cited report arrives in the experiment's Research tab ready to inform your design.
Deep Research is enabled organization by organization. If you don't see a Research tab on your experiments, contact us to enable it.
How to Start a Research Run
Open any draft or running experiment and start a conversation with the assistant.
Ask for what you want to know, for example: "Research which ligands work best for Suzuki couplings of aryl chlorides at low temperature."
The assistant turns your request into a standalone research question and shows it to you in an approval card before anything is sent (see Approving What Gets Searched).
Click Send it to start, or Change something to tell the assistant what to adjust.
Anyone in the experiment's team can start a run.
Approving What Gets Searched
Every run starts with a card titled Send this to literature search? It shows the full text that will leave SDLabs:
Research question — the question the research will answer
Focus areas — topics to prioritize
Constraints — limits such as conditions or scope
Nothing else from your conversation is included, and any word on the card may be used as a search term. Nothing is sent until you click Send it.
When writing the question, the assistant keeps the science — for example the chemistry and your parameter ranges — and leaves out measured results, company, site, team and project names, and how your optimization is going. If it leaves something out, it tells you, so you can put it back if you want.
While It Runs
A status bar in the chat shows progress and elapsed time.
The Research tab shows an indicator while a run is in progress.
You can keep working — the run continues in the background.
Reading the Report
Finished reports appear in the experiment's Research tab, each with a title, a short summary, and the full write-up.
Citations — hover a citation marker to see the reference; click it to jump to the full list under Sources. References link to the paper through its DOI.
Download — save any report as a PDF to share or archive.
Using the Findings in Your Experiment
When a report is ready, the assistant lets you know in the chat and asks how you'd like to use it — for example, to refine your parameter ranges or your expert context. You can also attach any finished report to a chat message to bring it into a conversation.
Who Can See Your Research
Reports belong to the team the experiment was created in (or to your personal workspace), and everyone in that team can read them.
Live progress updates are shown only to the person who started the run.
What Data Leaves SDLabs
Only the text you approve on the card leaves the platform. Your measurements, experiment data, uploaded documents, expert context and the rest of your conversation are not sent — unless you write them into the approved text yourself.
That text is used in two places:
AI models — the research is carried out by the model on Amazon Bedrock, running in the dedicated AWS environment in the same geography as your SDLabs deployment.
Public literature databases — the research agent turns the question into short keyword searches (a few words each) and sends them to public scholarly databases: OpenAlex, Crossref, ChemRxiv, PubMed and PubMed Central (NCBI), and Europe PMC. It also downloads openly available full texts of relevant papers from publisher and repository websites.
The finished report and a record of the run are stored encrypted in your SDLabs deployment's storage. Our product analytics record which buttons are used, never the content of your research questions.
Safeguards
Off by default. Deep Research is switched on per organization; when it is off, the Research tab is hidden and the assistant cannot start a run.
Your approval, every time. No run starts without your click on the approval card, and the card shows exactly what will be sent.
Question written to stand alone. The assistant is instructed to leave out results and identifying names, and to tell you what it removed.
Team-only access. Reports are visible only to members of the team they belong to.
Good to Know
Deep Research works from publicly available literature.
A run that fails or is cancelled does not appear in the Research tab — just ask the assistant to start it again.
An experiment cannot be deleted while a research run on it is in progress.
Where to Go Next
Meet Your SDLabs AI Assistant — what the assistant can and can't do, and where to find it.
The AI Assistant During a Running Experiment — putting research findings to work while your optimization runs.
