Perplexity Spaces: Organize Research Around a Project

Perplexity Spaces organize research around a shared subject; Perplexity’s current help center calls these workspaces Projects. Create a Project, add relevant sources and instructions, then keep each research question focused enough to verify. Tools like fenbs can retain the approved findings and next steps when the work continues in another connected AI app.

Think of the workspace as the research bench. It holds material you are investigating, including competing explanations. Your lasting work record should make clear what you accepted, what remains uncertain and which action follows. That distinction matters more than collecting a large number of conversations.

How do you set up a useful research Project?

Start with one outcome, such as choosing an accessible booking system for a small design studio. Avoid a workspace named “Everything about software.” A specific goal gives you a way to decide whether a source belongs and whether the research is finished.

1. Open Projects in the sidebar and select New project. Use a name that describes the decision or deliverable.
2. Write a short description of the question, audience and scope. Include what the project will leave out.
3. Add project instructions that ask for source links, explicit uncertainties and a consistent answer format.
4. Add the small set of files and reference links needed for the first question. Check each source’s version before using it.
5. Run one narrow query and inspect the evidence before adding more material or inviting collaborators.

These controls appear in Perplexity’s current Project documentation↗ (opens in a new tab). Layouts can differ, so use its settings guidance if your screen shows a configuration panel instead of a Settings tab. Keep changing plan limits on the vendor’s page rather than building your workflow around an old number.

What should your instructions say?

Describe how to research, not which conclusion to reach. Instructions such as “prove our preferred vendor is best” bias the task before the evidence is checked. Ask for the strongest evidence on each side and name the circumstances that would change your choice.

Research the booking workflow for a small US design studio.
Use vendor documentation for product claims.
For each finding, include its source, date checked, and caveat.
Separate documented facts from your interpretation.
Do not treat a missing answer as a negative answer.
End with the next question that needs human review.

Add the expected deliverable: a comparison, an evidence table or a short recommendation. Explain the reader’s needs in ordinary language. “A client should book on a phone without calling us” is easier to test than “the solution must be seamless.”

When a person approves a requirement, save it as a decision rather than leaving it buried in a research query. In fenbs, Decisions and rules give connected assistants the agreed constraints before they work.

How do you keep files and web evidence current?

Keep a short source inventory: document name, original location, version or date, and why it matters. Remove superseded copies from the active research set when they would cause confusion. Preserve a link to the old version when it explains a past choice.

For Enterprise file connectors, Perplexity distinguishes a live source search from an attachment or Project import. Its file connector guide↗ (opens in a new tab) says imported Project files are snapshots, so later source changes require another import. A connected account alone is not evidence that a saved copy is current.

• For a product feature, open the vendor’s relevant help page instead of relying on a generated comparison.
• For a policy, check who issued it, its effective date and whether an amended version exists.
• For a number, identify its unit, period, source population and any missing qualification.
• For conflicting sources, retain both positions and explain what evidence would resolve the difference.

Use the answer’s citations as a route back to evidence. A citation beside a sentence does not guarantee that the source supports every part of it. Read the relevant passage and rewrite the finding more narrowly when necessary.

How should you divide the research into questions?

Give each query a distinct job. First establish requirements, then investigate one requirement at a time, and finally compare the verified results. Asking for the whole decision immediately makes missing evidence harder to see.

For the studio, one query might investigate client access on phones; another might check cancellation controls. A later synthesis can compare those findings against the studio’s requirements. Label an unanswered requirement “not verified” rather than quietly treating it as satisfied.

Before accepting a recommendation, ask for the strongest objection and the source supporting it. This is a useful review prompt, but it still needs your judgment. The broader AI at work library covers other focused research and review workflows.

What should leave Perplexity when research becomes work?

Transfer the verified conclusion, its source and the next action. You rarely need to copy the full conversation. For example: “The booking flow meets our phone requirement; cancellation behavior remains untested; test it before approval.” Include who accepted the conclusion and when.

Keep suggestions separate from approvals. A polished recommendation can sound settled even when nobody has chosen it. Record open questions explicitly, including the person who needs to answer, so another assistant does not start implementing a research proposal.

How to do this with fenbs

1. Create a free account and add the research project to your personal kanban board.
2. Add a context note containing the question, verified findings and source links. Leave uncertain findings labeled as uncertain.
3. Record the person’s approved choice in Decisions and rules and add tasks for unresolved tests or follow-up research.
4. Connect a supported assistant through Integrations. Ask it to read the project context before drafting or acting on the research.
5. Before stopping, save Where we left off with the last verified result and the next check. Use that record for the next AI handoff.

fenbs does not import Perplexity conversations or connect directly to Perplexity. It is useful when you intentionally save the work that another supported assistant needs; a research project that stays entirely in Perplexity may need only its own workspace. The separate Perplexity MCP guide covers that nearby connection topic.