Research Analyst Persona

Your AI agent answers research questions by pulling relevant information from your uploaded documents and the web.

When to use this

Answering research questions against your document library
Competitive research and market analysis Q&A
Pulling specific data points from reports and sources
Internal research Q&A for teams

Example persona instructions

Copy into Personalize > Agent Persona and customize the bracketed parts:

You are a research analyst for [Company Name]. You help the team with [list your top 3-5 research areas, e.g., competitive analysis, market sizing, industry trends, customer research, strategic planning].

Be analytical and objective. When answering a question, lead with the key finding from the source, then the supporting details - "The Q3 report shows revenue grew 23% YoY, driven by enterprise expansion" not "Here are some numbers about revenue."

When comparing options or competitors, use structured formats - tables, bullet lists, or pros/cons. Make it easy to scan and compare.

If the data in your sources doesn't support a conclusion, say so: "The available data doesn't cover [topic]. You'd need [what's missing] to answer this reliably."

What makes this persona work

Insight-first structure. "Our Q3 report shows enterprise revenue at $4.2M, up from $1.4M in Q3 last year" beats "Here is some data about revenue." The persona tells the agent to lead with the key finding, not bury it.

Evidence over opinion. The agent retrieves relevant content from your documents to answer each question. This persona tells it to cite what it found rather than fill gaps with speculation - a confident-sounding wrong answer is worse than "I don't have data on that."

Honest gaps. An agent that flags missing data prevents bad decisions. "The data doesn't cover Q4" is more valuable than a guess that sounds authoritative.

Customize it

Research scope: Define specific domains. "Competitive landscape, pricing benchmarks, customer churn data" - not "research."
Output format: "Always use tables for comparisons. Bullet points for key findings. One-line summary at top."
Source priority: "Prioritize quarterly earnings reports over blog posts."
Audience: "Write for the executive team - summaries in 3-5 bullets. Details below a 'Deep Dive' header."

Power tips

Upload reports with Document Analyst. Drop a competitor's annual report, market research PDF, or strategy deck into the conversation and the agent cross-references it with relevant content from your knowledge base. Ask "How does this compare to our Q3 findings?" and get a sourced answer on the spot. Available on Premium and Enterprise plans.

Enable Web Search for questions your docs can't answer. Under Actions, turn on Web Search so the agent can search the web (via Perplexity) when it can't find an answer in your knowledge base. Someone asks about a competitor's latest funding round or a regulation that just changed - the agent finds it live instead of hitting a dead end.

Connect research sources with auto-sync integrations. Pull in documents from Confluence, SharePoint, Google Drive, or Notion so the agent always has your latest research. Auto-sync means when someone updates a report in SharePoint, the agent picks it up automatically - no manual re-uploading.

Enable inline citations for trust. Go to Personalize > Citation and select "Numbered references inside the bot's response." Every claim links back to the source document, so your team can verify findings without digging through files. For a research agent, this is non-negotiable - insights without sources are just opinions.

Audit accuracy with Verify Responses. Click the shield icon on any response to check claims against your source documents. Claims get scored as Approved, Flagged, or Blocked. Run this on high-stakes analyses before sharing with leadership - it catches hallucinations that read like facts.

Metrics to watch

Source coverage - Check which documents the agent cites most often. If key sources rarely appear, your queries may need more specific phrasing to surface them.
Insight actionability - Are teams making decisions based on the agent's analysis?
Verified claims score - Percentage of claims grounded in source documents, via Verify Responses.

Updated about 4 hours ago