Research [vendor category] for [team or use]. My hard requirements are [list]; preferences are [list]; exclusions are [list]. Consider [named candidates or discovery scope]. For each candidate, verify the relevant claims on a specific source page and record the checked date. Drop any candidate that fails a hard requirement and explain why. Return at most [number] finalists with strengths, trade-offs, unresolved questions, and links. Do not contact vendors, create accounts, accept terms, or share our requirements without my approval.
Vendor research shortlist with evidence and gaps
Compare vendors against stated requirements, link each finding to its source, and return a shortlist with unresolved questions.
Copy-ready brief
Example output
Example shortlist for illustration only: Requirement: Data export in a documented format. Candidate A: Meets requirement according to [specific documentation URL], checked [date]. Trade-off: The documentation does not describe automated exports. Candidate B: Unknown. The public page says “portable data” but gives no format. Ask vendor to confirm before advancing. Candidate C: Dropped because the required region is not listed on its availability page. [Source URL] Next decision: Review A and B; no vendor has been contacted.
Trigger and cadence
- The reader supplies a category, requirements, and a decision deadline.
- One research pass followed by updates when requirements or sources change.
Inputs
- Required capabilities, exclusions, geography, and data needs.
- Candidate vendors or a defined discovery scope.
- Evaluation criteria and the date evidence should be checked.
Steps
- Convert the requirements into a comparison checklist.
- Find candidate vendors from the defined scope.
- Verify each claim on a source page and record its checked date.
- Drop candidates that fail a hard requirement, with reasons.
- Return a shortlist, trade-offs, and questions for a human to confirm.
Approval boundary
Contacting a vendor, starting a trial, or accepting terms waits for approval.
Stop conditions
- A hard requirement is ambiguous.
- A vendor claim has no usable source.
- The search would require sharing private requirements externally.
Expected output
A compact shortlist with evidence, exclusions, trade-offs, and next questions.
Receipt fields
- Requirements and sources checked.
- Candidate inclusion and exclusion reasons.
- Open questions and any approved outreach.
Recovery when it fails
- Mark unsupported cells unknown and revisit them when a source becomes available.
- Recheck the shortlist if a hard requirement or decision deadline changes.
On this page
Start with hard requirements
A vendor research shortlist should answer a specific buying question. The first step is to separate requirements that eliminate a candidate from preferences that help rank acceptable candidates. OperatorNest can research public sources and prepare a comparison, then hold any outreach or account creation for your approval. The example output is illustrative, not a record of a purchase decision.
Write hard requirements as testable statements. “Supports export in CSV” can be checked against a documentation page. “Easy to use” needs a defined task or hands-on evaluation before it becomes useful. Include geography, access needs, file formats, and service constraints when they affect the choice.
Record the evidence beside the claim
Each comparison cell should have a source link and checked date, or say unknown. A vendor homepage may describe a capability at a high level while the documentation limits it to certain accounts. Prefer the page with the specific condition. If the source is an article or review, attribute the claim and avoid treating it as the vendor’s own commitment.
For example, Sam needs an export format their team can use. Candidate A documents the format but does not mention automated export. Candidate B says “portable data” without naming a format. The shortlist should mark A as meeting the format requirement and automated export as unknown; B’s format stays unknown. It should not quietly mark both as complete.
Record exclusions, too. If Candidate C lacks a required region, note the exact source that led to removal. That lets Sam reconsider C later if the requirement changes or the availability page is updated. A rejected candidate disappearing without explanation wastes the next research pass.
Return a decision-ready shortlist
For each finalist, include the relevant strengths, meaningful trade-offs, and questions that remain. Keep the list short enough for a person to inspect. A good result may be two candidates and three open questions rather than a forced top five.
The research assistant use case covers broader investigation. This workflow ends at a shortlist. Contacting a vendor, opening an account, accepting terms, or sharing internal requirements is a separate step. The operator can prepare outreach questions, but the exact recipient and message wait for approval.
When a source is unavailable, mark the affected claim unknown and keep the rest of the comparison. If a hard requirement changes, rerun the filtering from the recorded candidate set and check whether previously excluded vendors now meet the criteria. The receipt should retain the original criteria, all source pages, checked dates, and each inclusion decision.
Run it on a schedule
Plan the timing and missed-run behavior with the recurring task planner. For documented schedule controls across assistants, read scheduled AI tasks compared.
Related tools
- AI model picker
Choose a model class for a founder task. Compare sourced model examples by context, input types, price tier and open weights.
- AI agent approval policy
Set clear approval rules for an AI agent. Choose systems and risk limits, then copy a policy that says what waits and what can run.
- AI task brief generator
Turn a rough request into a task brief with a goal, needed inputs, steps, approval points, receipt details, and a repeat schedule.
Common questions
How many vendors should appear on the final list?
Choose a number you can review, often three to five. Keep rejected candidates in the receipt with the reason they were removed.
Should an unsupported claim count as a failed requirement?
Mark it unknown. A candidate can remain under consideration, but the missing fact must be confirmed before a decision that depends on it.
Can the operator ask vendors for missing answers?
It can draft questions. Sending them or disclosing your requirements to a vendor waits for your approval.
Keep reading
- An AI research assistant that turns a question into a sourced briefIt researches a question or a standing topic and delivers a sourced brief, once or on a recurring schedule.
- A browser of its own, for real web tasksIts own browser to research, compare and fill in forms across the web, checking with you before it submits or pays.
- Approvals before anything that mattersIt asks before it sends, pays, books, deletes or posts. You approve, edit or hold from any channel.
- A receipt for everything your operator doesEvery task keeps a receipt: what was asked, what ran, which model did it, what it cost, and every approval.
Hand off your first task tonight.
Tell us your email and what you'd hand off first. We'll send your access details and help you set up your operator.