Intellivy

Intellivy is a consumer research platform that polls a real panel to test Amazon listings, images, prices and product concepts. Agents pull finished polls and their aggregated results into a flow and hand them to an AI model for a listing recommendation.

2 actions API key available

Intellivy website Platform Documentation Capability data verified 2026-08-26

A listing test finishes and the product team wants the verdict without opening Intellivy

List Polls returns the polls on the account so the flow can locate the finished test by title

Get Poll Results returns that poll's questions, aggregated results, total response count and Intellivy's own summary

The agent confirms the poll it read is the one the question was about, and that the response count is worth acting on

The verdict posts to the listing owner with the summary, the response count and the two options that were tested

The listing owner approves the change before anything on the live product page moves

What This Integration Enables

Intellivy polls a real panel of consumers on a specific commercial question: does this main image beat that one, does this title read better, does this price look right, is this concept worth building. FlowRunner agents read the finished work. Two operations is the entire external surface of this API, and that is a measured fact about the vendor rather than an unfinished connector. The API is read only. Polls are created and run in the Intellivy dashboard, and nothing in this connector can create, edit or delete one.

That narrow surface is the argument for using it in a flow rather than against it. A research tool earns its cost at exactly one moment, which is when somebody is about to change something and wants to know first. Get Poll Results returns the questions, the aggregated answers, the total number of responses and the summary Intellivy generates from them, in a shape an agent can put in front of a decision maker or hand to a model for a recommendation. List Polls is how a flow finds the poll to read. Everything consequential still happens somewhere else, on the listing, in the catalog, on the price, which is why this page's human gate sits at that boundary rather than inside Intellivy.

Without FlowRunner

Results locked in a dashboard Reading a finished test means someone logs in, screenshots the summary and pastes it into a thread

Tests untracked over time Each poll is judged on its own, and nobody can say which listings have been tested and which have not

Opinion arrives before evidence The listing changes while the panel result is still sitting unread

With FlowRunner

Results reach the decision The finished poll, its response count and its summary land in the channel where the call gets made

Tests logged as a record Every read appends to a sheet, so tested and untested listings are a list rather than a memory

Evidence arrives before the change The verdict is in front of the owner at the moment the listing edit is proposed, not after it ships

Use Case Scenarios

The verdict reaches the channel where the decision is made

A test on two candidate main images completes. On a schedule, the agent calls List Polls, finds the finished poll by title, and calls Get Poll Results for the aggregated answers, the response count and Intellivy's summary. It posts all of it into the merchandising channel in Slack with the two images side by side and the summary quoted rather than paraphrased. Nobody logs into Intellivy to find out what the panel said, and the result is in the thread where the listing change is actually being argued.
A running record of what has been tested

Every time the agent reads a poll, it appends a row to a workbook in Google Sheets: the poll title, the total response count, the date, the summary, and the ASIN the test was about. Over a quarter that sheet answers a question the dashboard does not, which is which listings have been tested at all. The gaps in the sheet become the research backlog, and the tested rows sit next to sales data so a change can be read against what happened after it.
A recommendation drafted, not decided

For a completed concept test, the agent hands the raw questions and aggregated responses from Get Poll Results to a model with the product context attached, and asks for a written recommendation on the listing copy. The draft, the panel numbers and the model's reasoning go to the listing owner together, with the draft clearly labeled as a proposal. If the owner accepts it, the approved copy is what gets written to the listing in Amazon Seller Central. The model writes. It does not publish.

Human-in-Loop Highlight

Everything this connector can do is a read, so the risk is entirely in what a flow does next, and the next thing is usually a change to a live listing. A main image swap or a price change on an active product page takes effect immediately, affects placement and advertising economics from that moment, and cannot be undone by changing it back, because the sessions and the sales that ran against the wrong version already happened. Two properties of this API make an unattended change worse than it looks. There is no operation to re-run a poll, so an agent that acts on a thin result has no way to go get a better one. And List Polls is an undocumented route: it is verified to exist and to require the key, but its filters, its response envelope and whether it paginates are not published, so a flow that believes it has surveyed every poll on the account cannot prove it. A later poll that contradicts the one the agent read could be sitting on a page nothing fetched. So the agent stops before the edit and posts to the listing owner: "Poll 'Main image A vs B, kitchen scale' is complete. Intellivy's summary favors image B, on the response count shown. This came from the undocumented poll list route, so I cannot guarantee it is the most recent test for this ASIN. Approve the swap to image B, or check the dashboard first?" This is the digital andon cord placed exactly where the evidence runs out. The agent gathered the research and stated what it does not know. A person owns the change to the live page.

Agent processes routinely

Detects exception requiring judgment

Clear match Continues automatically

Ambiguous Routes to human via preferred channel

Human decides

Agent resumes with decision

Agent Capabilities

2 actions

Polls

List Polls Returns the polls on the Intellivy account, which is how a flow discovers the poll ids that Get Poll Results needs. It takes no input, so it doubles as the connection test. This route is undocumented: it is verified to exist and to require the API key, but its filters, its response envelope and whether it paginates are not published, so the body is returned exactly as the API sends it and query parameters pass straight through.
Get Poll Results Returns a single poll with its title, its questions and the aggregated results, including the total number of responses and the summary Intellivy generates from them. This is the documented route and the way a flow reads the outcome of a listing or image test. The poll id is a UUID, while each nested question carries its own plain integer id, so the two identifiers are not interchangeable.

Frequently Asked Questions

What can FlowRunner do with Intellivy?

FlowRunner agents can run List Polls and Get Poll Results in Intellivy.

Does connecting Intellivy to FlowRunner require OAuth?

No. Intellivy connects to FlowRunner with an API key, no OAuth flow required.

Can Intellivy trigger a FlowRunner workflow automatically?

Intellivy doesn't currently expose triggers in FlowRunner. It connects as an action step inside workflows started by another trigger.

Start building with Intellivy

Free plan, no card required. Connect in minutes.

Start Building Free Talk to us