AI Assistant (Chat)

Aperture AI is a chat assistant built into the platform. You can ask it questions about your pricing data, competitors, and product catalog; tell it to apply filters, create lists, manage alerts, manage category attributes, manage MAP prices, adjust product-matching settings, export a CSV, build boards, or visualize data as charts and tables; and receive proactive alerts and a daily brief directly in the conversation. All of this works without leaving the page you are on.

Note

The AI Insights on the Dashboard and Competitors & Retailers pages, and the AI suggestions on Price Perception, are separate from this assistant and are not covered here.

Opening the assistant

When the chat panel is closed, a compact AI tab appears pinned to the right edge of the screen. Click it to open the panel. It slides in as a drawer on the right side of the screen.

When the Ask page is available, an Open Ask button in the panel header opens the conversation on the full Ask page. The × button closes the panel.

When the panel is open, the header shows the current conversation’s title (or “New chat”). Click the title to open the conversation switcher. A strip below the header reads “Sees what you see:” followed by the current page name and any active filters. Behind the scenes, the assistant also receives the live data currently on your screen with every message you send, so its answers reflect your actual context, not just the page name.

Asking questions

Type your question in the input field and press Enter to send, or click the send button. To start a new line without sending, press Shift + Enter. A line under the input reminds you that the AI can make mistakes and that you can type @ to mention a product, competitor, or list.

While the assistant is writing an answer, a stop button (“Stop generating”) appears in place of the send button. Click it to cancel the current response. After the answer is complete, the assistant may spend a few more seconds preparing follow-up questions and shows “Finishing up…”. If you send a message during that time, the status reads “Sending your question…” and the message goes out as soon as the assistant finishes.

@-mention entity picker

Type @ in the message input to open a typeahead picker. It shows matching products, competitors, and lists from your workspace. Navigate the list with the arrow keys and select with Enter or Tab, or dismiss it with Escape.

Choosing an entity from the picker resolves it to an exact ID before the message is sent, so the assistant uses the right product or list without having to guess or re-search.

Suggested questions

When the conversation is empty, the panel shows a landing view with your recent conversations, example questions grouped by area, and the message input in the middle. Click any example question to send it immediately. The starter questions are a fixed set, the same on every page. After a reply, up to three follow-up questions written for that specific answer may appear below it, and clicking one sends it. Some replies show no follow-ups. They are a separate list from the landing’s example questions.

Reading responses

Answers stream in real time. During generation, a status line below the messages shows “Thinking…”. Tool steps appear as rows describing what the assistant is running, for example “Searching products…”. If a response takes longer than usual, the status reads “Still working… this can take a few seconds for deeper analysis.”

Responses are formatted in Markdown. Bold text, bullet lists, numbered steps, headings, and tables are all rendered. Links in responses open in a new browser tab.

Under each finished answer, a Copy response button copies the answer. Pasting into an email or Slack keeps the bold text, lists, and tables. Thumbs-up and thumbs-down buttons record whether the answer was useful. After a thumbs-down, a “What went wrong?” row offers the reasons “Wrong numbers”, “Wrong products or competitors”, “Didn’t show the data”, “Too long”, and “Something else”. Conversations that were shared with you are read-only, so they show Copy but no thumbs.

Citation chips

Answers built from a product search or from price/availability movers carry a “View on Products” or “View on Movers” chip below the message. Clicking it navigates you directly to that page with the same filters applied, so you can verify the numbers yourself.

What you can ask

The assistant can answer questions and take actions across the full platform. It uses live data tools behind the scenes, so answers reflect current data rather than a snapshot.

Data questions

Below is a representative sample of what you can ask:

Topic	Example questions
Dashboard KPIs	”Give me a quick overview” · “What’s my overall price position?”
Competitors	”Which competitor has the lowest price index?” · “Who am I competing with?” · “Where are my biggest assortment gaps?”
Products	”Find me coffee products under €10” · “Find me black metal storage products”
Price comparison (retailer)	“How am I priced on Product X vs the market?” · “Who is cheaper than me on this product?”
Retailers reselling you (brand)	“Which retailers carry Product X?” · “Which retailer sells it cheapest?” · “Who’s breaking MAP on it?”
Price history	”Has the price of Product X changed in the last 30 days?”
Price distribution	”How are my prices spread across the market?”
Movers	”What are the biggest price drops this week?” · “Which products went up in price?”
New Products	”What new products appeared at competitors this month?”
Lists	”What lists do I have?” · “How many products are in my Watchlist?”
Alerts	”What alert rules do I have?” · “Which alerts fired this week?”
MAP Violations	”What were the worst MAP violations this week?” · “What’s my MAP compliance rate?”
Matches & Match Review	”How many products are pending match review?” · “What are my lowest-confidence matches?” · “How good is my product matching overall?”
Match rules	”What is the match confidence for Bottled Water?” · “Which products are missing from this match rule?” · “Which attribute is cutting my matches?”
Market Benchmarks	”Which products need comparable review?” · “How many Equivalent Matches do I have?”
Price Perception	”What’s in my price perception basket?” · “Suggest some KVI products for me” · “What price should I set on my KVI products to stay competitive?”
Attribution	”What attributes does Furniture > Sofas have?”
Data quality	”Is my catalog data ready for export?” · “Which categories have the most missing attributes?”
Digital Shelf	”What is my Digital Shelf Score by retailer?” · “Which listings are closest to their review target?” · “What are the deepest promotion cuts this week?” · “Which retailers don’t list our SKUs?” · “Which products fail content compliance at Walmart and why?” · “What is our share of search?”

Tip

For pricing advice, ask the assistant to optimize your KVI, or Key Value Item, selection under a strategy: “aggressive”, which prioritizes low prices; “balanced”, the default; or “premium”, which protects margin. It returns suggested prices for you to review; it never changes live prices itself.

Note

The assistant can tell you how many lists you have and how many products are in each, but it can’t list out a list’s contents directly in the conversation. Ask it to open the list instead, for example “Show me my list named Watchlist”. Once you’re on that list’s page, you can ask follow-up questions about the products shown there.

Visualizing data

Ask the assistant to see, plot, chart, or visualize your data and it renders the result inline in the conversation: a bar, line, or area chart, a ranked data table, or a set of stat cards. The result is built from your real, live data, never invented. Examples:

• “Show me my price distribution as a chart”
• “Plot the price history of Product X over the last 90 days”
• “Chart my biggest price drops this week”
• “Make a scorecard for [Competitor]”, retailer perspective
• “Chart which retailers carry the most of my products”, brand perspective

Charts and tables can carry a “View full chart”, “View full table”, or “View full comparison” link, shown with an arrow, that opens the corresponding page for deeper analysis. If you follow up with something like “now show that as a table” or “change it to the last 7 days”, the assistant updates the same visualization instead of starting over. It can’t change the visual styling, such as colors, fonts, or layout, only the underlying data and settings.

Each card shows a “Data as of” time, for example “Data as of today, 09:30”. A chart in the conversation is a snapshot taken when you asked, and the time never goes later than your last data import. When the data is more than 2 days old, the stamp turns amber and a tooltip explains why: either no new data has been imported for more than 2 days, or the chart is an old snapshot and you should ask again for current numbers.

In the panel, each card has an Ask about this button. Click it and a chip above the input reads “Next answer is about:” followed by the card’s title. Your next question is then about that card. Click the × on the chip to stop. The chip clears after you send the question.

Saving a chart to a board

After the assistant renders a chart, ask it to add or save the chart to a board, for example “add that chart to my board” or “pin this chart”. This works from any page. You don’t need to be on the Ask page first. Review the chart details in the confirmation card and click Add; the assistant saves it to a board in your Ask workspace (creating one for you if you don’t already have one), where it re-fetches live data every time it loads.

When the Ask page is available, each card in the panel also has an Add to board menu. It lists your boards and a New board with this option.

You can also ask the assistant to change a chart already saved on a board, for example “change the price-history item to 90 days” or “switch my movers table to show the top 20”. Review and click Update in the confirmation card to apply the change.

Filtering and navigation

Ask the assistant to apply, change, or remove filters on any page, and it will do so immediately in your browser. You do not need to touch the filter controls yourself. Examples:

• “Show me products priced 10–30% above the market”
• “Filter by brand X”
• “Show me black leather chairs”, filters by product attribute
• “Show me my out-of-stock products”, filters by availability
• “Clear all filters”
• “Take me to the Prices page”
• “Show me my list named Watchlist”

Note

When you ask the assistant to remove or clear filters, it always performs the action, even if no filters appear active. The assistant reads the current filter state from the page in real time, not from the conversation history.

Actions

The assistant can perform write operations on your behalf. Some are immediate; others require you to review a confirmation card before anything is saved.

Confirmed actions

For the following operations the assistant presents an inline confirmation card directly in the conversation, not a pop-up dialog. Review the details, such as name, product count, alert type, item count, product, price, recipient, or chart details as appropriate, then click Cancel to dismiss, or click the action button. The action button’s label matches what it does: Confirm for creating a new list, alert rule, or category attribute, or setting a MAP price, Update for changing a MAP price or editing a chart, Remove for removing a MAP price, Send for a MAP compliance notice or a content notice, Approve or Reject for a match, Delete for deleting a list, alert rule, category attribute, or board, Create for a board template, and Add for a new chart. No data is written until you click that button.

Action	Trigger phrase
Create a product list	”Create a list of coffee products”
Delete a list	”Delete my Watchlist”
Create an alert rule	”Alert me when a competitor goes below my price on Product X”
Delete an alert rule	”Delete my price-drop alert”
Delete a board	”Delete my Pricing board”
Create a board from a template	”Create a board using the Pricing overview template”
Add a chart to a board	”Add that chart to my board”
Update a chart already on a board	”Change the price-history item to 90 days”
Add a category attribute	”Add an attribute called Material to Furniture > Sofas”
Delete a category attribute	”Delete the Material attribute from Furniture > Sofas”
Add or change a MAP price (brand)	“Set a MAP price of €49.99 on Product X”
Remove a MAP price (brand)	“Remove the MAP price on Product X”
Send a MAP compliance notice to a retailer (brand)	“Notify [Retailer] about this MAP violation”
Send a content notice to a retailer	”Email [Retailer] about the content issues on this listing”
Approve or reject a product match	”Approve this match”

Deleting a list, an alert rule, a category attribute, or a board, removing a MAP price, and rejecting a match are destructive, and those confirmation cards are highlighted in red.

Immediate actions (no confirmation)

These actions take effect immediately and do not require a confirmation step:

Action	Example
Add products to an existing list	”Add Product X to my Watchlist”
Remove products from a list	”Remove Product X from my Watchlist”
Export the product catalog to CSV	”Export my coffee products to CSV”
Enable or disable an alert rule	”Disable my price-drop alert”
Subscribe or unsubscribe from an alert rule	”Add me as a recipient on my price-drop alert” · “Stop notifying me about that alert”
Send a test notification for an alert rule	”Send a test notification for my price-drop alert”
Run an alert rule’s check now	”Check my price-drop alert now”
Triage alert events	”Mark my recent alerts as read” · “Snooze this alert for 24 hours”
Update an alert rule’s name or threshold	”Update the threshold on my overpriced alert to 5%“
Update your KVI selection for Price Perception	”Set these products as my KVI selection”
Change product-matching settings for a category or the global rule	”Set the minimum match confidence for Beverages to 0.9”
Create a board	”Create a new board called Pricing”
Rename a board	”Rename my board to Q4 Pricing”
Add, remove, move, or resize an item on a board	”Add the price distribution item to my board”
Update a MAP violation case’s status or add a note (brand)	“Mark this MAP case as resolved”
Update a content case’s status or add a note	”Mark this content case as notified”

Note

A CSV export requested through chat is capped at 5,000 products. For a larger export, use the export button on the Products page instead.

The matching settings the assistant can change are the minimum match confidence, the price-closeness weight (low, medium, or high), whether out-of-stock products can still be matched, and whether to fall back to the next-best candidate when the best one is out of stock. Ask the assistant for the current values first, since a change replaces all four settings together.

Resolving a content case requires a reason: content corrected, accepted difference, false positive, or other. The assistant asks for it if you don’t give one.

Undo

Some immediate write actions show an Undo chip below the assistant’s reply: adding or removing products from a list; enabling, disabling, or updating an alert rule’s name or threshold; subscribing to or unsubscribing from an alert rule; updating your KVI selection when it replaces a previous selection; adding, removing, or moving an item on a board; and reverting a MAP violation case or content case status change. Resizing a board item, and creating or renaming a board, do not offer an Undo chip. Click the chip to reverse the action. The chip shows a spinner while the reversal is in progress and changes to Undone on success.

Three confirmed actions also offer an undo on their result card: creating a list (“Delete this list”), creating an alert rule (“Delete this alert”), and adding a chart to a board (“Remove from board”). Editing a chart on a board can be undone too.

The Changes button in the panel header, shown as a clock icon with a count, lists the confirmed actions and undoable changes the assistant made in the conversation, newest first. Immediate writes with no undo, such as creating or renaming a board, resizing an item, or sending a test notification, do not appear. A row has an Open button when there is something to link to, and an Undo button while the change can still be reversed. Undoing from the list also marks the matching chip as Undone, and the other way round. The button is hidden when the conversation made no changes, and it is not shown for conversations shared with you.

Managing boards from chat

Ask the assistant to build out a board in your Ask workspace, a saved page of charts and data tiles, from any page you are chatting on:

• Create a board, e.g. “Create a new board called Pricing”. Takes effect immediately.
• Rename a board, e.g. “Rename my board to Q4 Pricing”. Takes effect immediately.
• Add a pre-built item: Key numbers, Price index, Price position trend, Out of stock, Assortment gaps, Price distribution, Shelf Score, Ratings and reviews, Content compliance, or Promotions, e.g. “Add the price distribution item to my board”. Takes effect immediately.
• Remove, move, or resize an item, e.g. “Remove the price index item from my board”. Takes effect immediately.
• Delete a board, or apply a template (Pricing overview or Stock and assortment) to create a new board from it. Both show a confirmation card first.

Adding, removing, or moving an item shows an Undo chip so you can reverse it immediately. To save a generated chart itself as a board item, see Saving a chart to a board above.

Proactive messages

If proactive messages are enabled for your workspace, certain events are pushed into the chat automatically, without you needing to ask anything. These Proactive Messages land in a dedicated conversation for the day, labelled with the date (for example “Brief · Fri 4 Sep”), which the assistant creates automatically the first time one arrives. Proactive alerts and the scheduled daily brief are each turned on separately, so your workspace may have one, both, or neither active.

The assistant maintains a persistent background connection even while the chat panel is closed. When a message arrives while the panel is closed, the launcher tab pulses with a colored ring and briefly expands into a wider pill labeled “AI Alert” or “AI Daily brief”. Dismiss it early with the × button, or let it collapse on its own. A numbered unread badge also appears on the tab. Open the panel to read the full message. The badge clears as soon as the panel opens, whether or not you have switched to that day’s conversation yet. On the Ask page, that conversation keeps its own unread marker in the conversation list until you open it. The chat panel’s switcher shows no unread marker.

Two kinds of Proactive Messages exist:

Proactive alerts

When an alert rule fires, the assistant delivers the event into that day’s conversation as an amber “Proactive alert” message. This arrives in addition to the rule’s configured email and Slack delivery. Those are the channels you configure on the rule itself. See Notifications for details on how alert events are delivered across channels.

When the Ask page is available, the message carries an Open in Ask chip. It continues the alert in its own conversation, which starts with the alert, a live view, and three suggested questions.

Daily brief

A sky-blue “Daily brief” message arrives once each morning, in your workspace’s timezone, summarising what changed since the previous day: the biggest competitor price movers, newly discovered products, MAP violations, and any alerts that fired. Even on a day with nothing notable to report, you still get a short brief noting it was a quiet day.

You can also request the daily brief on demand at any time by asking: “Give me my daily brief” or “What’s new today?”. The assistant returns the same summary inline in whichever conversation you’re in.

Some proactive messages include a quick “Was this useful?” prompt with thumbs-up/thumbs-down buttons. Your answer is recorded as feedback on that message, and the prompt is replaced with a thank-you note once you respond.

Perspective-aware behaviour

The assistant adapts to whether you are viewing the retailer or brand perspective:

• Retailer perspective. Competitor-focused tools are available: competitor price index, common product comparisons, price-vs-market comparisons, assortment-gap analysis, and competitor-undercut alert creation.
• Brand perspective. The retailers-reselling overview and per-product retailer distribution are available, showing which retailers carry a product, at what price, and their MAP status. Competitor-centric tools are not shown. Alert rule creation is available too, for MAP violation, price-decrease, and price-increase alerts. For a MAP violation alert, name one retailer to watch it alone, or leave it unset to watch every retailer reselling your products. Adding, changing, or removing a MAP price, working a MAP violation case, and sending a MAP compliance notice to a retailer are brand-only actions too.

MAP violation and MAP compliance questions work on both perspectives: a retailer asks about MAP on their own listings, a brand about the retailers reselling its products. Digital Shelf questions, including content compliance cases and content notices, also work on both perspectives.

Tools not applicable to the active perspective are removed automatically; the assistant will not offer or attempt them.

Chat memory

Tell the assistant to remember something, such as a preference, a standing fact, or a priority, and it saves a short note. The note can be up to 200 characters; it distills longer requests down to a single short sentence. It applies the note to every future conversation, even after you archive your conversations. For example: “Remember that I always want prices shown in EUR” or “Remember that our biggest competitor is [Competitor].” You can also ask it to forget something you no longer want it to remember.

When it saves a note, a “Saved to memory — manage” chip appears below its reply. Click the chip, or choose Chat memory from the header menu, to open the Chat memory menu. It lists every note you have saved, each with the date it was added. Click the × next to a note to forget it individually, or click Forget all. A second click confirms, since it clears everything at once.

You can save up to 25 notes. If you are already at the limit, the assistant will tell you and ask which existing note to forget or merge before it can save a new one.

Note

Archiving a conversation, described in Conversation history below, does not erase your saved memory notes. The two are independent.

Conversation history

The assistant keeps a running list of conversations, so different lines of inquiry don’t get tangled together. The panel header shows the current conversation’s title (or “New chat”). It doubles as a switcher: click it to see your other conversations and pick one, or click New chat to start a fresh conversation. New chat is available as a row in the switcher and as the new-chat button in the header. Selecting a conversation reloads its history automatically, and every conversation stays available across sessions.

To archive the conversation you’re currently viewing, open the header menu and choose Archive conversation. Archiving does not delete your messages; it just removes the conversation from the switcher and moves you to another one of your conversations, starting a new topic if none remain.

On the Ask page, the conversation list groups today’s daily brief under Today and other conversations under Recent. Each row has a More menu with Rename and Archive. An Archived toggle at the bottom of the list shows archived conversations, where Unarchive restores one.

If a response fails, an error message appears with a Retry button. Clicking it resends your last message exactly as typed without you having to retype it.

Sharing a conversation

Choose Share… from the header menu to share a conversation you own. You can share it with your whole workspace or with specific members. Shared conversations are read-only for the people you share with. Daily brief conversations can’t be shared, and only the owner can share or archive a conversation.

Conversations others have shared with you appear under Shared with me in the switcher, each marked “from” the owner’s name. When you open one, a bar at the bottom reads “Shared by [name] · Read-only” in place of the message input. Click Continue in my own thread to copy the conversation into one of your own and keep chatting there. If the owner stops sharing a conversation while you have it open, you see “This thread is no longer shared with you.”

Caution

The assistant can make mistakes. Always verify important numbers before acting on them.