Chord Release Notes: May 20 – June 1, 2026

This release focuses on making Chord AI feel more like a feedback loop than a one-shot answer service: users can now react to individual answers with praise or critique, see what the assistant has learned about their data, and trust that follow-up questions are grounded in the actual rows from the prior turn. Chord CDP gets richer connection-level metadata and filtering, and the Chord platform picks up automated status alerts.

New features

Chord AI

Inline praise & critique reactions

Users can now react to individual SQL answers with a praise or critique directly in the thread as a message. The assistant recognizes the reaction, classifies it as praise or critique, and saves it against the specific answer. Vague critiques ("this is wrong") now trigger an inviting follow-up asking for specifics, while specific critiques anchored to a metric or value get an acknowledgment.

"Saved to Copilot memory" chip

When users drop a rule mid-conversation and the assistant captures it to memory, the bot bubble now shows a "Saved to Copilot memory" caption so it's obvious the rule was retained.

Chord AI Slack bot

App Home tab

The Chord AI App Home tab is now wired up: linked users see a personalized greeting confirming which tenant they're connected to and a one-click button to switch tenants. The connected tenant name doubles as a deep link into that tenant in Hub. Users who haven't connected yet see a Connect Account call-to-action.

Recent conversations on App Home

The App Home tab now lists your most recent threads with the bot, each row deep-linking back to the original Slack thread. The list is scoped per tenant, so switching tenants refreshes the view to show only that tenant's history. First-time users see an empty-state prompt to ask their first question.

"Open in Hub" button scoped to direct messages

The "Open in Hub" deep-link button on Slack bot answers now only appears in 1:1 DMs. In shared channels and group DMs the button is hidden, since Slack renders identical message blocks to every viewer and the Hub thread is scoped to the original asker. "Show reasoning" and "Show SQL" continue to appear everywhere.

Full support for group DMs

The bot now works end-to-end in multi-person direct messages — clarifier replies, non-owner follow-ups, and continuations in a group DM behave the same as in channels. Previously the first @mention reached the bot but follow-up messages silently failed, causing conversations to stall.

One-click trace lookup for triage

A new "Get trace ID" message shortcut lets Chord engineers fetch the diagnostic trace for any Chord AI answer directly from Slack, so customer-reported answer-quality issues can be triaged without back-and-forth. The trace returns as an ephemeral reply visible only to the requester.

Praise & critique reactions reach Hub feedback

When a user replies in a Slack thread with praise or critique, the Slack bot now auto-submits that reaction through the same feedback chain as the Hub Copilot UI — landing a thumbs up/down on the specific answer, a row in the Hub Copilot feedback admin view, and kicking off the feedback evaluation pipeline. The bot also responds with a tailored "Noted your positive/negative feedback" acknowledgement instead of a generic caption.

Find additional Chord AI Slack bot documentation here!

CDP

Site and destination tag badges

Environment (Production / Staging), Direction (Front-end / Back-end), and Delivery Mode (Cloud / Device) now render as compact badges on site cards, destination cards, and the Connections table — making it easy to scan a long list and pick out what you're looking for.

Tag-based filtering on the Connections page

The Connections page Status dropdown is replaced by two labelled filter buttons — Sources and Destinations — each opening a popover with checkbox groups for environment, direction, delivery mode, and connection state. Combine filters across sides for AND logic across sources and destinations.

Bing Ads Consent Mode

The Bing Ads destination configuration now includes an Ad Consent Category field, bringing it in line with the Google Ads and GA4 consent-mode fields shipped last release.

Platform

Recharge API token validation

Saving or testing a Recharge data source now validates the API token's scopes up front and surfaces specific missing-scope errors, so connection failures are diagnosable without guesswork.

Platform status alert snackbar

A new bottom-right status snackbar surfaces unresolved incidents and active maintenance from status.chord.co automatically, with severity-tiered styling (maintenance, minor, major, critical) and per-incident dismissal that re-surfaces when the status page posts an update. Operates alongside the existing top-of-page system banner for manual announcements.

Improvements

Chord AI

Follow-ups grounded in prior result rows

Follow-up questions now receive a sample of the actual rows from the prior question's answer. Asking "why is March lower than the others?" produces grounded analysis referencing the real values from the prior table instead of speculative guesses. When the prior query legitimately returned zero rows, the assistant now says so explicitly rather than rendering a confusing empty answer.

Current-date anchor in follow-ups and clarifications

Follow-up and off-topic responses now resolve relative dates against today instead of the model's training cutoff. Questions like "what day is today?" or references to specific months no longer get treated as future dates.

Routes sync questions to Chord Data Activations

When users ask about syncing a segment to a destination like Klaviyo, the assistant now points to Chord's native Data Activations feature instead of recommending external reverse-ETL tools.

Longer answers without truncation

The output cap for assistant responses has been raised significantly, so long-form analyses and detailed SQL reasoning no longer get silently truncated mid-thought.

User chat bubbles preserve line breaks

Multi-line questions typed into Copilot now render with their original line breaks instead of collapsing into a single paragraph.

CDP

Device-mode destinations flagged in Logs dropdown

Device-mode destinations (which fire from the browser and don't flow through the function pipeline) are now greyed out with a "Device Mode" tooltip in the API Destinations & Functions Logs dropdown, so users stop selecting them only to find an empty events table.

Compact environment & direction labels in selects

Site, connection, source, and destination dropdowns now show inline meta text (e.g., "Prod · Front-end") next to each name, making the right pick obvious without opening each one. The Functions Logs toolbar is also reorganized for clarity.

Fixes

• A bug that wiped the typed reason on inline-feedback praise has been fixed; the message now persists alongside the praise reaction.

• Fixed broken layout on Pinned Reports and Quickstart cards: pin button is back in the top-right corner, category chips no longer get clipped, and confirmation dialog buttons have visible spacing.

• Chord-built reports can now be unpinned directly from the Pinned Reports grid, matching the behavior already present on the report detail page.

• Fixed an issue where pages briefly showed a stray horizontal scrollbar at the bottom of the viewport.