Changelog

A batch of refinements from recent releases:

• The agent no longer stays silent after preparing a reply and loading a follow-up skill in the same turn, a case that could leave a customer without an answer.
• The agent now answers in the language the customer insists on, even after switching mid-conversation.
• A journey HTTP node now receives the conversation it is acting on, renders \n in a JSON body as real line breaks, and shows its variable chips more clearly.
• Journeys can count a customer's inbound documents and branch or respond on that count.
• The command menu now reaches the all-channels page.
• Inbox operators now land on the shared inbox right after signing in.
• A journey step no longer offers a business-hours window the engine does not honor.
• Brands can keep the Spanish opening question mark in outbound messages, with the option shown only where the mark applies.
• A tool call can no longer reach a customer as raw message text.
• The {{person.phone}} variable now resolves, and a blank value says so plainly instead of sending an empty token.

A journey can now branch on how a conversation closed, so what happens next follows the real outcome (the customer reached the goal, asked to stop, or went quiet). The close reason is saved as a step output you can reuse downstream and is reported on the journey_run.ended webhook, and the builder now lays out one branch per close reason for you instead of asking you to type each value by hand.

Boom now supports Instagram as a native channel: connect your account and the agent sends and receives Instagram direct messages through the same journeys and inbox as your other channels. Instagram and Messenger both appear on your channel settings overview, so every connection lives in one place.

A journey can now route a customer from one initiative into another partway through, so you can hand a contact from an onboarding flow into a research flow without starting over. The receiving initiative picks up the conversation in context.

A few refinements from this release:

• Journey messages can now format the variables they insert, showing a value as currency, a date, a running total, or a chosen item from a list, so a message reads exactly how you intend.
• Journeys can branch on how recently a contact last messaged on WhatsApp, which makes it simple to route around the messaging window.
• A journey step's outputs are now easier to find and reuse as variables in the steps that follow.
• A journey validation notice now shows its real message instead of an internal label.
• The toll-free SMS setup now collects the business details required to file and shows your submitted application in place of an empty form.

Journeys can now include a step that sends a free-text WhatsApp message you write yourself, not only a pre-approved template. The builder flags when the 24-hour messaging window is closed, so you always know when a template is still required.

A smaller refinement from this release:

• The channel settings pages now show one consistent, accurate status for each channel, so a setup step you have not reached yet no longer looks like an error and each channel reads clearly as ready or not at a glance.

A few smaller fixes and refinements from this release:

• Conversation handoffs now show who moved a conversation to the AI, naming the teammate who did it or marking it as automatic, instead of implying a person every time.
• The chat widget no longer fights the host page's layout on mobile, so it opens and scrolls cleanly on phones.
• When the agent reports how many records match, it now gives an exact count and flags when a list is only partial, instead of implying it showed everything.
• SMS messages now fill in variables like the recipient's name, respect a working opt-out (reply STOP), and support per-message emoji.
• The document dashboard now counts formats like screenshots and unreadable PDFs from the files themselves, so the totals match what customers actually sent.

One phone number can now serve several journeys at once. Keyword routing sends each incoming message to a different journey based on what the customer writes, so price can start your quoting flow while invoice goes to billing, with a catch-all for everything else.

A few smaller fixes and refinements from this release:

• The web widget on mobile no longer lets the page behind it scroll or shift, sits in the right place, and typing in it no longer lags on iPhone.
• Insights key quotes now show only what the customer actually said, instead of occasionally quoting the agent's own messages under a customer's name.
• Yes/no guiding questions now record a "no" as a no, so their charts reflect real answers instead of reading as all yes.
• The inbox timeline now names who handed a conversation to the AI, a teammate by name or the automatic routing, instead of a single generic label.
• The agent answers about your records more precisely: when someone asks how many contacts match or to find one by an identifier like an account number, it now counts exactly, says when a list is only partial, and can look a contact up by that identifier directly.

You can now reach customers over SMS straight from your journeys. A new send node composes the message and shows a live preview with the opt-out footer and the exact number of billable segments before anything goes out, so SMS reminders and campaigns can run right alongside your existing flows.

The web widget got a cleaner redesign: the visitor panel is now one tinted surface with the agent's replies sitting directly on it, and the launcher comes in two styles so it stands out on any site. The refresh also improves color contrast and fixes several iOS keyboard and scrolling issues.

A few smaller fixes and refinements from this release:

• Settings pages now use the breadcrumb as the page title, with a single consistent card underneath.
• Stopping an agent run no longer unassigns the conversation, so it stays with whoever was handling it.

A few smaller fixes and refinements from this release:

• Segments that sort by a number column now order correctly.
• Custom object types can now require that every search result carries key attributes, so the agent always has details like a product slug instead of guessing them.
• More web widget polish: correct styling for the components the panel renders, faster appearance updates, and a lighter panel payload.

The web widget now supports images in both directions. A visitor can attach a photo for the agent to read, and the agent or an operator can send one back, all natively inside the panel. Images are proxied through Boom so nothing loads from a third party, and the inbox composer gains an attach button for widget replies.

You can now use Edit on an active or paused initiative to update its objective, guiding questions, and agent instructions, no need to pause or relaunch it. The change reaches the very next conversation, and removing a guiding question that already collected answers stays blocked to protect your data.

The web widget now offers a compact panel height, capped around 560px, so it can sit on your site like a chat card instead of stretching the full height of the page. Pick it from the widget's appearance settings; the default stays full height, and the mobile layout is unaffected either way.

A few smaller fixes and refinements from this release:

• The web widget now renders links as clickable chips with a real label, instead of showing the raw URL.
• The web widget opens instantly on a visitor's first click, and we fixed an issue that could keep it from loading once installed on a live site.
• Each message of a multi-part reply in the web widget now shows as its own chat bubble instead of being merged together, and the widget keeps the conversation open across a single-page app's route changes.
• The web widget's composer now centers your text and highlights with your brand's accent color when focused, and clicking outside the panel closes it.
• When a customer opts out by texting a keyword like STOP, the conversation now shows an internal note explaining why and hands off to a person, instead of going silent.

A few smaller fixes and refinements from this release:

• A conversation held for a human handoff now resumes with the agent when a new journey starts, so it no longer stays silent when no one picked it up.
• Replies are no longer cut short by an unusual character in the response, so customers receive the full message.

A few smaller fixes and refinements from this release:

• The inbox now shows why an outbound message failed right in the conversation, so you no longer have to leave it to find out.
• A journey run that stops now surfaces in the inbox and in the run drawer, so a halted run is visible instead of silent.
• The inbox now renders automated sends as automated instead of labeling them as if a teammate sent them.
• When the agent cannot read an attached file, it now tells the customer why, not just that it failed.
• The template preview now fills in a template's own sample values instead of positional guesses, so the preview matches what customers actually receive.
• Template button links are now fit to their approved base, preventing a broken or duplicated destination URL.
• Attachment filenames no longer skew the conversation's detected language, so replies stay in the customer's language.

Campaigns now have their own section, separate from your recurring Initiatives: use one to reach a set of customers a single time, with a clear start and end, and get a report back. Your existing links keep working, so nothing you saved breaks.

A few smaller fixes and refinements from this release:

• Your webhook events now echo back your own event id on all three event types, so you can match a Boom event to the record on your side without keeping a translation table.
• The Webhooks settings show each destination as an environment matrix instead of a flat row, so it is clearer where each event actually goes.
• Settings, Webhooks now links straight to the public documentation.
• A webhook test ping that partially succeeds no longer reports as a total failure.
• A template inside an open 24 hour window no longer locks the inbox composer, so you can keep replying freely.
• Publishing a journey is now blocked when a send step uses a template that cannot be sent, catching the problem before it reaches customers.

You can now narrow the inbox to the conversations talked to on a given day. The date filter offers Today, Yesterday, the last 7 days, or a custom range, so finding the right conversation no longer means scrolling the whole list.

When the AI agent escalates a conversation to a person, Boom can now tell your systems in real time. A new conversation.escalated webhook fires with the conversation id and a direct link to the Shared Inbox, so your team can step in without watching the inbox all day.

Building a journey now starts with a clearer entry. The entry trigger reads as a card at the top of the canvas, and its title tells you at a glance which trigger starts the journey. You configure it in a panel right beside the card, so setting up how a journey begins stays in one place.

Your journey analytics now show where success happens. The successes chart breaks down by outcome node, so you can see which ending each conversion reached instead of one combined total.

A few smaller fixes and refinements from this release:

• The Publish button no longer shows on a draft that has never been published, so the first action you see matches what the draft can actually do.
• The analytics dashboards no longer get stranded when a custom date range is left without dates.
• Dark mode was rebuilt on a cleaner elevation ladder, so surfaces read with more consistent depth.

You can now filter the People directory by any attribute: pick a key from your organization's attribute catalog, type an exact value, and the list narrows to the people whose data carries it. This makes it easy to find a specific person by a field like an RFC, a tier, or a user type, even across very large directories.

A few smaller improvements.

• Conversations: improved reliability when a customer sends messages back to back, so the agent finishes its in-progress reply and still answers the next message instead of occasionally leaving it unanswered.

You can now duplicate an initiative right from the list, using its row menu. The copy lands in Draft with your full setup (context, questions, success metrics) but zero participants, so you can adjust it and launch without rebuilding from scratch.

Every picker in the journey builder (WhatsApp templates, sending numbers, segments, variables, and more) is now searchable, so you can jump straight to the option you need instead of scrolling through a long list.

The WhatsApp template preview now renders real formatting (bold, italics, strikethrough, monospace, and lists), so you see exactly what your customers will receive instead of raw text. Buttons also now show up reliably while you're editing a template.

A few smaller improvements.

• Journeys: fixed a bug where a message-sending step could skip resending a WhatsApp message on a retry, so customers now reliably receive it even after a retry.

Journey analytics now shows per message and per path attribution, so your success metric is tied to the specific step and route that produced it instead of only the initiative total. A new attribution view in Routes shows how much each step contributed, and a Paths and outcomes section traces the route through the journey that earned the result, so you can see what is actually converting rather than just the aggregate.

A few smaller improvements.

• Segments: the segments list is now paginated, so large workspaces load faster and stay responsive instead of rendering every row at once.
• Data and Variables: the Data and Variables sections are now available in every workspace, no longer limited to selected accounts, and the initiative sidebar is grouped into clearer sections.
• Agent context: you can now give your agent a richer briefing, with the initiative context limit raised to 20,000 characters.

A smaller improvement.

• Data export: the data view export now includes the participant's phone number, so contact details come through alongside your exported results.

When the AI agent escalates a conversation to a person, the inbox now shows an internal note summarizing the handoff right in the thread. Your team can see that the escalation happened and why, without digging through other tools.

Your messages can now pull data from related records like orders and fulfillments, so a WhatsApp message can include the tracking number, order status, or shipping details directly. Before, a journey could filter on that data but not print it, which left blanks in the message; now the value resolves across the full path from contact to order to fulfillment.

Segments can now filter on the absence of a relationship, not only its presence. Build audiences like people with no completed checkout or no order in the last month, which makes abandoned cart and win back campaigns straightforward, and a negated condition can now carry its own filters for finer targeting.

A few smaller improvements.

• Leave without saving: the prompt that appears when you leave a journey with unsaved changes now has a clear exit button, so it is obvious how to discard and go.
• Journey decision node: the decision step now renders and lets you edit every condition value correctly, including the shapes that previously showed blank.

Boom now ingests Shopify fulfillments and order transactions as first-class data. You can build segments and trigger journeys on what actually happened to an order: message a customer when their order ships, chase an expired or failed payment, or follow up after a refund. It works on your existing Shopify connection with no new permissions, and covers your stores right away.

The new Random Split node lets you A/B test inside a single journey. Add it to the canvas, give each branch a share of your audience, and send different openers, delays, or agent modes to each group without publishing separate journeys or splitting your list by hand. Every contact is assigned once and stays on their branch even if a step retries, so the comparison stays clean. Per-branch results are coming next.

Leaving the journey builder with unsaved edits now stops to ask before anything is lost, with a clear Save and leave or Keep editing. A stray click to the sidebar or a closed tab no longer discards a journey you were still working on.

A few smaller improvements.

• Reply counting: a reply the agent holds back is no longer recorded as a dropped message, so your delivery and reply stats reflect what actually happened.
• Links in the inbox: tracked links now resolve to their real destination on the path the inbox reads, so they open correctly when you click them.

The journey builder now supports undo and redo on the canvas. Press Ctrl+Z to step back and Ctrl+Y to step forward through your edits, so an accidental change or a deleted node is one keystroke away from being restored instead of rebuilt by hand.

A few smaller improvements.

• Test phone numbers: add a test number with a country selector so it is never malformed, and test numbers now expire and are managed from Settings.
• Initiative tester: the tester now shows what a customer actually receives (media, footer and buttons) and reports honestly how many steps of the journey it covered.
• Message formatting: cleaned up stray commas and flattened lists that could appear in some outbound messages.
• Tidy your journey canvas: large journeys now arrange into a clean, straight layout, and a new Tidy action re-flows the canvas whenever it gets messy.
• Invalid nodes flagged on the canvas: a misconfigured node, or a message node whose template variables are not set, is now marked right on the canvas instead of a generic banner, so you catch it before publishing.
• Clearer handling of files and documents: when the agent receives a file it cannot read, it now names the file type and size, and it reports how many pages a document has instead of guessing.

The new Message Logs page lists every message with its delivery status and the reason a send failed, and lets you search by recipient, so you can tell why something did not arrive without asking engineering. Delivery status is more accurate too, because out of order provider callbacks no longer roll a delivered message back to an earlier state.

You can now subscribe to outbound webhooks and be notified when a journey run starts, when it ends (with the reason, such as a node failure, a cancellation or a timeout), and when an enrollment is rejected. Webhook URLs support environment variables and resolve their own credential per environment, and you can send a test ping without launching a journey.

You can now export a conversation as a PDF from the shared inbox. Pick the time range you want, and the export leaves out internal agent notes so the document only shows the customer-facing exchange. It is handy when you need a clean, shareable record of a conversation, for example to show a payment processor that an exchange actually took place.

The agent no longer asks a customer again for something they already told it. It now reads the answers captured earlier in the conversation, so replies pick up where the customer left off instead of repeating questions.

You can now stop, pause, and resume a running campaign directly. Stopping retires a campaign so it takes no new contacts, without having to publish a replacement. Pause holds a campaign and resume picks it back up later. In every case, conversations already in progress finish on their own. The controls are available in the builder, on the initiative, and through the API.

A few smaller improvements from today.

• WhatsApp template headers: upload an image straight from your computer for a template's media header, instead of only pasting a URL.
• Journey Analytics speed: the analytics dashboard loads much faster on high-volume initiatives, with the same numbers.
• Navigation: the main menu is reorganized into clearer sections (Initiatives, Shared Inbox, Customer Data, and settings), with usage now at the top level.

Routes is a new view inside Journey Analytics that reads a journey as an outline instead of a diagram: one row per step, with branches indented under the split that creates them, run counts on each branch, and an estimate of the operational time each step saved. It also counts how many AI messages were sent at each step, and shares the version and date filters you already have set on the dashboard.

Segments now support is between, a new date operator that matches a sliding window, like more than 90 and less than 97 days ago, instead of a fixed calendar range. Because the window moves forward every day, a segment built on it can enroll the same person again and again on a set cadence, for example re-surveying customers every 90 days instead of only once.

Build cart recovery journeys for your Shopify store. The products a shopper left in their cart and your checkout domain are now available as journey variables, so you can message customers about the exact items they abandoned and send them straight back to checkout.

A few smaller improvements from today.

• Journey analytics polish: switch charts between counts and values, funnel step percentages are now accurate, and the KPI cards and their change indicators are easier to read.
• Large audiences: segments now evaluate reliably for very large contact bases and report an exact size instead of stopping at 100,000.
• Bulk transcript export: a new API endpoint returns all of an initiative's conversations in pages of up to 100.

Measure a journey with more than one success metric, each with its own name and aggregation (count, sum, average, or max). Attribution now starts from the first outbound message so replies across WhatsApp and email are counted correctly, a value metric can pull its amount from a linked record such as an order total, and you can filter results by journey version to compare performance before and after a change.

A few smaller improvements from today.

• API-launched initiatives: launching an initiative through the API now enrolls participants and starts sending, instead of leaving the journey in draft.
• Conversation wait status: a journey conversation node now shows why it is still waiting and when it will close, instead of an opaque running state.
• Done view in the shared inbox: the Done view no longer shows conversations that are no longer assigned to you.

A few smaller improvements from today.

• Conversation export in Data View: the export button in Data View now opens the full export menu (CSV, Excel, and transcripts) for your conversation data.
• Attachments in the sidekick: paste or drag and drop images and PDFs straight into the sidekick composer.
• Template preview in the sidekick: the sidekick's draft card now shows a real WhatsApp message preview instead of raw fields.
• Free-text inactivity timeout: set a journey conversation node's inactivity timeout as free text, with sensible minimum and maximum bounds.
• Conversations in the initiative sidebar: the Conversations view is now reachable directly from the initiative sidebar.

Run one journey against different environments. Set per-environment values with {{env.*}} (for example a different baseUrl for staging and production), and have an HTTP Request node authenticate with the secret that matches the journey's pinned environment. Test a journey against staging, then switch it to production, without duplicating the journey or editing each node.

A few smaller fixes from today.

• Template creation: creating a template no longer fails with a "Couldn't apply" error when a placeholder is missing an example value.
• Cleaner agent messages: fixed cases where line breaks showed as raw <br> and where list bullets could be dropped from a reply.
• No cut-off replies: agent responses no longer get truncated at the end of a list when handed off between steps.

A couple of smaller fixes from today.

• Campaign funnel counts every reply: the Replied stage now includes replies from conversation steps, not only steps that explicitly wait for a reply, so the funnel reflects real engagement instead of showing zero when customers did respond.
• Send Message variable picker: choosing First name in a message template no longer also selects the full customer name, so each placeholder maps to exactly the field you picked.

Every initiative now has a Journey Analytics dashboard that shows how a campaign is doing from end to end: a reach to reply funnel for each channel, KPI cards for messages sent and delivered, and a configurable success metric so you can define what counts as a win (a sale, a booking) and then see the success rate, the value generated, and the evidence behind each one. You can also open the journey map to read delivered, read, and reply counts on every node.

A few smaller improvements from today.

• Segment by when a contact was created: a contact's creation date is now a date attribute you can filter on in segments, so you can target people added in the last 30 days (or any window) from the builder, the API, and the sidekick.
• Clearer warning when a tested response has many fields: the Send Message, Send Email, and Dispatch field pickers now tell you when a tested body's field list is incomplete, instead of implying that a missing field does not exist.
• Fixed the Configure email link on journeys: the "Configure email" button on a Send Email readiness step now opens the right settings page instead of a broken link.

A few smaller improvements from today.

• Pick nested response fields in more journey nodes: the Send Message, Send Email, and Dispatch pickers can now bind to a specific field inside an HTTP node's response, not just the whole body, matching what the Decision and Case nodes already offered.
• Message variables when authoring by API or agent: building a journey through the API or the agent now exposes the same catalog of message template variables the visual picker uses, and publishing is blocked when a Send Message step points at a variable that would not resolve, so messages no longer go out with a blank spot.
• Renamed contacts show everywhere: renaming a contact in the inbox now updates their name across the list, header, search, and live updates, instead of only in the context panel.
• Add identity stays available in the inbox: the option to add an identity no longer disappears after the first one when a contact has only linked profiles, so you can keep attaching identities.

Journeys can now forward documents a customer sends over WhatsApp (IDs, payment receipts) to your own systems. The new engagement.documents variable resolves to secure, time-limited file links inside an HTTP Request node, so a journey can hand those files straight to your API. The agent also detects password-protected PDFs so they are handled cleanly.

The data your agent extracts from conversations can now be used as journey variables. Reference an extracted field inside a journey (in a condition, a message, or an HTTP request) and the value flows through automatically, so what a customer tells the agent can drive what happens next.

Operators can now add extra phone numbers and emails to a contact directly from the shared inbox, so a customer who writes in from a second number still resolves to the same person. Contact matching also got more reliable for numbers written with or without the extra Mexico country-code digit.

A few smaller improvements from today.

• Organization switcher in the shared inbox: operators who work across several organizations can now switch between them right from the inbox sidebar.
• Preview your sources: data sources now open in a preview modal and show up in the sidebar, so you can check what you connected without leaving the page.
• Easier HTTP node authoring: the journey HTTP Request node now auto-quotes JSON values, validates the body inline, and ships with sample payloads.
• Simpler edge connections in the builder: connecting two journey nodes now snaps to the nearest handle, so you no longer have to hit the exact dot.
• Cleaner add-data flow: the add-data wizard waits until you save before creating anything, so abandoned setups no longer leave stray resources behind.
• Publish guardrail for extracted variables: publishing a journey now warns you when a step uses an extracted field that has no conversation to fill it, before it comes back empty.

When a customer sends a PDF over WhatsApp, like a bank statement or a form, the Concierge now reads its text and can use it in the conversation and in your reports. If the file is password-protected, the agent recognizes that instead of treating it as unreadable, so it can ask for an unlocked copy.

The Sources page now covers every integration, your database, Shopify, and Skio, in one place, so you can view and edit your own sync queries without looping in a Boom engineer.

A few smaller improvements from the last few days.

• Higher API rate limits: public API keys can now make up to 1,000 requests per minute, up from 100, so heavier integrations stop hitting the ceiling.
• Accurate segment member counts: segment lists and detail pages now show the real, live number of active members instead of a count that could lag behind.
• Fully translated journey builder: labels, toolbar actions, and validation messages in the journey builder now follow your language instead of showing Spanish regardless of locale.
• Journey run history over the API: you can now read a contact's step-by-step journey run, which nodes ran and whether they succeeded, over the API and MCP.

The new email builder lets you design branded emails visually: drag in logos, headings, text, buttons, images, dividers, and footers, insert {{variables}} inline, and manage your own image library. Emails render with email-safe HTML so they look right in every inbox.

The extraction Data View got more analytical: numeric fields now show total, median, and range, you can export the data to CSV, and you can click a quote to open the exact conversation it came from. It is easier to trust and act on what the agent extracted.

Segments can now update reactively: when a person's data or events change, their segment membership updates right away instead of waiting for a scheduled sweep. Journeys triggered by a segment enroll people the moment they qualify.

Journeys now include a Send Email node, so an automated flow can reach someone by email the same way it does over WhatsApp. Point the node at a published email template and it sends with each recipient's real data.

A few smaller improvements from the last few days.

• Extraction schema over the API: you can now read and set an initiative's data-extraction schema programmatically over the API and MCP, not just field by field in the builder.
• Attribute distributions in the sidekick: the operator sidekick can summarize how a person attribute is distributed across your audience, with personal data kept safe.

From the contacts table you can now download your contacts as CSV or Excel, either the rows you have selected or every contact matching the current view and search. The export always includes every extraction field, even columns hidden by the column picker, so you get the complete data no matter how the table is arranged.

You can now author a complete journey without touching the builder canvas. The new write surface lets you create, edit, validate, and publish an entire journey (every node type and every connection) from the public API, MCP, and the in-app Sidekick. Build a flow by describing it in chat, fork a live journey into a draft, or manage your journeys straight from your own tools.

A few smaller improvements from today.

• Multi-select list fields: extraction schemas now support a closed-list field that accepts several values, so answers like abandonment reasons stay consistent instead of drifting into free text.
• Cleaner responses view: the Respuestas tab groups its data sources into a single stacked pile and lays its charts out in an adaptive two-column grid.
• CSV import preview now stays neatly inside its dialog instead of overflowing the table.
• Friendlier empty state: the Initiatives screen shows an animated illustration when there is nothing yet.
• Clearer errors when a Sidekick draft cannot be saved, now showing the real reason instead of a generic message.

The journey builder got a visual refresh: a cleaner canvas with clearer node cards and more helpful editing affordances. Building an automated flow (enrollment, send windows, messages, wait for reply branches, and AI managed conversation steps) is now easier to read and lay out.

You can now do more than read from the public REST API and MCP server. Launch research Initiatives, create and manage Segments, and add, list, and stop Participants, including pulling a participant's full conversation transcript. Journeys are now readable through the API too. Every action runs on the same definition behind the API, the MCP tools, and the product, so your own systems and agents stay consistent with what you see in the app.

A few smaller improvements from this week:

• Abandoned cart reach: contacts who have a phone number but were showing as unreachable are now correctly matched to WhatsApp, so more of your abandoned cart customers actually get your messages.

The visual journey builder now splits waiting and talking into two dedicated blocks. Wait for reply pauses a journey until the customer responds, and Handle conversation is a first-class card your AI agent uses to hold a back and forth toward a goal. You can expand and edit a block in place and drag it to reorder, so building a multi-step flow is clearer and faster.

Creating an initiative is now a guided four-step wizard: set it up, configure the agent, choose what data to extract, and shape the journey. Each step keeps the choices in front of you, so you can launch a research or outreach initiative with confidence and step back to adjust anything before you go live.

A few smaller improvements from the last couple of days.

• Editable customer details: edit a customer's name right in the inbox, and your change overrides the synced data everywhere the agent uses it.
• Simpler WhatsApp templates: message templates are now managed with just your phone number, no technical account identifier needed.
• WhatsApp location shares now render as a shared location instead of a blank bubble, and your AI agent can read them too.
• Richer CSV uploads: every column in an uploaded CSV is captured as a customer attribute, and the upload wizard asks for an ID column so contacts match cleanly.
• Clearer journey conditions: boolean decision steps now use a simple True or False selector, and an unset condition is caught before you publish.
• The dashboard now follows your light or dark theme.

The journey builder now uses clearer building blocks so it is easier to see what each step does. The old Conversation step becomes Wait for reply, a new Handle conversation block lets you choose whether AI or a human takes over, and the confusing send-immediately toggle is replaced by an explicit send-window step. Building an outbound journey now reads the way it runs.

The Metrics dashboard has a refreshed layout with a new funnel overview, softer sentiment colors, and clearer labels, so your results are easier to read at a glance. Response rate and the conversation funnel now count real inbound replies, so the numbers reflect what actually happened.

The metrics view got a rework that makes results easier to read at a glance. A new funnel shows how contacts move through each stage, the response rate and funnel are now based on real inbound messages, and the layout, colors, and labels are cleaner so you can spot what is working faster.

A few smaller improvements from today.

• Contact uploads now capture every column in your file as a person attribute, so no data from your upload is left behind.
• The contact upload wizard now asks for an ID column, so records reconcile cleanly instead of creating duplicates.
• Journey decision conditions now use a clear True/False selector and can no longer be published while left unset.

You can now maintain a Do Not Contact list per organization from the new dashboard under People, so opted-out contacts are never messaged again by mistake. A contact is added automatically when they reply STOP, use the one-click unsubscribe link in an email, or is added manually, and every outreach campaign, journey message, and automated follow-up checks the list before sending. A team member can still send a message by hand when it is truly needed, with a clear warning and a permanent audit log, and the dashboard now supports search and pagination to keep long lists easy to manage.

A few smaller improvements from today.

• WhatsApp replies from your AI agent now arrive faster, thanks to a streamlined reply pipeline.
• The segments builder is denser and easier to scan, with a cleaner header and no more horizontal scrolling on smaller screens.
• Your AI agent can now see images customers send in a conversation, not just read the text, so it responds with the full picture in mind.

A few smaller improvements from today.

• Journeys can now branch on any event property, not just the reserved set, so decision and case steps react to the full detail of your custom events.
• Reports now hide cards that have no data (like reward eligibility when no reward is configured), so the view stays clean and shows only what is relevant.
• The journey builder node editor is now fully localized, so the controls read in your language throughout.
• Revenue figures are more accurate: discounts that are already used up no longer count against net amount.

You can now add a Case node in the journey builder to route people down many paths based on a single attribute value, instead of chaining several decision steps. Each branch has its own outcome and there is a required default, so everyone lands somewhere and no one silently falls through.

You can now configure your agent per channel from a new settings screen: pick a default agent, toggle AI on or off, and set the namespace for each channel independently, with a fallback to your organization defaults. WhatsApp and web chat can behave differently without extra setup.

The public REST API and MCP server can now create and manage your research Initiatives, not just read them: create a draft, update it, cancel or archive it, and link a WhatsApp outreach template to an initiative. You can also create, list, and check the approval status of WhatsApp templates, so you can run campaigns end to end from your own systems.

The in-app assistant can now draft a complete email template from a simple chat prompt, filling in the copy and the right customer variables for you. Describe the email you want and start from a working draft instead of a blank canvas.

A few smaller improvements from today.

• Email templates now show variable chips consistently everywhere, and a bug that sent a raw token instead of the value is fixed.
• You can now publish or unpublish an email template, and the send-test button shows a loading state so it no longer looks stuck.
• New themed starter templates give you image-forward, styled emails to begin from, with variables available in the subject line.
• The engagement detail view has a new Data tab, so you can see a person's context and workflow state at a glance.
• Funnel and coverage numbers in reports now use honest denominators and rounded, attributed cards, so the figures you show match reality.

You can now author on-brand email templates in a rich editor: write your copy, drop in customer variables from a slash menu, and start from a gallery of ready-made templates (welcome, onboarding, reactivation, and more). Send a test to yourself, then send for real from your own verified domain, so email becomes a fully operated channel alongside WhatsApp.

In the visual Journey Builder, WhatsApp message steps can now pull fields from your custom events as variables. When a journey is triggered by an event, pick its properties (like plan_tier or order_total) from a dropdown and drop them straight into your message, so each send is personalized with the data that started the flow.

The public REST API and MCP server now read your research Initiatives and the data they collect: list and get initiatives, participants, and their extracted variables, plus a per-initiative summary of engagement and results. Personal and internal fields stay hidden, so your own systems and agents can pull insights safely.

A few smaller improvements from this week.

• You can now reference context files inside an initiative's context block with an @ mention, and they render as inline chips so it's clear which files the agent is using.
• The initiative detail view now shows attached assets as chips, so you can see what's linked at a glance.
• The initiative tester can now impersonate pre-launch enrollment requests, so you can preview exactly how an enrollment will behave before you go live.
• The initiative tester opener no longer invents a placeholder name, so test conversations start with a clean, realistic first message.
• The agent no longer shows internal tool preamble in its replies, keeping customer-facing messages clean.

A few smaller improvements from this week.

• Archived templates are now viewable: a new "Show archived" toggle on the templates list lets you find and review templates you previously removed.
• WhatsApp templates now submit for approval automatically when you create them, so you no longer need a separate manual step.
• The initiative tester now opens with a preview of the journey's real WhatsApp template instead of a generated message, so what you test matches what customers receive.
• The journey builder no longer discards unsaved changes when you switch browser tabs and come back, protecting work in progress on the canvas.
• Survey re-extraction is unblocked: initiatives without a fixed schema can be re-analyzed, so you can refresh survey results after making changes.

You can now enroll people into a journey by picking them manually from your CDP or by taking a snapshot of an entire segment, with no CSV upload required. Each source can be added as pending or start the journey right away.

Journeys are now available to every organization, no longer behind a preview flag. Build automated, multi-step flows over WhatsApp in the visual builder: branch on customer data with decision nodes, wait, send messages, run a conversation, and dispatch events, with each path ending in its own outcome. Access is role-based: admins have full control, and other roles get read-only access.

Message templates that attach a PDF now show the first page of the document in the WhatsApp media preview, instead of a generic file icon. You can confirm you attached the right document before sending.

A new public REST API lets you create and preview segments and read your CDP catalog (relationships, custom object types, and relationship types) from your own systems. The same data is available through an MCP server, so your own agents can query Boom directly.

A handful of smaller improvements across segments, journeys, and templates:

• Personalize messages with computed values: calculated segment values, such as a summed subscription amount, can now be used as variables in journey messages.
• Import people by CSV: add people to the CDP People tab through a CSV upload or one by one.
• Cleaner template management: Boom now creates fewer duplicate template variations per round, and follow-ups reuse approved copy.
• Smoother CSV enrollment: the add-contacts wizard handles column mapping and large files more reliably.
• Accurate relative times: recently created records no longer show timestamps in the future.
• Tidier initiative assets: the assets inherited from your organization now start collapsed.

When you create an initiative, Boom now builds the first draft of its journey and outreach messages automatically in the background, including a message for each follow-up round, reusing your approved templates when they fit. Starting a new campaign goes from a blank canvas to a ready-to-review draft.

You can now set how far the agent digs in on each guiding question, from no follow-up to a deep, multi-turn probe. Use lighter follow-ups for quick survey-style questions and deeper ones where you want richer, more detailed answers. This gives you finer control over the length and depth of every interview.

The Wait step in journeys can now hold a contact until a specific date and time, or until the next weekday you choose (for example, the next Monday at 9 AM) in your organization timezone. Before, a wait could only pause for a relative amount of time. Now you can align messages with launches, deadlines, or business hours.

A few smaller upgrades:

• A zoomable image viewer in conversations, WhatsApp style, with pinch, scroll, and double-click zoom plus thumbnails.
• Click to copy any value in the conversation context panel.
• Starting a New message is no longer blocked when the contact already has a conversation.
• Connect a database through an SSH bastion (jump host), for data behind a private network.

You can now build a segment from a library of common templates (for Shopify, Skio, and connected databases) instead of starting from a blank filter. Pick a template, adjust it, and your audience is ready. Segments are also now reachable over the public API.

Creating an initiative is now just an objective and a bit of context, and you are done. The long multi-field setup form is gone, so you can launch a research or outreach initiative in a fraction of the time.

A handful of smaller fixes across the inbox and engagement views:

• The engagement timeline now shows which branch a decision took and how long a delay lasted, in English, Spanish, and Portuguese.
• The conversation list filter can now include archived initiatives.
• Runs lists show a contact's email instead of a raw internal id.
• You can only reply to a conversation that is assigned to you, so replies no longer get lost in the unassigned view.
• A cleaner events table with corrected relative timestamps.

The Shared Inbox is now on for every team on your plan, with no setup required. Anyone with inbox access can see and reply to conversations across WhatsApp and your other channels from one place.

WhatsApp template buttons can now carry personalized link variables, so each recipient gets their own URL (a unique reward link, a tailored checkout) resolved at send time. Template header images are also saved now, so they appear in the shared inbox instead of showing as text only.

The inbox got a round of polish: resolved conversations now look visually distinct, the All view shows only open conversations with a count badge, actions update the list instantly, and the context panel surfaces a contact's channel and identity even when there is no match in your data yet.

Journey messages can now pull in variables from your data and from the events that triggered them, so each customer gets a message written for their context. Picking the event that starts a journey is clearer too, with a branded event picker.

A new People directory lets you browse everyone in your customer base in one paginated table, then open any profile to see their attributes, events, and conversations across tabs. Find the right person and understand them at a glance.

Segments can now use computed attributes like number of orders or total spent, calculated for you from each person's history, plus match on list and tag values. Build audiences around behavior, not just static fields.

You can now connect your Skio subscriptions to Boom from the Integrations page, using a self-serve API key. Bring your subscription data in so your segments and journeys can act on it.

Journeys got three new controls: a frequency cap so a customer never gets too many messages, a node that fires an event into the CDP as the journey runs, and the ability to remove a customer from a journey the moment a chosen event happens.

Boom now supports more than one WhatsApp business account at once. Choose which number sends each step of a journey, and templates stay scoped to the account they belong to across journeys and the inbox, so the right sender and the right templates always line up.

Segments can now filter across related records and reach into nested attributes. Build a segment that follows the link from a customer to their orders, or matches on a value buried inside a JSON field, all from the same builder.

Describe the audience you want in plain language and the agent builds the segment for you, turning your request into the right conditions, including links to related records and nested fields. No manual filter setup required.

A new analytics dashboard shows how your conversations are going, with volume over time, response times, an hourly activity heatmap, and an operator leaderboard. Filter by the dimensions that matter and see where to focus.

Boom is now available end-to-end in Portuguese (pt-BR), both the operator dashboard and the agent itself. Set Portuguese as your workspace's default language and run conversations natively in Brazil.

The shared inbox got a rework. Every conversation row now shows which channel and number it came from, a new filter bar lets you slice by channel, and search was rebuilt to scan across all sessions of a conversation, returning results in milliseconds instead of seconds.

You can now configure exactly what your operators see next to each conversation: customer attributes, related records like orders or loans, and deep links into your own systems. Set it up once and every agent gets the context they need to resolve faster.

Save your team's most common responses and drop them into any conversation with a / shortcut or a button. Edit before sending, and search across all of them. Quick replies are separate from WhatsApp templates, so they're yours to manage.

The new segments builder lets you combine conditions with AND/OR in a simple three-step flow. Schedule a segment to refresh daily at the hour you choose, and export the full membership to CSV.

Boom can now read your customer data straight from your own database, with no CSVs and no middle layer. Model real-world relationships like one buyer across many merchants, push behavioral events such as a completed payment, and segment on what your customers actually do. A documented public API keeps everything in sync with your systems.

A journey can now enroll a customer the instant something happens, like a completed order or a churn-risk signal, straight from a CDP event. There's no CSV and no waiting for the next segment refresh. The right message goes out in the moment.

You can now chat with your agent right inside the initiative. It's a live test that runs the real pipeline and streams responses as they generate, so you can try multi-turn flows and edge cases before any customer sees them.

Connect your Shopify store in a few clicks and Boom keeps customers, orders, and your full product catalog continuously in sync. The agent can reference real products in the conversation, with their images and descriptions.

Leave internal notes in a conversation thread to pass context to a teammate, drop a reminder, or note a handoff detail. Notes stay private to your team and never reach the customer.

Upload your documents, like product sheets, policies, and FAQs, and the agent searches them to answer customer questions. It pulls the right passage from the right file, so replies stay grounded in your own content.

Boom now runs on WhatsApp, Instagram, Facebook, web chat, and email (beta), with support for multiple WhatsApp numbers per workspace. Every channel lands in one shared inbox, and the agent answers with the same customer context no matter where the conversation starts.

Beyond downloading a single conversation, you can now export the transcripts of many conversations at once, as a single consolidated .txt you can hand off or analyze in bulk.

Your agent can now send images, PDFs, and video in WhatsApp conversations, not only text. It shares receipts, product sheets, or onboarding docs right inside the chat, picking the right file for the moment.

Contacts is now a dedicated section in the initiative sidebar, easier to reach, with a unified view of engagement status, filters, and bulk actions to start journeys.

You can now download the full transcript of any conversation, as plain text (.txt) for quick sharing, or as JSON (.json) to feed into your own analysis.

We deployed a new version of Boom Voice. It sounds dramatically more natural, perfect for Research campaigns running over voice.

You can now run Boom Surveys: shorter, focused conversations to ask your customers a specific question and get fast answers.

Use Surveys when a full Research interview is more than the moment calls for: a quick pulse after a release, a single follow-up question, or a targeted check-in with a segment.

We built an agent at Boom that controls the entire platform.

It can create research initiatives. Change settings. Add context about your company. All through a conversation.

Read more →

Define your research objective. Add the guiding questions you want answered. Boom does the rest.

Boom Research runs personalized, context-aware customer interviews at scale over WhatsApp, email, and voice, turning what used to take weeks of recruiting and scheduling into a 2-day workflow. Each conversation feels human, follows the thread your customer opens up, and stays focused on the questions that matter to you.

When the conversations are done, Boom synthesizes every transcript into an actionable insight report, so you go from "let's talk to users" to "here's what we learned" in days, not quarters.