Shared inbox
AI handles most conversations end to end; your team resolves, snoozes, or takes escalations.
The infrastructure your customer conversations run on
Most AI agents deflect tickets. Boom drives revenue and retention. It runs the conversation; you keep your logic and get the variables.
Shared InboxAI runs it, humans step in
CDPPeople, objects, events
WorkflowAgent
ExtractionTalk becomes columns
one agent · one growing contextcontext accumulates
The system
Four components, one connected loop
The inbox runs the conversation, data adds context, workflows pick the next step, and extraction structures the result.
Customer data platform
Customer data from Shopify, Skio, or your database, read live by the agent. No SQL.
Workflow builder
Initiatives give the agent an outcome; journeys map the triggers, sends, and branches around it.
Extraction
Pulls fields from conversations into typed columns you can sort and export to CSV or Excel.
Assigned to AI
Hi Mariana, happy to help. I can open the new account for you right here.
Great, what do you need from me?
Just confirm the email on file and I will take care of the rest.
How they connect
The inbox and the data sit side by side
The context panel opens the customer account next to the thread, and the agent reads the same data, so it answers with the order or balance instead of asking.
- Account context inside the chat
- Agent reads people, objects, events live
- Escalation pings the team
Win-back, past due1,284 matches
Personcountry = MX
Subscriptionstatus = past_due
Plantier = pro
Entering this segment enrolls a journeyJourney
From a segment to a conversation
Segments decide who gets reached, and when
Build a segment up to three relationships deep, no SQL. When a customer enters, the journey sends, waits, branches, or hands the objective to the agent.
142 conversations
longwideone-hot
| Customer | churn_reason | promise_date | plan |
|---|---|---|---|
| Mariana G. | price | 2026-07-03 | pro |
| Daniel R. | missing feature | none | starter |
| Priya S. | switched vendor | 2026-07-09 | pro |
| Tomás L. | price | 2026-07-12 | scale |
The conversation becomes data
Every closed conversation turns into a column
Extraction reads each conversation, including audio and images, and fills your typed schema: one row per customer on the Engagements dashboard, exported to CSV or Excel.
Offline eval gate70% threshold
Runtime scoringScored by AI
PIICardsCanaryDrift
Human feedbackthumbs train v+1
feedback trains the next version
ships only above thresholdsharper each version
Trust and safety
The agent is graded before and while it runs
Every change must pass an offline eval gate and regression baseline. Live, each conversation is scored for PII, card numbers, prompt leaks, and drift; team feedback trains the next version.
- Runtime scoring on every conversation
- RBAC, encrypted credentials, read-only replicas
- SOC 2, GDPR, ISO 27001, ISO 42001 in progress
FAQ
Questions teams ask first
Is this a bot builder?
No. You do not script the dialog. You give the agent an outcome and context; it navigates the conversation and improves from team feedback. Journeys handle the path around it: triggers, sends, delays, branches.
Do we keep our customer data and our logic?
Yes. Boom sits on top of what you have. It runs the conversation and extracts the variables, which flow back to you by export, so your domain logic stays yours.
Can we connect our own database?
Yes. Besides Shopify and Skio, Boom connects to a read-only copy of your database over an encrypted bastion, plus a REST events API. It reads on a schedule, so it is near real time.
Where does my team fit?
In the shared inbox. AI handles the volume; your team resolves, snoozes, or takes escalations, and a human must own a conversation to reply. Every agent change also clears an offline eval gate.
Explore each component
See the system on your own customers
In a 15-minute walkthrough, we run your use case through the inbox, data, and extraction.