Breakout

Language models do the selling work: the inbound agent answers product questions from the customer's knowledge base, qualifies against the ideal profile and books meetings, while the outbound agent researches accounts, writes follow ups grounded in the triggering signal and handles replies. A real layer works without them, since the Signals plan sells visitor identification, waterfall enrichment, buying group mapping and alerts on their own. Responses come from OpenAI and Anthropic models.

The inbound agent converses, qualifies and books in real time with no review step, by design, and the outbound agent sorts replies, responds and books meetings on its own. Oversight shows up in a few places: Spoc builds agents from a brief that the customer approves before anything ships, outbound messages leave through the customer's own sequencers, and a human in the loop is part of the stated AI posture. Which actions need approval and how a misbehaving agent is paused are not described.

OpenAI and Anthropic are named as the large language model providers, customer prompts and conversations go to them under contracts that forbid training, and an AI subprocessor list sits in the trust center beside an AI posture summary. Scoring against the ideal profile and the signal model are described by inputs rather than logic. Whether the website agent tells visitors it is software, and whether outbound copy is marked as AI written, is not addressed.

Customer stories from Ketch, Beautiful.ai, HackerEarth, Barti, TeamOhana and Recruit CRM describe switching from Drift or Intercom and building an SDR motion without hiring SDRs, told through quotes rather than measured results. Headline counters cover companies identified, identification accuracy, open rates and speed to lead. Meetings booked, pipeline created or conversion rates for a named customer, with a period and baseline, are not given.

Person level identification runs only where visitors have consented through the customer's own banner or consent mechanism, and the customer's privacy notice and lawful basis govern it. Outbound email and LinkedIn messages go out through the customer's sequencers, so unsubscribe handling and sending rules sit with those tools and with the customer. CAN SPAM, GDPR marketing consent and LinkedIn messaging rules for the outbound agent are not addressed by name.

Breakout acts as processor for customer service data, never sells it, and names its lawful bases for its own processing, contract, legitimate interests, consent where required and legal obligation. Transfers from Europe, the UK and Switzerland rely on standard contractual clauses, service data is deleted within 90 days of contract end, and California rights are listed. A signable DPA is not public, and the trust center shows only part of the subprocessor list without access.

Company and person matches for website visitors come from unnamed third party data providers, and waterfall enrichment adds contact details from further providers in sequence. More than 700 outside signals, from hiring and funding to technology changes and job moves, feed the outbound agent without their sources being named. Licensing terms, supplier names and any warranty on the data are not given.

Identification runs from a script on the customer's own website, CRM and tool connections use their native interfaces, and LinkedIn follow ups are sent through the customer's existing sequencers rather than by Breakout logging into accounts. The open point is that LinkedIn sending depends on whichever sequencer the customer uses, and Breakout takes no position on how that sequencer handles LinkedIn's automation rules. Social feeds from LinkedIn and X appear as embedded content on the customer's site.

Service data is never used to train Breakout's own models or those of its AI subprocessors, and OpenAI and Anthropic receive prompts and conversations under contracts that prohibit training on them. Data is used only on the customer's instructions and never for Breakout's own purposes. Service data is deleted within 90 days after the contract ends, or sooner on request.

Website visitors talk with an AI agent that opens with context about their company and history with the seller, and outbound prospects receive emails and LinkedIn messages written in the customer's voice. Whether the agent introduces itself as software, and whether outbound copy carries any AI label, is not stated, and Breakout takes no position on Article 50 of the EU AI Act. Nothing in the marketing sells passing as human.

Visitors are matched against CRM records to surface open opportunities, closed lost deals, customers and account owners, and high value visitors route to the owning rep in real time with Slack and email alerts. More than 55 GTM tools connect for reading, writing and actions, Clay and Attio among them, with HubSpot in the vendor stack and custom integrations on Enterprise. Sync direction, field mapping and an open API are not described.

Breakout runs as a hosted service on infrastructure in the United States, and data from Europe, the UK and Switzerland moves there under standard contractual clauses. An EU processing region is claimed without a named location, provider or way for a customer to choose it. Hosting providers and tenancy are not described beyond that.

Breakout holds a SOC 2 Type II attestation covering security, availability and confidentiality, tested over a 12 month window by an independent CPA firm. A trust center run on Sprinto lists controls across product, data, network, endpoint and corporate security, with 37 policies and procedures available on request. Data is encrypted with TLS in transit and AES 256 at rest, with role based access and multifactor authentication, and SSO comes with Enterprise.

Signals starts at $500 a month for visitor identification, enrichment and buying group mapping, and Growth starts at $2,000 a month for the inbound AI SDR, routing, scheduling and multichannel plays. A slider estimates the Growth price by usage, and seats are unlimited on every plan. The usage unit behind the starting prices and the rate above them are not spelled out, and Enterprise is quoted. A 30 day trial needs no credit card.

Customers keep all rights to their data, and service data is deleted within 90 days after the contract ends, or earlier on request. Access stops as soon as the agreement terminates, and either side can end it with notice. An export path for conversations, identified visitors and account dossiers is not described, and renewal and notice terms for paid plans are not public.

Outbound follow ups leave through the customer's own email and LinkedIn sequencers, so mailbox warmup, volume caps and reputation monitoring belong to those tools. Breakout chooses when to send, triggering a message at the moment a signal fires, and sorts replies into interest, objection, out of office and referral. Limits on how many signal triggered messages an account or contact can receive are not described.

Marketing and revenue teams at B2B software companies are the core buyers, with use cases for SDR teams, RevOps, demand generation, paid traffic and enterprise accounts. Live customers include Ketch, NetApp, CaptivateIQ, MoEngage, HackerEarth, Beautiful.ai, Axiom Law, Revenue Grid and Savant Labs, mostly venture backed software firms. Language support for the website agent and data coverage outside the US are not broken out.