ooligo

SoundHound AI

ai-customer-experience conversational-ai · ai-support-agent · voice-ai
AI-NATIVE API
Customer Success
6.6 /10

What it is

SoundHound AI (Nasdaq: SOUN) started as a voice-recognition company for cars and devices and has become a roll-up of enterprise conversational-AI vendors. It bought SYNQ3, a restaurant phone-ordering business, in late 2023 for about $25M; Amelia, the enterprise customer-service and employee-help agent platform, in 2024 for about $80M; Interactions, a contact-centre voice-automation vendor, in September 2025 for about $60M; and LivePerson, the digital-messaging platform behind Conversational Cloud, in a deal that closed on 2026-09-04. LivePerson is now a wholly owned subsidiary and its shares no longer trade on Nasdaq.

The products are being consolidated into OASYS (Orchestrated Agent System), announced on 2026-05-05. SoundHound describes it as the best of Amelia, SoundHound, Interactions and SYNQ3 combined, with agents built once and deployed to phone, SMS, web chat, social, kiosks and in-car systems, and an orchestration layer that coordinates several agents inside one conversation. Management has said every customer will eventually migrate to OASYS. LivePerson’s messaging layer, which carries about 1B customer messages a month, is being folded in as OASYS’s digital channel.

The company is growing and losing money. Q3 2026 revenue was $62.4M, up 48.5% year on year and including about $14M from LivePerson for the 26 days after close, with a GAAP net loss of $46.4M. Full-year 2026 guidance is $250M–$260M, and SoundHound targets positive adjusted EBITDA in the second half of 2027.

Why it shows up on shortlists

  • Analyst cover. Gartner’s July 2026 Magic Quadrant for Conversational AI Platforms named SoundHound AI one of four Leaders, beside Google, Salesforce and Kore.ai. For procurement teams that need a Leader-quadrant name, it qualifies.
  • Voice is the core competence. SoundHound owns its speech recognition and language stack rather than reselling a cloud provider’s, and Amelia, Interactions and SYNQ3 bring years of high-volume phone automation; named Amelia customers include Allina Health and Stellantis. If most of your contact volume is phone calls, that depth matters more than chat-first vendors’ claims.
  • Voice plus messaging under one contract. With LivePerson, one vendor now covers voice agents, web and app messaging, and the human-agent workspace. SoundHound says the combined company counts 25 Fortune 100 companies as customers.

The LivePerson question

If you run LivePerson Conversational Cloud, you did not choose this vendor; you inherited it. Three facts set the decision. First, the platform is not being shut down: SoundHound’s plan is to fold it into OASYS, and management says it has engaged LivePerson’s top 90 accounts, worth over $100M in ARR, with full platform and organizational integration targeted for the start of 2027. Second, no public date exists for how long Conversational Cloud will run side by side with OASYS, or on what pricing customers move. Third, the balance sheet risk that hung over LivePerson is gone: SoundHound retired nearly all of LivePerson’s secured debt at close, paying about $261M to noteholders on top of about $43M to shareholders.

The practical path: at your next renewal, ask for the Conversational Cloud end-of-support date, the OASYS migration plan for your bots and integrations, and a price hold through migration, all in writing. If SoundHound cannot give them, run a competitive evaluation in the same cycle rather than renewing for multiple years.

Pricing reality

SoundHound publishes no rate card. Contracts are annual and custom, and the company is moving customers from upfront licences to hosted, recurring billing. Expect separate line items for platform usage, voice telephony, implementation and professional services. The only public per-minute figure is a competitor’s model: Retell AI estimates Amelia’s effective cost at $0.42–$0.60 a minute at 50,000 minutes a year, which is a sales comparison, not a quote.

Best for

Contact-centre and CX leaders at large enterprises whose volume is mostly voice (insurance, healthcare, utilities, automotive, restaurant chains) and who want phone and digital agents from one vendor with Leader-quadrant analyst coverage. Also existing Amelia, Interactions and LivePerson customers deciding whether to stay.

Skip it if you are a software company resolving email and chat tickets on Zendesk or Intercom; a help-desk-native agent goes live faster at a fraction of the cost.

Versus the alternatives

The market-share leaders in the same quadrant are Google and Salesforce Agentforce. Pick Agentforce when Service Cloud is your system of record; pick Google’s suite when you are a Google Cloud shop. Kore.ai is the closest like-for-like: pick it over SoundHound when you need a platform you build and govern yourself across several CRMs and CCaaS vendors, with a deterministic layer for regulated scripts and a vendor that is not mid-merger.

The fastest-growing entrant is Sierra; pick it for an outcome-priced, vendor-built agent for a consumer brand. Decagon and Ada fit digital-first support teams. For voice-only deployments, PolyAI and Parloa are narrower and faster to stand up.

If none fit, start with the AI agent inside your existing CCaaS or help desk on your top ten intents, and revisit a platform once you have documented flows.

Watch-outs

  • Four platforms are being merged into one. Amelia, Interactions, SYNQ3 and LivePerson customers are all on a path to OASYS, and migrations of this size slip. Guard: tie any multi-year term to a written migration plan, keep the right to stay on your current platform until OASYS passes your acceptance tests, and keep exports of your intents, flows and transcripts current.
  • The vendor is not yet profitable. SoundHound lost $46.4M in Q3 2026 and is cutting costs to reach adjusted-EBITDA breakeven in late 2027, which can mean thinner support and dropped products. Guard: put named support levels and service credits in the contract, and review the quarterly results before each renewal.
  • Pricing is opaque and services-heavy. With no rate card and a shift to hosted billing, year-two costs are hard to predict. Guard: ask for a three-year total cost covering usage, telephony, services and support, and compare it with a per-resolution quote from a help-desk-native agent on the same volume.