The CX-AI and voice-AI vendor landscape.

Enterprise CX-AI platform with an "Agent OS" framing and a deep set of named product surfaces (Agent Studio, Agent SDK, Insights, Voice, Live Assist). Sold into Fortune 500 CX organizations on multi-quarter rollout cycles.

Wins on F500 logo reach, AI-specific compliance posture (ISO 42001, EU AI Act), and platform depth when the buyer has a dedicated engineering team to build against the Agent SDK.

Mid-market and consumer-leaning CX-AI vendor with an "AI Concierge" brand metaphor. AOPs (Agent Operating Procedures) let non-engineers define agent behavior in natural language; Decagon Labs runs internal model research.

Wins on growth-marketing reach, glossary-scale SEO, and digital-native consumer brand references (Chime, Notion, Duolingo, gopuff).

No-code self-serve CX-AI vendor that has been in the category longer than most current entrants. Built on a chat-first foundation and now extending into voice and broader operator workflows.

Wins on self-serve onboarding maturity, chat depth, and tiered pricing that lets a single CX manager start small.

Intercom's AI agent layer, bundled with the Intercom messaging platform. Fin sits on top of an existing chat surface that many SaaS support orgs already run, so the deploy story is shorter when Intercom is already in place.

Wins when Intercom is already the system of record for support conversations and the team wants AI without changing platforms.

Real-time coaching and AI for contact-center floors. Cresta started as a copilot for human agents in sales and support and has expanded into autonomous agent workflows from that base.

Wins in contact centers with large human-agent populations where real-time coaching, after-call summaries, and quality assurance lift human performance.

Voice-first conversational AI platform with strong EU and DACH presence. Larger telco and enterprise voice deployments are the core use case, with multi-channel orchestration alongside.

Wins in EU enterprise voice rollouts, especially telco and utilities, where regional presence and language coverage matter.

Enterprise conversational AI platform with a long history in multi-channel orchestration. Strong in regulated industries and complex multi-system integrations.

Wins on multi-channel orchestration breadth, EU enterprise references, and depth of legacy-system integration tooling.

Support-ticket automation vendor with a focus on Salesforce Service Cloud and Zendesk environments. Triage, deflection, and ticket-routing are the core surfaces.

Wins on Salesforce-native support automation where the buyer is already inside the Salesforce ecosystem and wants AI inline with existing case workflows.

Text-to-speech and voice-cloning provider. Heavy presence in content production (audiobooks, dubbing, character voices) and increasingly in real-time agent voice.

Wins on voice quality, language coverage, and the breadth of available voices. The default TTS choice for many voice-agent builders today.

Low-latency text-to-speech infrastructure focused on real-time conversational use. Built around a state-space-model architecture optimized for time-to-first-audio.

Wins on first-audio latency for real-time voice agents where every hundred milliseconds matters.

Speech-to-text and transcription infrastructure. One of the most-used STT engines in production voice-agent pipelines.

Wins on streaming STT throughput, language coverage, and the operational maturity of the API in real-time agent workloads.

Voice agent orchestration layer that sits on top of TTS, STT, and an LLM of the developer's choice. Vapi handles the call-routing, turn-taking, and tool-call plumbing so the developer focuses on the prompt and the integrations.

Wins for engineering teams building bespoke voice agents who want a managed orchestration layer without writing the realtime plumbing themselves.

Voice agent infrastructure in the same orchestration-layer category as Vapi. Provides the realtime stack so developers can wire up an LLM-driven phone agent.

Wins for developer-led voice projects that need a fast path from prototype to a phone number that picks up.

Voice agent infrastructure focused on inbound phone support, with templated agent personas and SMB-friendly onboarding.

Wins for small-business inbound use cases where the buyer wants a phone agent without standing up a custom realtime stack.

Voice agent infrastructure with a focus on enterprise outbound and inbound at high concurrency. Self-hostable models and a developer-first API surface.

Wins on outbound voice at scale, where call concurrency and per-minute economics dominate the decision.

Emotion-aware voice and speech models. Hume's wedge is reading and responding to prosody and affect alongside the literal transcript.

Wins in use cases where the agent needs to recognize emotional state (mental health intake, sensitive support scenarios) and adapt response style.

WebRTC infrastructure, originally for video calling, now widely used as the realtime media layer underneath voice agents. More dev-tool than CX vendor.

Wins as the WebRTC substrate of choice when a team is building voice agents from primitives and needs production-grade media plumbing.

Cloud contact-center platform with a long enterprise track record. AI features (Five9 Agent Assist, Five9 IVA) have been added on top of the existing CCaaS surface over the last several years.

Wins when the buyer has a deep existing Five9 footprint and wants to add AI without replacing the contact-center platform.

One of the largest enterprise CCaaS platforms. NICE has wrapped a broad AI portfolio (Enlighten, Copilot, Autopilot) on top of the CXone contact-center core.

Wins in large enterprise contact-center procurement where the buyer wants a single vendor for CCaaS, WFM, QM, and AI under one roof.

The oldest player in the contact-center category, with both on-premise and cloud product lines. AI capabilities live across Genesys Cloud CX and partner ecosystem integrations.

Wins in regulated enterprise environments with long-running Genesys deployments where displacement risk is the dominant procurement concern.

Newer cloud-native contact-center platform, smaller and faster-moving than Five9, NICE, or Genesys. Strong AI roadmap with Autopilot and Copilot surfaces.

Wins for mid-market and lower-enterprise CCaaS buyers who want a cloud-native platform without the procurement weight of the largest incumbents.

The AI agent bundled with Intercom. Easiest path for teams already standardized on the Intercom messaging platform.

Wins when the buyer is already on Intercom and the question is "can we turn on AI" rather than "which AI platform should we adopt."

Conversational marketing and sales platform. Originally a chat tool for B2B websites; now part of Salesloft and increasingly AI-augmented.

Wins for revenue teams using chat as a top-of-funnel pipeline tool rather than a post-sale support channel.

Zendesk's built-in AI deflection and answer bot, packaged with the Zendesk Suite. The path of least resistance for teams already on Zendesk.

Wins for teams on Zendesk that want a deflection bot and after-call summarization without standing up a separate AI vendor.

HubSpot's chatbot builder, included with the HubSpot Service Hub. A starter surface for teams that want AI inside the HubSpot stack they already pay for.

Wins for HubSpot-native shops at the lower end of the market where the chatbot is one of several AI features on an existing seat.

Outbound sales orchestration platform with built-in prospect data and a fast-growing AI surface. Used heavily by SDR teams running cold outbound at scale.

Wins for outbound sales teams that need a single platform for prospect data, sequencing, and AI-assisted outreach.

Sales engagement platform that has expanded into conversational AI (Drift acquisition) and revenue orchestration. Enterprise revenue-team workhorse.

Wins for enterprise revenue orgs running sequenced outbound and inbound conversations across the funnel.

The other large sales engagement platform alongside Salesloft. Used by enterprise revenue teams for sequencing, deal management, and increasingly AI-assisted call analysis.

Wins in enterprise revenue stacks where the buyer wants deep deal intelligence and forecasting alongside the engagement surface.

Whose name is on the audit row?

When the AI closes a case, updates a record, or commits a calendar entry, whose identity does the system log? If the audit row says "the AI did it," you have a black box your operators cannot review or stand behind. If the audit row names a specific human operator and shows the AI's role as a tool that human used, you have a system that fits inside the way regulated industries already work.

Where does the customer data live?

In the vendor's cloud, or in your CRM? This is the single highest-leverage question in the category and the one most homepages will not answer directly. Ask: if we terminate this contract in year three, what happens to the conversation history, the customer records, the agent traces, and the analytics store? The answer reveals the actual lock-in shape of the deal.

Can you swap the voice substrate?

Is the voice pipeline a chained stack (separate STT, LLM, TTS) where individual components can be replaced as the frontier shifts, or is it a single speech-to-speech vendor that locks the entire voice surface to one model provider? Voice quality and latency are improving on a quarterly cadence; the product you sign today will be running on different substrates next year, and you want to know which architecture lets that happen without a re-platform.

What public benchmark are they on?

Internal benchmarks (Sierra has τ-voice and τ-knowledge; Decagon has Decagon Labs) are interesting research, but the vendor designs and runs the test. Ask whether there is any external, multi-vendor benchmark the vendor agrees to be on. Vorel is building one (CXBench). The category as a whole is moving toward public evaluation, and the vendors who refuse to be ranked externally are telling you something.

Who is the protagonist of the product?

Read the vendor's product pages and ask whose face shows up on the screenshots. Is it the bot, the AI persona, the "concierge"? Or is it the human operator, the case manager, the dispatcher, the supervisor? The protagonist of the marketing copy is almost always the protagonist of the product, and that protagonist determines how the workflow surfaces are designed. Pick the shape that matches how your team actually works.