AI SDR & Outbound · Conversational Voice Calling Agents
Best AI Calling Tools in 2026
Also known as: AI Voice Agent, AI Cold Calling
Ultra-low latency conversational voice agents and AI outbound tele-calling platforms. Powered by sub-second speech-to-text, LLM decision reasoning, and expressive neural speech synthesis to conduct qualification surveys, inbound routing, and automated appointment follow-ups.
How AI Calling tools segment by buyer profile
High defensibility (Moat ≥ 3.5) with predictable 5-seat monthly TCO and no mandatory annual lock-in.
- PolyAI~$49/mo (base) · Moat 4.5/5 · L3
- Bland.ai~$299/mo (base) · Moat 4.5/5 · L3
- Dialpad Ai~$75/mo · Moat 4.4/5 · L4
- Mailshake~$60/mo · Moat 4.1/5 · L2
Proprietary data or multi-agent orchestration built for mid-market and enterprise governance.
- Bland.ai~$299/mo (base) · Moat 4.5/5 · Enterprise
- Gong.ioCustom quote · Moat 4.5/5 · Enterprise
- PolyAI~$49/mo (base) · Moat 4.5/5 · SOC 2, GDPR
- Dialpad Ai~$75/mo · Moat 4.4/5 · Enterprise
Fastest onboarding for solo operators and 5-person teams under $250/mo or with a free plan.
- Fathom VideoCustom quote · Free tier · Moat 4.3/5
- Phonely AICustom quote · Free tier · Moat 3.9/5
- Sybill AICustom quote · Free tier · Moat 3.8/5
- ContactOutCustom quote · Free tier · Moat 2.6/5
Verify per-seat multipliers, credit exhaustion rules, or annual lock-in clauses before signing.
- Bland.aiUsage credit burn — Telephony Carrier Filtering
- Retell AIUsage credit burn — Latency degradation
- Dialpad AiPer-seat scaling — Acoustic transcription degradation
- Synthflow AIAnnual lock-in — Telephony Latency Jitter
Interactive procurement & TCO comparison matrix
| Free Option | Primary Watchout & Stack | Visit & shortlist | |||||
|---|---|---|---|---|---|---|---|
Bland.ai✓ Verified Infrastructure-level conversational voice platform delivering sub-second latency for enterprise inbound and outbound phone automation. | ~$299/mo (base) | Free plan · $299/mo · Enterprise planFreemium | Free plan | L3 · Conditional automation | 4.5 / 5.0Proprietary moat (category leader) | ⚠️ Telephony Carrier Filtering: High-volume outbound calling velocity faces downstream carrier STIR/SHAKEN attestation drops an ✓ Twilio Segment | |
Gong.io✓ Verified Revenue AI operating system that captures multi-channel sales interactions to automate deal execution, pipeline forecasting, and conversational intelligence. | Custom quote | Custom / enterprise pricingEnterprise / quote | No | L3 · Conditional automation | 4.5 / 5.0Proprietary moat (category leader) | ⚠️ Telephony Latency and Webhook Sync Constraints: High-volume call centers experience processing lags when transcribing simultaneous multi-hour c ✓ Twilio Segment | |
Retell AI✓ Verified Developer-first conversational voice API platform enabling real-time, low-latency inbound and outbound telephone call automation. | Custom quote | Usage-based pricingUsage-based | No | L3 · Conditional automation | 4.1 / 5.0Proprietary moat | ⚠️ Latency degradation: Compound network jitter across multi-model pipeline stages can elevate latency past acceptable ✓ HubSpot✓ Twilio Segment✓ n8n🛡️ GDPR🛡️ HIPAA | |
Clari Copilot✓ Verified Predictive conversation intelligence and real-time sales coaching platform that transcribes calls, analyzes buyer signals, and updates CRM deal data. | Custom quote | Custom / enterprise pricingEnterprise / quote | No | L3 · Conditional automation | 3.9 / 5.0Proprietary moat | ⚠️ Meeting Bot Rejection: Virtual recording bots face participant rejection or calendar exclusion in strict corporate env ✓ Twilio Segment | |
Dialpad Ai✓ Verified Cloud-native unified communications and contact center platform providing real-time speech recognition, live agent coaching, and automated call summaries. | ~$75/moPer-seat scaling | $15–$25/seat/mo · Enterprise planPer seat | No | L4 · Autonomous agent | 4.4 / 5.0Proprietary moat (category leader) | ⚠️ Acoustic transcription degradation: High ambient background noise and packet loss over degraded cellular SIP links noticeably incre ✓ Twilio Segment | |
Synthflow AI✓ Verified No-code enterprise voice AI platform that automates inbound customer service and outbound phone outreach via real-time speech orchestration and native. | ~$2,500/mo (base) | From $30,000/yrSubscription | No | L3 · Conditional automation | 4.1 / 5.0Proprietary moat | ⚠️ Telephony Latency Jitter: Network hops across third-party SIP providers and distributed LLM inference points can generate ✓ Salesforce✓ HubSpot✓ Zapier🛡️ SOC 2🛡️ GDPR | |
Vapi.ai✓ Verified Voice AI developer platform orchestrating ultra-low-latency telephony, speech recognition, LLM reasoning, and neural voice synthesis via real-time APIs. | Custom quote | Usage-based pricingUsage-based | No | L3 · Conditional automation | 4.1 / 5.0Proprietary moat | ⚠️ Stacked Provider Latency: Chaining separate outside speech-to-text, LLM, and voice APIs can still add brief reply lag if ✓ Slack🛡️ HIPAA | |
PolyAI✓ Verified Enterprise conversational voice AI platform powered by proprietary spoken-language models for customer engagement and GTM. | ~$49/mo (base)Essentials tier | $49–$299/mo · Enterprise planSubscription | No | L3 · Conditional automation | 4.5 / 5.0Proprietary moat (category leader) | ⚠️ Enterprise Implementation & Per-Minute Commitment Barrier: PolyAI scopes commercial engagements around custom implementation packages and ongoing per-minu ✓ Salesforce✓ Zendesk🛡️ SOC 2🛡️ GDPR | |
Voiceflow✓ Verified Enterprise AI agent platform that customer support and CX teams use to build, test, and deploy conversational AI agents across chat and voice. | Custom quote | Custom / enterprise pricingEnterprise / quote | No | L3 · Conditional automation | 4.4 / 5.0Proprietary moat (category leader) | ⚠️ State Machine Complexity: Large enterprise support workflows spanning hundreds of nested decision nodes and custom API ca | |
Balto✓ Verified Real-time contact center intelligence platform delivering live agent guidance, automated quality assurance, and conversational analytics during voice calls. | Custom quote | Custom / enterprise pricingEnterprise / quote | No | L3 · Conditional automation | 4.2 / 5.0Proprietary moat | ⚠️ Audio Stream Degradation: Packet loss, acoustic crosstalk, and degraded VoIP codecs hinder live transcription fidelity, p ✓ Twilio Segment | |
Consio✓ Verified Voice commerce and autonomous telephone support platform purpose-built for Shopify merchants to automate inbound customer calls and order modifications. | ~$60/mo (base) | From $60/moSubscription | No | L3 · Conditional automation | 3.4 / 5.0Workflow-embedded (strong) | ⚠️ Ecosystem Lock-in: Consio remains tightly coupled to the Shopify data model, leaving merchants on alternative arch ✓ Shopify✓ WooCommerce✓ Twilio Segment | |
Sybill AI✓ Verified Multimodal conversation intelligence platform combining non-verbal visual tracking and speech transcription to automate sales follow-ups and CRM field. | Custom quotePer-seat scaling | Custom / enterprise pricingEnterprise / quote | Free plan | L3 · Conditional automation | 3.8 / 5.0Proprietary moat | ⚠️ Camera Occlusion Sensitivity: Multimodal visual sentiment models fail to gather behavioral signals whenever prospective buyer ✓ Slack✓ Microsoft Teams✓ Zoom🛡️ SOC 2🛡️ GDPR | |
WINN.AI✓ Verified Real-time AI sales copilot providing live in-call playbook tracking, dynamic objection handling battle cards, and automated post-call CRM data. | Custom quote | Custom / enterprise pricingEnterprise / quote | No | L3 · Conditional automation | 3.9 / 5.0Proprietary moat | ⚠️ CRM Schema Drift: Automated field population relies on rigid object schemas; unexpected API modifications or cust ✓ Salesforce✓ HubSpot✓ Slack🛡️ SOC 2🛡️ GDPR | |
Superleap✓ Verified Enterprise AI CRM and conversational voice platform deploying autonomous SuperAgents to conduct inbound qualification, outbound calling, and automated. | Custom quote | Custom / enterprise pricingEnterprise / quote | No | L3 · Conditional automation | 3.3 / 5.0Workflow-embedded (strong) | ⚠️ Telephony Latency and Turn-Taking Overhead: Network hops across regional SIP trunks and cascaded inference steps can occasionally induce la ✓ Mailchimp✓ Intercom | |
Mailshake✓ Verified Sales engagement and outbound automation platform orchestrating multi-channel email sequencing, cold outreach deliverability, integrated VoIP calling. | ~$60/moPer-seat scaling | From $12/moSubscription | No | L2 · Workflow copilot | 4.1 / 5.0Proprietary moat | ⚠️ Autonomous Conversational Voice Capabilities: Unlike dedicated voice agents that conduct autonomous bi-directional speech conversations, the ✓ Salesforce✓ HubSpot✓ Pipedrive🛡️ CCPA | |
Invoca✓ Verified Enterprise conversation intelligence and AI revenue agent platform that tracks inbound calls, analyzes spoken dialogue via deep learning, and automates. | Custom quote | Custom / enterprise pricingEnterprise / quote | No | L3 · Conditional automation | 4.2 / 5.0Proprietary moat | ⚠️ Number Pool Exhaustion: High-traffic web spikes can deplete dynamic number pools, causing attribution collisions or una ✓ Salesforce✓ HubSpot✓ Slack | |
Coval✓ Verified Voice and chat AI simulation, evaluation, and production observability platform backed by $31.8M in Series A funding. | ~$100/mo (base)Starter tier | $100–$4,500/mo · Enterprise planSubscription | No | L3 · Conditional automation | 3.3 / 5.0Workflow-embedded (strong) | ⚠️ Primary Operational Trade-Off: In practical enterprise deployments, the primary trade-off is that the $100/month Starter plan ✓ Slack✓ Model Context Protocol (MCP)🛡️ SOC 2🛡️ GDPR | |
Air.ai✓ Verified Autonomous conversational voice engine executing multi-turn inbound and outbound phone calls using low-latency neural speech pipelines. | Custom quote | Custom / enterprise pricingEnterprise / quote | No | L3 · Conditional automation | 3.3 / 5.0Workflow-embedded (strong) | ⚠️ Telephony Latency and Network Jitter: Packet loss or transport lag on standard PSTN connections occasionally degrades turn-taking syn | |
Fathom Video✓ Verified AI-powered meeting recording, transcription, and summarization software that extracts action items and syncs notes to CRMs directly from video conferences. | Custom quotePer-seat scaling | Custom / enterprise pricingEnterprise / quote | Free plan | L2 · Workflow copilot | 4.3 / 5.0Proprietary moat (category leader) | ⚠️ Conference Platform Constraints: Third-party platform API changes or host-side bot admission denials can disrupt automated captu ✓ Salesforce✓ HubSpot✓ Microsoft Teams | |
Leaping AI✓ Verified Enterprise AI voice and text agents that answer calls, qualify leads, and book appointments 24/7. | Custom quote | Pricing not disclosedNot disclosed | — | L3 · Conditional automation | 3.2 / 5.0Workflow-embedded (strong) | ⚠️ Carrier Registration & STIR/SHAKEN Bottleneck: Running high-volume outbound voice and SMS speed-to-lead campaigns requires strict A2P 10DLC an ✓ Salesforce✓ HubSpot✓ Zendesk🛡️ GDPR | |
Terrakotta✓ Verified AI voice-cloned voicemail and power-dialing platform built for high-volume phone outreach and real estate brokers. | Custom quote | Pricing not disclosedNot disclosed | — | L3 · Conditional automation | 2.6 / 5.0Workflow-embedded | ⚠️ TCPA & Telemarketing Regulatory Barrier: Deploying automated AI voice-cloned voicemail drops and SMS follow-ups requires strict complian ✓ Salesforce✓ HubSpot✓ Outreach | |
Parloa✓ Verified Enterprise conversational voice calling platform orchestrating natural, low-latency customer service dialogues and automated multi-turn phone workflows. | Custom quote | Custom / enterprise pricingEnterprise / quote | No | L3 · Conditional automation | 4.0 / 5.0Proprietary moat | ⚠️ Telephony Latency Jitter: Network packet loss and transcontinental SIP routing introduce latency spikes that degrade turn | |
Cekura✓ Verified Automated QA, adversarial simulation, and production call monitoring platform built for enterprise Voice AI and Chat AI agents. | ~$150/moPer-seat scaling | Free plan · From $30/seat/mo · Enterprise planFreemium | Free plan | L3 · Conditional automation | 2.5 / 5.0Workflow-embedded | ⚠️ Primary Operational Trade-Off: In practical enterprise deployments, the primary trade-off is that Cekura is a specialized QA, 🛡️ HIPAA | |
Phonely AI✓ Verified Phonely AI develops enterprise-grade conversational voice agents powered by its proprietary Alma telephony model to automate inbound and outbound phone. | Custom quote | Custom / enterprise pricingEnterprise / quote | Free plan | L3 · Conditional automation | 3.9 / 5.0Proprietary moat | ⚠️ Telephony Latency and Network Jitter: Packet loss and public PSTN carrier jitter can degrade Alma voice synthesis, occasionally forci ✓ Slack🛡️ SOC 2🛡️ GDPR | |
Simple AI✓ Verified Autonomous AI voice agents that convert inbound calls, follow up on leads, and drive upsells like top sales reps. | Custom quote | Pricing not disclosedNot disclosed | — | L3 · Conditional automation | 3.8 / 5.0Proprietary moat | ⚠️ Catalog & POS Integration Learning Curve: Because Simple AI executes live transactions, upsells, and bookings directly inside enterprise 🛡️ SOC 2🛡️ HIPAA | |
Attention.tech✓ Verified Attention.tech delivers conversational voice intelligence agents that transcribe sales interactions and automate downstream CRM pipeline hygiene. | Custom quote | Custom / enterprise pricingEnterprise / quote | No | L3 · Conditional automation | 3.6 / 5.0Proprietary moat | ⚠️ Schema drift vulnerability: Arbitrary additions of custom CRM validation rules frequently trigger silent synchronization fa ✓ Microsoft Teams✓ Zoom | |
Kaspr✓ Verified B2B contact enrichment platform and Chrome extension providing direct European phone numbers and verified business emails from LinkedIn profiles. | Custom quote | Custom / enterprise pricingEnterprise / quote | Free plan | L2 · Workflow copilot | 2.3 / 5.0Workflow-embedded | ⚠️ DOM Dependency: Third-party interface adjustments by LinkedIn can disrupt client-side DOM scraping elements, re ✓ Salesforce✓ LinkedIn✓ Zapier🛡️ GDPR | |
Roark✓ Verified Audio-native Voice AI testing, simulation replay, and 500+ metric production evaluation platform backed by Y Combinator. | ~$500/mo (base)Team tier | Free plan · $500–$4,000/mo · Enterprise planFreemium | Free plan | L3 · Conditional automation | 2.4 / 5.0Workflow-embedded | ⚠️ Primary Operational Trade-Off: In practical enterprise deployments, the primary trade-off is that Roark bills simulation minut 🛡️ SOC 2🛡️ HIPAA | |
Swordfish AI✓ Verified Autonomous conversational calling and contact verification platform orchestrating outbound agent workflows for enterprise sales development pipelines. | Custom quote | Custom / enterprise pricingEnterprise / quote | No | L3 · Conditional automation | 2.5 / 5.0Workflow-embedded | ⚠️ Carrier Filtering Overhead: High-volume automated outbound dialing runs the continuous risk of carrier-level spam tagging, ✓ Outreach | |
Marr Labs✓ Verified Compliant, auditable AI voice agent platform automating inbound and outbound customer calls for regulated sales, lending, and service teams. | Custom quote | Custom / enterprise pricingEnterprise / quote | No | L4 · Autonomous agent | 3.3 / 5.0Workflow-embedded (strong) | ⚠️ Primary Operational Trade-Off: In practical enterprise deployments, the primary trade-off is that following its acquisition by | |
Second Nature✓ Verified AI sales roleplay and coaching platform using interactive voice and video avatars to train and certify revenue teams. | Custom quote | Pricing not disclosedNot disclosed | — | L3 · Conditional automation | 3.2 / 5.0Workflow-embedded (strong) | ⚠️ Primary Operational Trade-Off: In practical enterprise enablement deployments, the primary trade-off is that simulated AI role ✓ Zoom | |
Hamming AI✓ Verified Enterprise voice and chat agent testing, adversarial red-teaming, and real-time production call monitoring platform. | Custom quote | Custom / enterprise pricingEnterprise / quote | No | L3 · Conditional automation | 3.0 / 5.0Workflow-embedded (strong) | ⚠️ Primary Operational Trade-Off: In practical enterprise deployments, the primary trade-off is that Hamming AI does not publish 🛡️ SOC 2🛡️ HIPAA | |
Freya✓ Verified Enterprise voice AI and ultra-low-latency text-to-speech platform (Adam V1) automating multilingual customer calls and voice agent workflows. | Custom quote | Custom / enterprise pricingEnterprise / quote | No | L4 · Autonomous agent | 2.8 / 5.0Workflow-embedded | ⚠️ Primary Operational Trade-Off: In practical enterprise deployments, the primary trade-off is that deploying custom-voiced, low ✓ Outreach🛡️ GDPR | |
Thoughtly✓ Verified Human-like AI voice agent platform that responds to leads in seconds, answers questions, qualifies prospects, and books meetings automatically. | Custom quote | Usage-based pricingUsage-based | No | L4 · Autonomous agent | 2.9 / 5.0Workflow-embedded | ⚠️ Primary Operational Trade-Off: In practical enterprise deployments, the primary trade-off is that automated outbound voice cal ✓ Salesforce✓ HubSpot | |
ContactOut✓ Verified B2B contact intelligence platform providing verified direct phone numbers and personal email addresses for outbound sales and recruiting teams. | Custom quote | Custom / enterprise pricingEnterprise / quote | Free plan | L2 · Workflow copilot | 2.6 / 5.0Workflow-embedded | ⚠️ Platform Dependency: The client extension relies on DOM parsing across third-party social networks, introducing vuln ✓ Salesforce✓ LinkedIn✓ Twilio Segment🛡️ GDPR🛡️ CCPA | |
Samora AI✓ Verified Multilingual enterprise voice agents that automate high-volume inbound and outbound calls with safe escalation. | Custom quote | Pricing not disclosedNot disclosed | — | L3 · Conditional automation | 2.6 / 5.0Workflow-embedded | ⚠️ Human Escalation Queue Dependency: Because Samora AI is architected around safe human-in-the-loop escalation for high-risk edge ca ✓ Salesforce✓ HubSpot | |
Acrely✓ Verified Controllable enterprise AI voice agents deployable in cloud or on-prem for sales and customer operations. | Custom quote | Pricing not disclosedNot disclosed | — | L3 · Conditional automation | 2.3 / 5.0Workflow-embedded | ⚠️ On-Premises Telephony Integration Complexity: Deploying self-hosted voice models and wiring custom enterprise SIP trunks and internal APIs in ✓ Salesforce✓ Zendesk | |
Prospeo✓ Verified B2B contact enrichment and data verification platform providing verified email addresses, mobile direct dials, and automated LinkedIn scraping. | Custom quote | Custom / enterprise pricingEnterprise / quote | Free plan | L2 · Workflow copilot | 2.3 / 5.0Workflow-embedded | ⚠️ Catch-All Ambiguity: Severe server configurations that mask corporate mailboxes behind generic catch-all rules degra ✓ Salesforce✓ HubSpot✓ Outreach🛡️ GDPR | |
Structurely✓ Verified Conversational AI voice and SMS inside sales agent platform automating two-way lead qualification and booking. | Custom quote | Usage-based pricingUsage-based | No | L3 · Conditional automation | 2.6 / 5.0Workflow-embedded | ⚠️ Vertical Concentration in Real Estate & Consumer Lending: Structurely's pre-trained conversation scripts and CRM connectors are heavily optimized for rea ✓ Outreach | |
RingtimeCatalog Autonomous conversational voice calling platform orchestrating outbound campaigns, qualification dialogues, and bidirectional CRM synchronization. | Custom / enterprise pricing | Custom / enterprise pricingEnterprise / quote | No | L3 · Conditional automation | 1.8 / 5.0Thin wrapper | — | |
Seamless.aiCatalog Real-time B2B sales prospecting search engine delivering direct dials, email addresses, and automated CRM contact enrichment. | Custom / enterprise pricing | Custom / enterprise pricingEnterprise / quote | No | L2 · Workflow copilot | 1.5 / 5.0Thin wrapper | — | |
VolcaCatalog Autonomous conversational voice agent platform orchestrating bidirectional outbound and inbound enterprise telephony workflows with real-time speech. | Custom / enterprise pricing | Custom / enterprise pricingEnterprise / quote | No | L3 · Conditional automation | 1.8 / 5.0Thin wrapper | — | |
RegalCatalog Enterprise AI voice and SMS sales agent platform automating outbound lead qualification, appointment scheduling, and high-consideration customer calls. | Usage-based pricing (est.) | Usage-based pricingUsage-based (est.) | No | L4 · Autonomous agent (est.) | Proprietary moat (category leader) (est.)Proprietary moat (category leader) | — | |
YoodliCatalog Enterprise AI voice roleplay and sales pitch certification simulator enabling revenue teams to rehearse multi-persona buyer calls with rubric scoring. | Free plan · $8–$20/mo · Enterprise plan (est.) | Free plan · $8–$20/mo · Enterprise plan (est.)Freemium (est.) | Free plan | L3 · Conditional automation (est.) | Proprietary moat (category leader) (est.)Proprietary moat (category leader) | — |
How to evaluate AI Calling software
- Latency and interruptions
- Run a live test call. Check how quickly the agent responds and whether it stops talking when the caller interrupts.
- Telephony
- Some platforms include phone numbers; others expect you to connect your own carrier or SIP trunk. This changes setup work and per-minute cost.
- Usage pricing
- Voice agents usually charge per minute, often on top of a platform fee. A five-seat floor understates cost at high call volume, so model your expected minutes.
- Calling consent rules
- In the US, the FCC ruled in February 2024 that AI-generated voices count as artificial voices under the TCPA, so outbound calls need prior consent. Check how the platform records consent and handles opt-outs.
- CRM and calendar integration
- Confirm that call outcomes, transcripts and booked meetings are written back to your CRM and calendar.
- Developer platform or ready-made agent
- API-first platforms give more control but need engineering time; no-code agents launch faster with less flexibility. Decide which your team can support.
Benchmark vendors we compare against: Bland.ai, Vapi.ai, Retell AI, Air.ai, Synthflow AI, PolyAI, Voiceflow.
Frequently asked questions: AI Calling
What latency is necessary for a natural conversational voice call?
Natural voice interaction requires end-to-end response latency under 600–800 milliseconds. Leading platforms achieve 400–500ms using streaming WebSocket architectures and edge voice models.
Are conversational voice agents compliant with telemarketing regulations?
Enterprise voice platforms provide TCPA compliance features, mandatory call recording disclosures, automatic DNC (Do Not Call) registry scrubbing, and explicit human escalation pathways.
Can I use an AI voice agent for cold calling in the US?
Only with care. The FCC ruled in February 2024 that AI-generated voices are artificial voices under the TCPA, which requires prior express consent for such calls. Talk to counsel before running outbound AI calls to lists without consent.
What should I check before buying an AI voice agent?
Response latency and interruption handling in a live call, whether phone numbers are included, per-minute pricing at your call volume, consent and opt-out handling, CRM and calendar write-back, and whether you need an API-first platform or a ready-made agent.
How much does AI Calling software cost in 2026?
Across 39 fact-checked AI Calling tools in Conversational Voice Calling Agents, public entry pricing starts at From $12/mo (Mailshake) and ranges up to From $30,000/yr (Synthflow AI). 28 tools offer custom enterprise contracts.
Which AI Calling tools offer a free plan or free trial?
9 tools in Conversational Voice Calling Agents offer a permanent free tier (Bland.ai, Sybill AI, Fathom Video, Cekura, Phonely AI).
Which AI Calling platforms rate highest for technical moat and AI autonomy?
Based on human fact-checked evaluations in Conversational Voice Calling Agents, the highest-rated platforms by moat depth are Bland.ai (4.5/5.0, L3 · Conditional automation), Gong.io (4.5/5.0, L3 · Conditional automation), PolyAI (4.5/5.0, L3 · Conditional automation), Dialpad Ai (4.4/5.0, L4 · Autonomous agent). 4 products in this subtrack operate at L4–L5 autonomous execution (Dialpad Ai, Marr Labs, Freya, Thoughtly).