AI agents for customer support

1) Thesis

Customer support is being restructured from seat-based agent software into outcome-based “resolution systems,” where AI agents handle conversations, take actions across enterprise systems, and escalate only exceptions to human staff.

2) Market map

Segment 1: AI-native horizontal resolution agents

Companies: Sierra, Decagon, Ada, Forethought, Intercom Fin, Ultimate.ai (now part of Zendesk)

Dynamics: These vendors are built around autonomous issue resolution rather than agent assistance; differentiation is shifting from language quality to integrations, policy execution, evaluation, and measurable resolution rates.

• Sierra — enterprise-facing autonomous customer-service agents that can execute actions across business systems.
• Decagon — AI agents for support workflows, with emphasis on enterprise integrations and resolution automation.
• Ada — automated customer-service platform with conversational AI and workflow integrations.
• Forethought — AI for ticket resolution, triage, self-service, and agent assistance.
• Intercom Fin — AI support agent embedded in Intercom’s customer-service platform.
• Ultimate.ai — AI-powered customer-service automation and agent assistance; acquired by Zendesk in 2021.

Segment 2: Incumbent customer-service and CRM platforms adding agents

Companies: Salesforce, Zendesk, ServiceNow, HubSpot, Freshworks

Dynamics: Incumbents have the strongest distribution, customer data, permissions, and workflow context; they can bundle AI into existing contracts, but their products may be less autonomous or less flexible than purpose-built challengers.

• Salesforce — Agentforce for service and CRM workflows.
• Zendesk — AI agents, copilots, knowledge tools, and automated ticket handling.
• ServiceNow — AI agents and workflow automation across IT, employee, and customer service.
• HubSpot — Breeze customer-service and CRM AI capabilities.
• Freshworks — Freddy AI for customer support automation, agent assistance, and self-service.

Segment 3: Contact-center and voice AI agents

Companies: Genesys, NICE, Five9, PolyAI, Parloa, Cognigy

Dynamics: Voice remains a large, difficult automation market because of latency, authentication, interruptions, compliance, and complex call flows; incumbents own telephony distribution while specialists often deliver better automation for specific use cases.

• Genesys — AI capabilities embedded in its cloud contact-center platform.
• NICE — CXone AI, agent assistance, automation, and contact-center analytics.
• Five9 — AI-powered contact-center automation and virtual-agent capabilities.
• PolyAI — enterprise voice assistants focused on handling customer calls.
• Parloa — conversational AI agents for enterprise contact centers.
• Cognigy — conversational AI and orchestration for voice and digital channels.

Segment 4: Customer-support intelligence, agent assist, and quality automation

Companies: Cresta, Observe.AI, ASAPP, Level AI, Balto

Dynamics: These vendors generally automate the human agent’s workflow rather than replace the entire support function; they sell into contact-center budgets and benefit from lower adoption risk, but face increasing overlap with CCaaS and CRM vendors.

• Cresta — real-time agent assistance, coaching, automation, and contact-center intelligence.
• Observe.AI — conversation intelligence, quality assurance, agent coaching, and automation.
• ASAPP — AI-powered agent assistance and customer-service automation.
• Level AI — quality management, conversation intelligence, and contact-center analytics.
• Balto — real-time guidance and compliance support for contact-center agents.

Segment 5: Vertical and commerce-native support platforms

Companies: Gorgias, Gladly, Kustomer, Intercom, Hyro, Yuma

Dynamics: Vertical or use-case specialization improves automation because the agent has narrower policies and richer domain context; commerce support is particularly attractive because agents can directly modify orders, refunds, subscriptions, and shipping workflows.

• Gorgias — customer support platform for ecommerce brands, with AI automation across order and customer-service workflows.
• Gladly — customer-service platform focused on unified customer conversations and agent productivity.
• Kustomer — CRM and support platform designed around customer timelines and omnichannel service; owned by Meta.
• Intercom — conversational customer-service platform with Fin AI Agent.
• Hyro — conversational AI for healthcare and other complex service environments.
• Yuma — AI support automation focused heavily on Shopify and ecommerce workflows.

3) Segment likely to capture the most value

Winner: workflow-integrated horizontal resolution platforms

This category includes the strongest parts of both AI-native vendors and incumbent service platforms: systems that can understand the request, retrieve trusted context, take actions in enterprise software, and own the resolution outcome.

It should capture the most value because:

1. It replaces labor, not just software seats. A ticket deflection or fully resolved interaction has substantially more economic value than summarization or agent suggestions.
2. It sits at the point of action. Refunds, cancellations, order changes, credits, account updates, and troubleshooting require permissions and workflow integrations.
3. It can monetize on outcomes. Pricing can evolve from per-seat SaaS toward automated resolutions, successful interactions, or support-cost reduction.
4. It accumulates proprietary operational data. Resolution outcomes, escalation patterns, policies, and exception handling create a feedback loop.
5. It benefits from distribution advantages. Salesforce, Zendesk, ServiceNow, and CCaaS platforms already own the system of record and buyer relationship.

The likely market structure is incumbents capturing the broad platform layer, while AI-native companies win where they offer materially higher automation, faster deployment, or superior performance in a particular workflow.

4) White-space opportunities

These are best viewed as apparent white spaces rather than absolute claims that no company is attempting them.

White space 1: Cross-company “case resolution” agents

A customer issue often spans multiple organizations—for example, a delayed shipment involving a merchant, carrier, payment processor, warranty provider, and bank. Existing agents are usually deployed by one enterprise and can act only within that enterprise’s systems.

Opportunity: an agent that securely coordinates across companies, verifies identity and authorization, exchanges evidence, tracks service-level commitments, and resolves the case end to end.

Potential initial markets include:

• ecommerce delivery and returns,
• travel disruption and refunds,
• insurance claims,
• telecom billing disputes,
• financial-services payment investigations.

The hard problem is not conversation; it is delegated authority, identity, liability, and interoperability across organizations.

White space 2: Outcome-guaranteed support automation for complex exceptions

Most current products optimize containment, deflection, handle time, or ticket automation. Few appear to offer a broadly applicable guarantee that a complex case will be resolved correctly—or that the vendor will share the financial downside when it is not.

Opportunity: a system that combines AI agents, human escalation, verification, and workflow execution under an outcome-based commercial model, such as:

• “resolved within 24 hours,”
• “refund correctly issued,”
• “claim completed without repeat contact,” or
• “customer retention preserved.”

This would require robust evaluation, audit trails, policy controls, and potentially performance guarantees or insurance. Less certain: some BPOs, vertical software companies, and AI-native vendors may already be experimenting with pieces of this model, but it is not yet a clearly established horizontal category.