Support Agent Platforms

The decision between building custom agents on a framework (like those in our Developer Frameworks article) versus buying a turnkey platform comes down to three factors: time to value, control requirements, and team capabilities.

Turnkey platforms get you to production in days or weeks, not months. They come with pre-built integrations to major helpdesks, trained models for common support scenarios, and analytics dashboards out of the box. The trade-off is less control over agent behaviour, pricing tied to resolution volume, and dependency on the vendor's roadmap for new features.

Building custom gives you full control over agent logic, the ability to integrate deeply with internal systems, and no per-resolution fees. But you need engineers who understand LLM orchestration, and you're responsible for reliability, monitoring, and iteration.

Many teams start with a turnkey platform for quick wins and build custom agents for specialised workflows that the platform can't handle well.

Intercom Fin is the most transparent platform in terms of published performance metrics. As of March 2026 (the Fin 3 launch), Intercom reports a 66% average resolution rate across 6,000+ customers, with over 20% of customers achieving 80%+ resolution rates.

Fin 3 introduced several capabilities: Fin Voice (voice channel support), Fin Vision (image understanding for screenshots and product photos), Fin Procedures (multi-step process automation replacing the earlier "tasks" feature), and Fin Email (automated email responses across inboxes). The Fin Copilot is now available to all agents with multilingual support.

Pricing is $0.99 per successful resolution, with a minimum of 50 resolutions per month. This per-resolution model means you pay for outcomes, not seat count — which aligns incentives but can become expensive at high volumes. A team resolving 10,000 tickets per month through Fin would pay roughly $6,600/month (at 66% resolution rate).

Zendesk made a significant strategic move in March 2026 by acquiring Forethought, a startup powering over 1 billion customer interactions per month. Forethought brought a suite of specialised agents: Solve (omnichannel resolution), Assist (agent workflow and draft responses), Discover (insight extraction and article generation), Triage (tagging and prioritisation), and Agent QA (scoring 100% of interactions).

Zendesk's own AI agents, trained on 18+ billion historical interactions, already claimed up to 80% autonomous resolution with support for 80+ languages. The Forethought acquisition deepens these capabilities, particularly in the quality assurance and knowledge management areas.

Zendesk's advantage is its enormous installed base and deep integrations with Shopify, Salesforce, Jira, and other business systems. For teams already on Zendesk, the AI agents are a natural extension — no new vendor relationship required. Forethought remains available to non-Zendesk customers for now, but the long-term integration trajectory is clear.

Salesforce launched Agentforce Contact Center in March 2026, unifying voice, digital channels, CRM data, and AI agents into a single native platform. Early deployments report 40-60% containment rates — the percentage of interactions fully resolved without human handoff.

The platform's key differentiator is deep CRM integration. Agents have real-time access to the full customer record, order history, case history, and account health data. This context makes them particularly effective for account-specific queries: travel changes, billing questions, subscription updates.

Agentforce is priced as part of the Salesforce platform license, making it most cost-effective for teams already invested in the Salesforce ecosystem. For teams not on Salesforce, the value proposition is weaker — you'd be adopting an entire platform for the AI agent capability.

Sierra, founded by Bret Taylor (former Salesforce co-CEO) and Clay Bavor (former Google VP), has reached a $4.5 billion valuation. The platform takes an enterprise-first approach: configurable AI agents that are pre-built for common support scenarios, with the technical complexity abstracted away.

Sierra's pitch is that support teams shouldn't need to understand agent frameworks or prompt engineering. You describe what you want the agent to do in business terms, and the platform handles the implementation. This removes the technical bottleneck but ties you to Sierra's timeline for changes and customisations.

Sierra is best suited for large enterprise teams that want a managed, battle-tested solution and are willing to trade customisation speed for reduced engineering overhead.

Decagon, raising $100M+ in funding, offers a distinctive approach with Agent Operating Procedures (AOPs). Instead of configuring agents through UI wizards, you define agent logic in natural language. This gives technical CX teams direct control over agent behaviour without writing code, but requires clear thinking about edge cases and decision trees.

Decagon supports chat, email, and voice channels, and emphasises fast omnichannel resolution. The AOP concept is powerful for teams that want full control without a programming language, but it requires more technical capability within the CX organisation than a platform like Sierra.

Ada focuses on self-service automation and multilingual support with high deflection rates. It's strongest for e-commerce and SaaS use cases where many customer questions can be resolved from knowledge base content without cross-system lookups. Ada is a good fit for teams with simpler support workflows that need broad language coverage.

Each platform occupies a different niche. Intercom Fin offers the most transparent metrics and a clean per-resolution pricing model. Zendesk AI is the obvious choice for Zendesk customers, especially with the Forethought acquisition strengthening its QA and knowledge capabilities. Salesforce Agentforce is unbeatable for deep CRM context. Sierra is the enterprise concierge option. Decagon gives technical CX teams the most direct control. Ada is the specialist for high-volume self-service.

All of these platforms are investing heavily in voice capabilities, multi-agent coordination, and tighter integration with the tools support teams already use. The competitive landscape is consolidating — expect more acquisitions like Zendesk/Forethought as platform vendors race to offer complete AI-powered support stacks.