Forethought AI agents by Zendesk: A 2026 overview and alternative

Disclosure: This article is published by eesel AI, a competitor of Zendesk. We encourage you to read Zendesk's own materials for their perspective.

On March 11, 2026, Zendesk completed the acquisition of Forethought, an AI agent platform for enterprise customer support that had been building since 2018. The deal terms were not disclosed, but they closed a chapter that included $115 million in total investment and a TechCrunch Battlefield win in the company's founding year.

Forethought's homepage now carries the announcement: "A new chapter begins: Forethought is now part of Zendesk." Zendesk has said it plans to integrate Forethought's technology into its Resolution Platform, accelerating its AI roadmap by over a year.

That context shapes what you're evaluating. Forethought is a Zendesk product now, with a roadmap driven by Zendesk's priorities and the deepest integrations pointing at Zendesk's own infrastructure.

This article covers what the platform does, how each agent works, what pricing looks like across three tiers, and where eesel AI fits for teams that want a transparent alternative outside the Zendesk ecosystem.

What Forethought AI agents by Zendesk is

Forethought is an AI agent suite for enterprise customer support. It doesn't work like a single chatbot you attach to a contact form. Instead, it ships four agents that each handle a different phase of the support lifecycle, feeding data between them rather than operating independently.

Co-founder Deon Nicholas put the founding thesis this way: "More than seven years ago, we set out with a simple but ambitious idea: AI could transform the customer experience." By acquisition time, the platform was supporting more than one billion monthly customer interactions.

Enterprise customers included Upwork, Grammarly, Airtable, Datadog, Fetch Rewards, Lattice, and Brex - primarily large support operations running high ticket volumes and existing Zendesk infrastructure.

The four core agents

Forethought divides support work into four roles. Each handles a specific phase, and they are built to connect with each other rather than act as standalone tools.

Solve

Solve is the autonomous resolution agent. When a ticket arrives, Solve reads the ticket content, retrieves the customer's history, searches the knowledge base, and decides whether to resolve the issue on its own or route it to a human.

When it resolves, it sends the reply, updates the ticket status, and closes the ticket. When it escalates, it passes the ticket with context attached. You configure the confidence threshold: below it, Solve routes to a human rather than attempting resolution.

Forethought reports that Solve achieves up to 98% resolution rates and reduces average first response time by 55%. The agent handles chat, email, voice, and social from one unified interface.

Solve connects to 70+ platforms including Salesforce Service Cloud, HubSpot, and Freshworks. The native Zendesk integration is the deepest.

Triage

Triage classifies and routes incoming tickets before any agent handles them. It reads ticket content and applies classification models to identify issue type, priority level, and where the ticket should go.

The Basic tier includes pre-built models covering common issue categories that work out of the box. Professional and Enterprise tiers add custom model training: you provide labeled historical ticket data, and Triage learns your organization's specific taxonomy. The classifier improves with more data and feedback.

QA rubrics let you define accuracy checks for the classifier. Basic supports 5, Professional supports 20, and Enterprise supports unlimited.

Assist

Assist is a real-time copilot for human agents. While a support agent is working a ticket, Assist appears as a sidebar panel showing suggested responses, relevant knowledge articles, and recommended next actions. Agents accept suggestions with a keystroke. The system tracks which suggestions were accepted and which were modified, and uses that feedback to improve.

In the Zendesk agent workspace, Assist appears as a native sidebar integration. No context switching required.

The system learns from your knowledge base content, your support policies, and your agents' editing patterns over time. Junior agents get in-context coaching rather than having to ask separately for guidance.

Discover

Discover analyzes patterns across your ticket history to surface gaps in your support operation. It identifies customer questions with no matching knowledge base article, recurring issue types that could be automated, and sudden spikes in a complaint category that might signal a product problem.

The Enterprise tier adds auto-drafting: when Discover identifies a knowledge gap, it can generate a draft article for a human to review and publish. In the base tier, Discover flags the gap; in Enterprise, it can propose the fix.

Three newer additions

In the months around its acquisition, Forethought shipped three extensions that broadened its autonomous capabilities.

Orchestrator (released April 3, 2026) coordinates across the full agent suite to optimize for business outcomes. Instead of treating all tickets the same way, Orchestrator reads signals like customer tier, churn risk, and issue type, then applies business logic to choose the right workflow. A high-value customer with a billing issue might be routed to a senior human agent with a priority flag; a standard FAQ gets resolved autonomously by Solve. You define the signals and goals; Orchestrator applies the routing.

Test Suite (released April 6, 2026) lets teams validate agent behavior before deploying changes to production. You write test scenarios - mock customer interactions with expected agent outcomes - and run them against current and proposed agent configurations. If a new Triage model change breaks an existing routing rule, Test Suite catches it before customers are affected. Tests cover behavioral assertions, response accuracy, and performance benchmarks.

Browser Agents (released April 29, 2026) extend autonomous resolution to tasks that APIs cannot handle. The agent controls a real browser instance to take action in legacy portals or third-party systems with no exposed API: navigating an admin portal to reset a customer account, filling a form to update billing information, or retrieving a document from a system that only exposes it through its UI. Browser-based automation runs in 5-30 seconds per task compared to milliseconds for API calls, and it is more fragile when the target UI changes. Browser Agents are suited for low-volume, high-value edge cases where no API path exists.

Pricing

All three Forethought tiers are sales-gated. No prices are published. The billing model combines a platform access fee with a committed usage cost based on deflection volume.

Source: Forethought via Zendesk

Annual or multi-year commitments are standard at the Professional and Enterprise levels. Getting to an actual number requires talking to sales. There is no self-serve trial and no pricing page with published rates.

Setup and implementation

A basic Zendesk connection - enabling Solve to read and reply to tickets - takes around 30 minutes. Intermediate setup, which adds custom Triage model training on historical ticket data and Assist integration for human agents, typically takes one to two days. Advanced configurations involving multi-brand support, complex autoflow rules, webhook integrations, and SLA-aware routing can take one to two weeks.

Forethought is not a self-serve product. Engaging Zendesk sales is the starting point, followed by an onboarding process covering integration setup, model training, and configuration review. Teams without dedicated support operations resources should budget implementation time accordingly.

Custom Triage model training requires labeled historical ticket data and iteration. Building autoflow rules for complex support domains requires domain expertise. The investment pays off at scale; at lower volumes, the effort-to-return ratio shifts.

Who this fits

Forethought is built for large enterprise support teams with high ticket volumes, existing Zendesk infrastructure, and the internal capacity to configure and maintain AI agents. The platform's reported 15x ROI reflects enterprise-scale deployments where deflection volume is high enough to absorb the setup and licensing costs.

For teams not already running Zendesk, the post-acquisition product direction is worth factoring in. Zendesk's stated roadmap focuses on deeper Resolution Platform integration through 2026 and into 2027. Teams choosing Forethought today are, in practice, committing to the Zendesk ecosystem.

An alternative for teams outside the Zendesk stack

eesel AI is an AI agent platform built to work across helpdesks rather than within one. It connects to Zendesk, Freshdesk, and more, answers tickets using your knowledge sources, and can be running in under 15 minutes.

Pricing is public: $0.40/task, with $50 in free credits on signup and no credit card required to start. There is no deflection-volume commitment and no sales process before you see a number.

For teams that need Zendesk-native orchestration and are committed to the Zendesk ecosystem long-term, Forethought's depth is a real advantage. For teams that want transparent pricing, a faster path to production, or support across multiple helpdesks without being tied to one vendor's roadmap, eesel AI is worth evaluating.