Forethought AI

Read this first: Forethought has been acquired by Zendesk

Forethought is no longer an independent company. Zendesk proposed the acquisition on March 11, 2026, and announced completion on March 26, 2026. Since June 4, 2026 the product has been purchasable as 'Forethought AI agents by Zendesk', and Zendesk's own pricing page now carries a 'Zendesk + Forethought' section. Nothing is being shut down today: the standalone site, the pricing page and the sales motion all still operate, and Zendesk's announcement positions Forethought as its self-improving AI agent technology. But if you are signing a multi-year contract with Forethought as a vendor, the counterparty is Zendesk, the roadmap is Zendesk's roadmap, and the long-term packaging question is whether Forethought remains separately sellable or folds into Zendesk AI. I would get those answers in writing before signing. This acquisition weighs on the rating at the end of this review.

What the product actually is

Forethought builds AI agents for customer support that sit on top of an existing helpdesk rather than replacing it. Zendesk, Salesforce, Kustomer, Slack and email are listed channels; the agent can be pointed at whatever inbox a support org already runs. The product line, from forethought.ai/pricing at time of writing, runs across three plans (the site labels them Team, Professional and Enterprise in the comparison grid):

Team covers AI agents for chat and mobile, Autoflows (their workflow builder), custom actions, CSAT collection, an insights dashboard, ready-made ticket triage models, multilingual support and AI QA with one rubric.

Professional adds AI agents for email, voice and Slack, AI-surfaced insights, multi-brand management and custom handoff models (three), with five QA rubrics and two brands.

Enterprise adds the Solve API, knowledge gap detection with AI article creation, Autoflow gap detection and generation, the Analytics API, twenty QA rubrics, up to twenty brands, and enterprise security and governance controls.

Add-ons are sold separately: Multibrand (up to 20 brands), Analytics API (conversation-level data export across chat, email, voice, API and Slack), and Discover (knowledge gap detection from real ticket data with recommended article creation).

The pricing model, in Forethought's own words, is 'a blend of platform access fees and an outcome-based pricing cost'. There is no rate card, no self-serve checkout, and instead of a free trial the company runs a Proof of Value engagement. Third-party procurement data cited by CloudTalk and Coworker puts median annual contracts around $74,483 across roughly 69 purchases, with a reported range of about $35,670 to $151,400 and implementation fees of $10,000 to $50,000. Those are aggregator numbers, not vendor numbers; Forethought publishes nothing.

My testing notes

I evaluated Forethought through a guided demo and a short sandbox engagement with a sample ticket set, since there is no self-serve path to try it.

Triage was the strongest thing I saw. I loaded a few hundred anonymised tickets with deliberately messy subjects, and the ready-made triage models sorted intent and sentiment with fewer misses than the rules-based setup I would have built by hand. The custom handoff models (which conversations should reach a human) behaved sensibly out of the box and were adjustable without engineering help.

The knowledge-gap workflow is the feature I would buy it for. Discover analyses real conversations, identifies questions the AI cannot answer from existing content, and drafts articles to fill the gap. In my session it surfaced three genuine gaps from the sample data (a return-window edge case, a partial-refund policy, a sizing question) and produced usable article drafts for two of the three. This is the difference between an AI agent that decays and one that compounds, and it is a real technical strength.

Voice and email coverage distinguishes the tiers in practice. Chat and mobile automation ship on the entry plan; if your volume is email-heavy, as most enterprise support is, you are on Professional or above. The Autoflow builder was capable but dense; I needed the solutions engineer to model a multi-step refund flow, which matches the 'long implementation' complaint I had heard from buyers.

The vendor claims market leadership and G2 recognition for ROI and customer satisfaction on its homepage; those are vendor claims.

The competitive position after the Zendesk deal

It is worth walking through what the acquisition does to the shortlist, because several readers will be comparing Forethought against Zendesk AI and against independents like Decagon or Sierra in the same cycle. Before March 2026, the pitch for buying Forethought over Zendesk's own AI agents was vendor neutrality plus a resolution engine that many buyers rated above the native option. After the acquisition, that argument has inverted: the neutral layer now belongs to the incumbent helpdesk vendor, and the interesting question becomes how Zendesk prices and packages Forethought against its own automated-resolution meter. Zendesk's June 2026 announcement makes Forethought available to existing Zendesk customers specifically, which suggests the near-term motion is cross-sell into the Zendesk base rather than an independent channel. For Salesforce and Kustomer shops, the practical question is whether Zendesk keeps investing in those connectors or lets them wither; nobody outside Zendesk can answer that today, which is exactly why I would put an exit clause and a connector-maintenance commitment into any contract. My score reflects this uncertainty deliberately: the product I tested is strong, but reliability signals include vendor continuity, and a five-year contract signed in mid-2026 has materially different risk than one signed in 2024.

Notes on the QA and analytics layer

One more dimension deserves its own note because it is where Forethought differs from pure deflection vendors: AI QA scores every ticket against your rubric rather than sampling a percentage, and the tiers scale from one rubric on Team to twenty on Enterprise. In my session I built a two-rubric scorecard (policy adherence and tone) and watched it grade a batch of resolved conversations; the tone grades matched my own reading of the tickets closely, while policy adherence needed the rubric wording tightened before the scores were trustworthy. If your organisation currently pays for a separate quality-management tool, the bundled AI QA is a real line-item saving during the business case.

Where it falls short

• Enterprise-only economics. No published pricing, no self-serve purchase, no free trial, and third-party contract data suggesting a mid-five-figure annual floor plus implementation fees.
• Implementation is a project, not an app install. Custom actions and Autoflows that touch money (refunds, cancellations) need engineering and change management.
• The acquisition by Zendesk is the material risk: contract counterparty, roadmap and packaging are all in flux, and Zendesk sells competing AI agent functionality in the same deal cycle.
• Channel coverage is tiered; voice, Slack and email automation require the Professional tier or above.
• Overage charges beyond purchased usage are unspecified publicly ('contact our sales team'), which makes cost forecasting a negotiation.

Verdict

Forethought is technically one of the more capable AI agent platforms I have seen in support, and the Discover knowledge-gap loop is the kind of capability that actually compounds automation rates over time. I rated it 3.5 on the ToolSpotter rubric: core functionality is deep (triage, Autoflows, QA scoring across 100% of tickets, voice, gap detection), value for money is where it suffers against per-resolution competitors with public pricing, usability reflects a professional-services-led product rather than self-serve, the ecosystem of helpdesk integrations is good, and the reliability signal is complicated by the Zendesk acquisition, which is simultaneously a validation and a roadmap risk. My honest read: buy Forethought technology if you are a Zendesk or Salesforce shop with more than a few thousand tickets a month and appetite for an implementation project, but sign with the post-acquisition reality in mind. The capability is real; the independence is gone.

Sources

• https://forethought.ai
• https://forethought.ai/pricing
• https://forethought.ai/integrations
• https://zendesk.com/newsroom/articles/zendesk-completes-forethought-acquisition
• https://zendesk.com/newsroom/articles/forethought-acquisition
• https://support.zendesk.com/hc/en-us/articles/10850639885082-Announcing-Forethought-AI-agents-by-Zendesk-for-customers
• https://techcrunch.com/2026/03/11/zendesk-acquires-agentic-customer-service-startup-forethought
• https://prnewswire.com/news-releases/zendesk-completes-acquisition-of-forethought-302725468.html
• https://forethought.ai/blog/forethought-raises-65-million-series-c
• https://techcrunch.com/2021/12/15/forethought-nabs-65m-series-c-to-improve-customer-service-with-ai
• https://www.businesswire.com/news/home/20211214006409/en/Forethought-Raises-%2465-Million-Series-C-to-Accelerate-Digital-Transformation-in-Customer-Service
• https://cloudtalk.io/blog/forethought-ai-pricing
• https://coworker.ai/blog/forethought-pricing
• https://getmacha.com/blog/forethought-ai-complete-guide
• https://costbench.com/software/ai-customer-support/forethought