Gong ships a genuine fleet of AI agents, not a single one, and they all sit inside the revenue cycle Gong already records. The umbrella name is Gong Agents, the thing that governs them is the Gong Revenue Harness, and the capability that changed the product in 2026 is Custom Agents, which lets a RevOps lead build a governed agent without engineering.
If you are evaluating Gong for its agents rather than its call recording, the useful question is not whether the agents are good. It is whether the work you want automated happens inside the revenue conversations Gong already sees. Below are the actual product names, the 2026 launches, and the honest boundary.
What Gong actually ships under "AI agent"
Gong's agents are deliberately narrow. Each one does a defined job inside a workflow a revenue team already runs, which is a different design philosophy from a general agent you point at anything.
- AI Tasker, inside Gong Engage: surfaces high-impact, actionable to-dos at the moment they matter, from post-call follow-ups to nudging a quiet prospect or sending an intro before a first meeting.
- AI Composer, inside Gong Engage: drafts follow-up emails from meeting and call transcripts, and rephrases messages into a professional, friendly, or persuasive tone.
- AI Data Extractor: turns conversations into structured attributes, which is what makes call content queryable rather than just searchable.
- AI Deep Researcher: produces structured reports and handles complex, multi-step business questions rather than single lookups.
- Custom Agents: agents your own team defines, generally available since June 2026 and governed by the same harness as the pre-built ones.
Sitting under all of them is the Gong Revenue Harness, which Gong describes as the evolution of its Agent Studio and Model Context Protocol support. It governs, orchestrates, and connects agents across the revenue cycle. The pitch is explicitly aimed at the failure mode Gong sees in early enterprise AI: generic tools that break in production because they lack revenue context, governance, and human-in-the-loop control.
The two 2026 launches that changed the product
Gong shipped two named product missions in 2026, which makes the agent timeline easy to follow. Both matter for anyone evaluating agents.
| Launch | Date | What it introduced | Why it matters |
|---|---|---|---|
| Mission Andromeda | 25 February 2026 | Gong Enable, conversational guidance, unified account management, secure AI interoperability | Extended the AI OS from analysis into enablement and account management |
| Mission Big Dipper | 24 June 2026 | Gong Revenue Harness, Custom Agents, expanded Gong Assistant, new Gong Enable capabilities | Added the governance and execution layer, and put agent building in non-engineering hands |
The Gong Enable capabilities from Big Dipper are worth naming because they are the least agent-like and the most immediately useful: Dry Run, which Gong said would let reps launch a role-play rehearsal from an upcoming calendar invite with the customer persona replicated from real account context; AI Coach gives personalized guidance straight after an AI Trainer session; and AI Builder for Scorecards generates coaching scorecards from the criteria found in your own closed-won calls.
What we found researching this post
Two things stood out, and both are warnings for anyone comparing vendors in this category.
The first is that Gong's product surface is hard to verify from outside. The obvious guess at a product URL, gong.io/product/ai-agents/, returned a 404 when we probed it on 7 September 2026; the page that does exist sits at gong.io/platform/ai-agents-for-revenue-teams, which is the one cited below. The pricing page resolves but returns a lead form rather than a price. The reliable sources were the press releases and the product blog, which is the same pattern we hit researching other enterprise vendors: newsrooms stay fetchable when product pages do not. We took every product name in this post from Gong's own writing, because aggregator posts in this category routinely invent plausible-sounding agent names that the vendor has never used.
The second is the pricing opacity, and it has a practical consequence. Gong states three things publicly: licenses are priced per user, there is a platform fee based on the number of users supported, and integrating your existing stack is free. That third point is more useful than it looks, because in tools that meter connectors, integration cost is often where the quote inflates. What you cannot do is model Gong's total cost before talking to sales, so treat any specific per-seat figure you find in a comparison post, including a confident one, as unverified.
The gap: work Gong's agents cannot see
Gong's agents reason over what Gong records: calls, emails, meetings, and the CRM data attached to them. That is a large and valuable surface, and it has a hard edge. If the work does not leave a trace in the revenue conversation, no Gong agent can act on it.
The work that falls outside includes:
- Anything in finance, product, support, or operations that never appears in a sales conversation.
- Cross-tool reconciliation where the answer lives in your billing system and your project tracker rather than in a call transcript.
- Recurring internal reporting that spans departments, where the sales view is one input among several.
- Work for teams without Gong licenses, since the agents live behind per-user pricing.
There is also a structural point about scope. Gong's agents are narrow and governed by design. That keeps them inside the revenue cycle and unavailable outside it, which is a coherent trade for a revenue AI platform and a real constraint if you were hoping for one tool that covers the whole company.
How to decide, by situation
- You already run Gong for call recording. The agents are the main reason to revisit what your contract covers, and Custom Agents means you are not waiting on Gong to build the specific one you need. Ask Gong which of these your plan includes, since it does not publish that.
- You have a RevOps function but no engineers to spare. Custom Agents is aimed exactly at you: Gong positions it as buildable without engineering support.
- You are a small team without Gong. Per-user licensing plus a platform fee makes Gong an enterprise purchase. Start with lighter AI agents for sales teams instead.
- Your problem is CRM hygiene. AI Data Extractor helps inside Gong's view, but a dedicated Salesforce data hygiene agent covers records Gong never touches.
- You need work done outside sales. Gong is the wrong scope. A general scheduled agent platform is the fit.
Where Gravity fits next to Gong
Gong is a revenue AI operating system for teams that live in calls and pipeline. Gravity is a general agent platform for recurring work across whatever tools you actually use. They overlap far less than the shared word "agent" suggests.
The practical split: let Gong's agents handle what happens inside the revenue conversation, drafting the follow-up, extracting the deal attributes, coaching the rep. Let a Gravity agent handle the work that crosses the boundary, like pulling won deals into a weekly finance summary, checking that new customers were provisioned in the product, and chasing the internal owner when onboarding stalls. That second category is where teams usually discover the seam, because it needs to read more than one system.
Cost shape differs too. Gong is quote-based per user plus a platform fee, which suits a funded sales organization. Gravity is a subscription with a free tier covering one agent at $0 per month and paid plans from $20 per month that include $20 of usage, with extra usage available beyond your plan. For how those models compare across the market, see AI agent pricing explained and the cheapest AI agent platforms.
For sales-specific workflows, see agents for sales call follow-up. If you are weighing one broad agent against several narrow ones, single agent versus multi-agent covers the architecture question directly. Related vendor explainers: Workday AI agents and Ramp's finance agents.
Frequently asked questions
What AI agents does Gong have?
Gong describes Gong Agents as more than a dozen specialized agents. Gong names five of them publicly: AI Tasker and AI Composer inside Gong Engage, AI Data Extractor, AI Deep Researcher, and Custom Agents that your own team builds. We could not find a published list of the full set.
What is the Gong Revenue Harness?
An agentic execution layer introduced with Mission Big Dipper in June 2026. It governs, orchestrates, and connects AI agents across the revenue cycle, and Gong describes it as the evolution of its Agent Studio and Model Context Protocol work.
How much do Gong's AI agents cost?
Gong does not publish pricing. It states that licenses are priced per user, that there is a platform fee based on the number of users supported, and that integrating your existing tech stack is free. Everything else is quote-based.
Can you build custom AI agents in Gong?
Yes. Custom Agents became generally available in June 2026 and are designed so a RevOps leader or sales manager can build and deploy a governed agent without engineering support.
Can Gong's agents work on data outside sales calls?
Not meaningfully. Gong's agents reason over the conversations and CRM data Gong records. Work in finance, product, or operations that never appears in a revenue conversation is outside their view.
What is the difference between Gong Engage and Gong Agents?
Gong Engage is the sales engagement product where AI Tasker and AI Composer are embedded in daily rep workflows. Gong Agents is the broader umbrella covering all of Gong's specialized and custom agents, governed by the Revenue Harness.
Sources
- Gong launches Mission Big Dipper; unveils industry-first Revenue Harness, Gong, 24 June 2026.
- Gong launches Mission Andromeda, expanding its Revenue AI OS to enablement and account management, Gong, 25 February 2026.
- Meet Gong Agents: specialized agents that actually work for your revenue team, Gong product blog, by Eilon Reshef, Chief Product Officer and co-founder, checked 7 September 2026.
- Gong Agents platform page, checked 7 September 2026.
- Gong pricing, per-user licenses and platform fee, checked 7 September 2026.