Scaling revenue AI beyond the basics with Gong Credits

When Amit Bendov and I started Gong over 10 years ago, we envisioned a new way to run revenue organizations. We started building a system that drives GTM productivity by leveraging real revenue data: actual customer conversations with customers — calls and emails.

Times were very different then. We had a strong conviction that technology would mature in a way that transforms how revenue organizations operate, and that no matter how AI evolved, it would depend on access to the right data. That belief led us to focus on capturing this critical data early on. Our working assumption was that no matter how AI would shape up, it would need access to the right data.

So we embarked on collecting the critical data: customer conversations. It’s hard to imagine now, but call recording was not available at the time, and we were forced to build a robust data capture layer ourselves. In fact, we still own a patent for recording web conferences.

At the same time, the AI landscape itself was still in its early stages. AI was also not a household term at the time, and the tools we take for granted today simply weren’t available.

Back then, we brought in data scientists to help us build models that understood the spoken word: transcription, summarization, classification, and more. Those models were hard to build, and compared to today’s world — hard to use. To teach Gong to “tag” conversations, RevOps professionals had to provide us with dozens of examples — a process that can be accomplished now with simple natural-language prompts.

Fast forward to today’s widespread availability of LLMs. We have introduced multiple agents that automate all the manual tasks that get in the way of driving revenue outcomes: they review calls, extract data, create content, find themes, and more. And, as our customers become more literate in operating such agents, they are setting up even more agents and rolling them out to automate more parts of their organization.

As an example, we introduced question-based AI Trackers a few months ago. This made our AI Tracker agent easier to use, letting customers find and tag revenue concepts with natural language instead of manually training the agent by providing examples. In the few months since launch, customers have created more AI Trackers than they did in the previous four years using pre-LLM technology.

Deploying AI agents used to require a serious technical lift. Now, it’s a few prompts. This shift drove us to rethink how we package Gong AI. Historically, we limited the amount of AI processing that customers can set up within Gong. That made sense when it was harder to build. But now, when customers can configure and deploy dozens of agents with simple prompts, we wanted to make sure we don’t let infrastructure economics (i.e. LLM costs) be the ceiling on what our customers can automate.

That’s why we are introducing Gong Credits, a currency that gives customers a controlled way to operate current and future agents beyond basic use.

We’ve built the credit model to align with the work that Gong does on behalf of customers, rather than force customers to think in terms of low-level tokens. Credits are consumed based on the volume of data Gong agents process — calls, emails, and more. For example, when a customer extracts structured data from a year’s worth of interactions, more credits would be consumed than for just a single month.

We’ve modeled the Gong system with customer patterns in mind:

• Credits are exclusively used for Gong Agents performing work behind the scenes, and not for AI that helps everyday user-facing tasks. Common activities such as writing and rephrasing emails, preparing for meetings, and using Gong Assistant are covered by customers’ existing Gong license, ensuring they only consume credits for advanced automation.
• Every customer receives a pool of credits included with their seats, sized to cover standard use of Gong agents. When customers like to expand beyond that standard and automate more work, they can purchase more credits. This means that the majority of customers would only need to purchase credits once they’ve decided to use agents above and beyond their current use.
• We’ve priced Gong Credits so that they are comparable — and usually cheaper — than running the same processing using third-party models such as Anthropic or OpenAI. Customers often export raw transcript data into external AI tools to generate insights and automate workflows. But the moment data leaves Gong, it loses the rich context we've spent years mapping across your accounts, deals, and interactions. Teams then spend time and tokens rebuilding that context in LLMs, but even then, the outputs are less reliable. They end up paying twice: once to capture the data and context in Gong, and again to reconstruct what Gong already knew. Customers who run agents on Gong versus a horizontal provider are likely to reduce their AI spend.
• We’ve included multiple mechanisms for transparency and control, to help customers control their costs. The Gong Credits admin center gives a clear view of credit balance and usage. When configuring an agent, in-product guidance provides an estimate of the ongoing agent cost.

With Gong Credits, the barrier to scaling AI is gone. Credits apply across AI Trackers, our agentic APIs, and MCP — and we're opening up our pre-built agents, like AI Call Reviewer and AI Data Extractor, to the same model. The more you use, the more credits you consume. No caps or hidden friction, just a straightforward way to run as many agents as your business needs.

Ten years ago, Amit and I operated on the core premise that capturing the right revenue data would be the foundation of driving productivity and intelligence within revenue organizations.

Now that we are able to not only capture data but also apply agentic AI to streamline revenue work, Gong Credits is a natural evolution that reduces barriers to such use by customers. We are excited to see how far customers want to take it.

To learn more about Gong Credits, visit our FAQ.