Scale Your Presentations Without Manual Work

Full video transcript

Hey everybody, Darby here with GenAI University, and today’s part three of our Gamma Mastery series, where we’re going into a case study on using the Gamma API to automate presentations and content at scale — how you can leverage the Gamma API inside of Zapier specifically in this example, and how you can get started using the Gamma API today to automate your marketing, sales, and operational content, turning ideas and documents into beautiful presentations, infographics, and more.

The building blocks we’re walking through today build on top of part one and part two of the mastery series, which lay a foundation for what you need in Gamma to generate any sort of branded presentation, document, or social post in a certain style guide. That’s what we’re going to tap into today with the API. Gamma came out with their Gamma 3.0 release, where you can now use the Gamma API to generate content through automations. There are a lot of different use cases, and I hope you can see a direction you can take this for your own business.

Here’s a Google Drive folder with a few pieces of a knowledge base used for the automation, plus a Google Form and Google Sheet managing the intake of data, which Zapier was monitoring to create outputs. Inside the Gamma dashboard, there’s an API Generated tab — if you click it and see nothing, it’s because you haven’t generated anything with the API yet, but once you start, outputs appear there. This particular deck took about three minutes from intake-form submission to a completed generation, without the presenter touching anything. It’s treated as a first draft that aligns with the given frameworks, which he’d review, modify, and pass along to the community member or client.

The document is 15 pages: a roadmap analysis based on the business owner’s top three challenges, an “understanding your business” slide reiterating the intake info, a target-audience snapshot, a single leverage point, three quick wins for using AI, a 7-day starter plan, a 30-day roadmap for transformation, other recommended AI workflows, a priority pyramid, a current-state/desired-state analysis with risks and hidden opportunities, and closing slides on available support and an immediate action plan. This framework could be adapted for other use cases, like a fitness or nutrition coach generating a custom plan instead of an AI roadmap.

The idea grew from a common problem: personalizing onboarding for community or client members. The trigger point was a Google Form asking role, what the person does, and their top three challenges. For the demo, the presenter cloned an intake for his other business, Side Hustle Games, describing it as live events and comedy-show experiences, with challenges around getting sponsor partners, building an audience through YouTube and social, and getting more people to play the game more often.

Once the form is submitted, the data flows to a Google Sheet, which serves as the trigger for a Zapier agent — the “decision intelligence generator.” He built this agent in Zapier using Claude to help design the workflow. Once a new row is triggered, the agent receives the member’s name, role, top challenges, and website; analyzes the business (a website URL helps it learn more); categorizes primary challenge areas; applies the presenter’s own method; and generates a comprehensive decision-intelligence profile matching the structure of the sample deck.

After that analysis step, the agent moves to generating the Gamma deck via the Gamma API. To use the API, you need a Gamma Pro plan; go to Settings, then Members, then API Keys, and create a key, which you then paste into Zapier’s Gamma connection. Every generation uses Gamma credits.

The agent saves the resulting generation ID back to the same spreadsheet row — an important detail, since a second, separate Zapier workflow (set on a delay) uses that ID to retrieve the finished deck. The first workflow also sends an immediate templated HTML email summarizing what was fed into Gamma, since Gamma decks take some time to generate (roughly 90 seconds or so in this workflow) and the agent step itself has no way to wait for that. So a second Zapier workflow monitors the generation-ID column, waits about three minutes, searches the Gamma account for that ID, retrieves the completed deck’s share URL, writes it back to the sheet, and then sends a Slack notification and an email so the presenter knows exactly which generation belongs to which person — useful when running many generations at once.

Getting the Gamma generation step configured in Zapier starts with aligning to a saved template — in this case, a custom “Market Sauce” theme using GenAI University colors, covered in part one of the series. Inside the Zapier Gamma action, you connect your Gamma account and configure input text, text mode, formatting, generation settings, theme, custom instructions, text tone, and image style. The integration also shows image models and their credit costs — Image 3 Fast used about 2 credits per image, while Ideogram 3, which the presenter used for his initial testing, used about 20 credits per image, which explains why he burned through credits quickly. You can also select card dimensions, the destination folder in your Gamma account, and workspace edit access for internal and external users.

He then ran another generation for a different example (referred to as “George’s”), showing the agent complete the full process — running the framework, formatting for Gamma, generating the deck, and sending the HTML email — and pulled up the resulting “decision intelligence roadmap” deck to show its image style, distinct from the earlier, more expensive image model.

Opening the deck’s “view source prompt” shows all of the instructions that were sent to Gamma: additional instructions, tone, theme, and image style. The custom theme used is named “GenAI Market Sauce Evolved,” built using the standard theme tools covered in part one. Format options include presentation, webpage, document, or social, and a custom style prompt shaped this deck’s look (“professional yet cool, fun, innovative, modern doodle art with black stick figures”). You can also opt out of images entirely, or choose from several AI image models.

The Zapier workflow also needs tone and target-audience fields — if left blank, the AI infers them from context, which the presenter suggested experimenting with. Additional instructions (limited to roughly 2,000 characters in this setup) were useful for keeping naming and outputs consistent across generations. Content depth is adjustable from minimal to concise to extensive; this example used extensive. The “generate, condense, or preserve” setting matters for cases like turning a fixed 10-page SOP into 50 branded decks without letting AI rewrite the content — for that, you’d choose “preserve.”

A key piece tying it together is giving the Zapier agent a knowledge source. Without business context, an AI agent won’t know your specific frameworks or details to reference, so the presenter synced this workflow’s knowledge source to the same Google Drive document referenced earlier, containing community information and a working knowledge base, to influence how content was produced.

With all of this in place, every new form submission creates a new row, triggers the analysis agent, generates the Gamma deck, waits via the second workflow, and finishes with a Slack and email notification containing the share link — a three-minute generated draft of a 15-page presentation.

In Q&A: Asked whether the same process works with Microsoft Excel and Zapier instead of Google Forms, the presenter said he hadn’t tried it directly but that you could add a Microsoft Excel tool in Zapier as the trigger instead of a Google Sheet, and that Zapier connects to most form tools (Typeform and others), so the intake side is flexible as long as Zapier has access to that tool.

Asked about using a similar flow for a responsive single-page dashboard, he confirmed you can customize how many pages or slides Gamma outputs — for example, generating a one-page deck styled like a dashboard, since you can embed content into a page and customize it.

A participant (Sawen) asked where to start, having joined partway through the series; the presenter pointed to part one (setting up your theme and brand/style guide) and part two (turning existing SOPs or content into marketing collateral), noting that once you have a consistent branded theme, you can reuse it across API-driven presentations, documents, and social posts.

Reflecting on the build, the presenter said his biggest takeaway was that a Zapier agent can use context from an intake form combined with a company’s own knowledge base to produce a strong first draft, even though there’s still a place for a human in the loop to review and refine it before it’s sent to someone as part of onboarding or coaching. He noted this generated first draft saves time and gives clarity on which levers matter, and that from there you might point someone toward setting up a Claude project or using Gemini for more nuanced follow-up recommendations, or apply the same idea to something like personalized nutrition or coaching plans based on someone’s specific situation.

Asked why he uses Zapier instead of Make or n8n, he said it comes down to personal preference: Zapier’s agent interface, including its conversational co-pilot for building and editing workflows, was easier for him to get this version-one workflow off the ground, even though Zapier can be more expensive at scale than alternatives, and he might rebuild the workflow in Make or n8n later to reduce token costs once it’s proven out. He also mentioned Zapier’s “pods” feature for coordinating multiple agents, and described another example workflow that scans his Google calendar each morning, summarizes upcoming sales calls, researches the companies and speakers involved, and sends a Slack summary — built in a similar way with a detailed prompt refined through back-and-forth with Claude.

He closed by thanking attendees for tuning in and pointing them to the community for more resources.