The Problem
Building a presentation is rarely hard — it's just slow. Someone has to decide what each slide says, find or create images that fit, lay everything out so it looks intentional rather than thrown together, and then repeat that process for the next deck. For a client who needed to produce presentations faster and more often than manual design work allowed, that slide-by-slide effort was the actual bottleneck — not a lack of ideas, but a lack of time to turn ideas into a finished deck.
Automations Limited's founder, Mustafa Haider, took on this project to solve that specific problem: remove the manual design step entirely, so a presentation could go from idea to finished deck without anyone opening a slide editor.
The Approach
The goal was straightforward to state and harder to build correctly: type a single prompt, and get back a complete, styled Google Slides presentation — content, images, and layout included, with nothing left to fill in by hand.
That meant chaining three separate services that don't natively talk to each other into one coherent pipeline:
- OpenAI, to turn a prompt into slide-by-slide content and layout instructions.
- Unsplash's API, to source images that actually match what each slide is about.
- Google Slides API, to take that content, those layout instructions, and those images, and assemble a real, styled presentation.
None of these tools were built to work together out of the box. The job was designing the workflow that made them act like one system — orchestrated end to end in n8n.
How It Works
The finished workflow runs in five stages, triggered by a single input:
1. A single prompt goes in. The user describes what the presentation is about — a topic, a purpose, an audience — in one line, with no separate steps for content, images, or design.
2. OpenAI generates the slide content and layout instructions. Rather than returning plain text, the model is prompted to produce structured output: what each slide should say, and how it should be arranged — title placement, body content, image placement, and slide type. Treating layout as structured data (not just written text) is what makes the next two stages possible.
3. The Unsplash API sources matching images. Using the topic and context from each slide, the workflow queries Unsplash for high-quality images relevant to that specific slide, rather than relying on generic or repeated stock visuals across the deck.
4. The Google Slides API builds and styles the deck in real time. With content, layout instructions, and images all available, the workflow calls the Google Slides API to create the actual slides — applying the specified layout, inserting the sourced images, and formatting text — producing a deck that's already styled, not just a wall of unformatted content.
5. n8n orchestrates the entire workflow end to end. Every stage above — the OpenAI call, the Unsplash lookup, the Google Slides build — runs as one connected workflow in n8n. There's no manual handoff between steps; the prompt goes in one end, and a finished presentation comes out the other.
Tools & Stack Used
- n8n — workflow orchestration connecting every stage of the pipeline
- OpenAI API — slide content generation and structured layout instructions
- Google Slides API — real-time slide creation and styling
- Unsplash API — image sourcing matched to slide content
The Outcome
The result is one-click creation of a fully-designed, media-rich presentation from a single prompt. What used to require manually writing slide content, hunting for images, and laying out each slide by hand is now handled entirely by the workflow — reducing manual design effort to zero.
Lessons Learned
The key decision that made this workflow actually work was treating "layout" as data, not as an afterthought. Early on, it would have been easy to have the AI model just generate slide text and handle formatting separately with fixed templates. Instead, prompting the model to output structured layout instructions alongside the content — which section goes where, what type of slide it is, where the image belongs — is what let the Google Slides API build a genuinely styled deck in real time, instead of dumping unformatted text onto blank slides.
The second lesson was about integration strategy. OpenAI, Unsplash, and Google Slides each have their own API, their own data format, and their own quirks. Writing custom point-to-point integration code for each connection would have meant maintaining three separate, brittle pieces of glue code. Orchestrating all three through a single n8n workflow instead meant one place to manage triggers, error handling, and data mapping between services — which made the system easier to build, easier to debug, and easier to extend later.
If your team regularly builds presentations, reports, or other recurring content by hand, this same approach — prompt in, finished output out — can likely be adapted to your workflow. Book a free automation audit to talk through what that would look like for your business.