Plus AI

Every AI deck tool solves the same first problem — producing slides quickly — and almost all of them create the same second problem.

The deck exists somewhere else. It arrives in Google Slides or PowerPoint as an export, which means the master template is wrong, the brand theme is absent, the text boxes are positioned rather than placed, and every edit from then on is a small fight with a file that was never native.

Plus AI's answer is to skip the export. It runs as an add-on inside Google Slides and an add-in inside PowerPoint, so the slides it makes are ordinary slides from the moment they appear.

Why Does Living Inside Google Slides Matter?

Plus AI installs as a native add-on in Google Slides and a native add-in in PowerPoint, which sounds like a distribution detail and is actually the product's whole argument.

Think about what happens after a deck is generated. Almost nobody presents a first draft. Somebody moves a chart, someone else objects to slide four, a colleague adds two slides from another deck, and the whole thing is reviewed by three people who will each change something.

All of that happens in Google Slides or PowerPoint, because that is where the collaboration, comments, version history and sharing live. A generated deck that arrives as an export has to survive that process as a second-class citizen.

Working natively means it never becomes an export. The slides use your existing theme and master, your colleagues can edit them without knowing a tool was involved, and comment threads and version history work exactly as they always did.

The corresponding constraint is real: you inherit the limits of the host application. Anything Google Slides cannot do, Plus AI cannot do inside it, and a tool with its own rendering engine can produce layouts neither host supports. That is a genuine trade, and for most business decks it favours being native.

The opposite bet is a tool that owns the whole artefact. Gamma keeps the deck in its own editor and exports as a secondary path, which suits a deck that lives on the web rather than in a shared drive.

Nobody presents a first draft. The question is what happens to the deck during the four rounds of edits that follow, and that all happens in Slides or PowerPoint.

Remix, Insert And Rewrite Are Three Different Jobs

Plus AI separates its editing actions into Insert, Rewrite and Remix, and that separation is better product design than a single "improve this slide" button.

Insert adds a new slide that fits what is already there — the deck is missing a section, and the tool writes one consistent with the surrounding argument. This is the action people use most, because decks grow by accretion rather than by planning.

Rewrite changes what a slide says while leaving its structure alone. Too wordy, wrong tone for the audience, needs to be blunter — the layout is fine and the words are not.

Remix does the inverse: same content, different arrangement. A slide with five bullet points might work far better as a two-column comparison or a simple diagram, and reformatting is a design decision most people cannot make quickly.

Keeping these apart matters because it lets you say what you actually want. A single general command has to infer whether you meant the words or the layout, and it will sometimes change both when you wanted one — which is how a slide you were happy with becomes one you have to fix.

An MCP Server Is An Unusual Thing To Ship

Plus AI lists a Presentation Agent API and an MCP Server alongside its ordinary features, and both signal something about where the product thinks it is going.

MCP is the protocol that lets an AI agent call external tools in a standard way. Shipping an MCP server means an agent — one you are using in a different application entirely — can create and edit presentations through Plus AI without a human opening the add-on.

The use that follows is worth thinking about. A weekly report that currently requires someone to read a dashboard and build a deck could instead be produced by an agent that pulls the data, writes the narrative and generates the slides on a schedule.

That is a genuinely different product from an assistant you open when you need help. It moves decks from something a person makes to something a process produces, and the person's job becomes reviewing rather than building.

It also has an obvious failure mode. Automatically generated decks nobody asked for are worse than no decks, and a weekly presentation produced by a schedule rather than a need is a meeting that should have been an email with extra steps.

For the diagrams inside those slides, a specialist like Napkin AI generates from prose and exports SVG, which is a different job from arranging a slide.

• Prefer native add-ons over exporters for anything a colleague will edit.
• Use Insert for gaps, Rewrite for words, Remix for layout — saying which saves a fix.
• Start from a document you already wrote rather than from a blank prompt.
• Set up brand and custom instructions once, since they apply to everything after.
• Automate a recurring deck only where someone actually reads it.

Document To Presentation Beats Prompt To Presentation

Plus AI converts documents into presentations, and this is a better starting point than a prompt for a reason that applies to every deck tool.

A prompt asks the model to invent both the content and the structure. It will produce something plausible, generically organised, and containing nothing you know — because you did not tell it anything.

A document contains your actual argument, your data and your conclusions. Converting it is a structural task rather than a creative one: identify the sections, decide what deserves a slide, condense the prose, keep the order that made sense in writing.

That is exactly the sort of transformation a model does well, and it is also the task humans do badly under time pressure — most bad decks are bad because someone pasted paragraphs onto slides rather than rethinking them for the format.

The practical consequence is that the best input is a document you have already written and edited. If the source is thin, the deck will be too, and no amount of reformatting fixes an argument that was not there in the first place.

If the source document does not exist yet, the research step comes first — an answer engine such as Felo is solving that half rather than this one.

Building A Deck That Survives Review

A generated deck that gets rebuilt by hand in review has cost you time rather than saved it.

The sequence below front-loads the decisions that reviewers otherwise force you to redo.

1. Set up custom branding and custom instructions before generating anything, so the first draft already looks like your organisation's decks.
2. Start from a written document rather than a prompt, and edit that document first.
3. Generate, then immediately delete the slides that exist only because the source had a section — decks are shorter than documents.
4. Use Remix on the two or three slides carrying the most information, since dense slides are where layout matters most.
5. Share it in Slides or PowerPoint for comments rather than exporting a PDF, so edits happen in the file rather than in an email thread.
6. Keep the generated speaker notes only if you will actually use them, since unused notes make a deck harder to maintain.

Plus AI Without The Pitch

• It runs inside your slide tool — a native Google Slides add-on and a native PowerPoint add-in, not an exporter.
• Three distinct actions: Insert adds new slides, Rewrite changes what a slide says, Remix reformats how it looks.
• Document to presentation turns an existing report or brief into a deck rather than starting from a prompt.
• Operated by Plus Docs, Inc., with SOC 2 Type II compliance stated.
• A Presentation Agent API and MCP server, which is unusual for a deck tool and points at programmatic use.

Plus AI: Platforms And Compliance