AI can create presentations, but who owns the version?

AI can create presentations now. That is no longer the hard part.

The hard part shows up five minutes later, when someone asks: which version is approved, which sources were used, who changed the pricing slide, and can this file be sent to a customer?

Google has added support in Gemini Enterprise Business edition for creating and editing slide presentations in Canvas. The deck can be exported to Google Slides, PowerPoint, or PDF. One small line in the documentation matters more than the export button: when you ask Gemini to update a slide presentation, it creates a completely new Canvas artifact. The old one is not modified or versioned for you.

Source: Gemini Enterprise Business edition release notes

Source: Create and edit documents and slides in Canvas

At the same time, more AI work is moving into shared project spaces. Manus announced in May that Projects are available on mobile, with shared instructions, files, project-specific skills, isolated connector setup, and scheduled workflows. That makes project context more portable. Good. But the controls need to travel with it.

Source: Manus on Projects on mobile

Source: Manus on shared instructions and project context

The problem is not that AI makes presentations

An AI-generated presentation is often an intermediate artifact, not a finished decision. It may be based on chat history, uploaded files, active connectors, and sources someone added for that run. That is exactly why it can be useful.

It is also why you need to know what it was built from.

A product deck can pull an old pricing model from a folder. Training material can mix a new policy with an older version. A sales presentation can look polished but miss the objection the customer raised in the last meeting.

This is not about distrusting AI. It is about treating the presentation like any other business document: source, owner, version, and approval.

Put ownership outside the chat

If the version only exists inside a chat, it is hard to manage. Writing latest in the file name is not enough. Decide a few simple things before AI creates material that will be used externally or in training:

• Where is the source folder?
• Which facts are approved for use?
• Who owns the presentation?
• How are versions named?
• Where is the change log written?
• When may the file be exported?
• Who must review it before use?

This does not need to be heavy. For many Hammer customers, a simple routine in Drive, OneDrive, Notion, or the project tool they already use is enough. The point is that the version decision should survive the chat.

A simple routine for AI-generated decks

Start with one page. Not a 40-page policy document. One page the team actually uses.

Write this down before the first deck is created:

• Sources: link the folder, document, or customer material AI may use.
• Owner: name the person responsible for the content, not just the person who wrote the prompt.
• Version name: use date and purpose, for example 2026-05-27_customer-intro_v1.
• Change log: note what AI changed, what the human changed, and why.
• Export rule: export only after sources and version have been checked.
• Stop rule: if the material includes pricing, legal claims, student data, customer promises, or security information, require extra review.

This sounds almost too simple. But it is exactly the kind of friction that stops a polished but wrong presentation from taking on a life of its own.

When mobile makes it easier, the stop point needs to be clearer

The Manus signal is worth taking seriously even if you do not use Manus. When projects with files, instructions, and connectors move to mobile, AI work becomes easier to continue between meetings, on the train, or right after a customer call.

That is practical. It also makes careless work cheaper.

So the same routine should be visible where the work happens: in the project description, in the folder, in the task card, or in the first instruction to the AI tool. If someone can ask AI to update a deck from a phone, that same person should see which version is approved and what must not be forwarded.

Short answer: who owns an AI-generated presentation?

The business owns it, not the AI tool. In practice, a named person needs to own the content, the sources, and the decision to use the presentation.

AI can help write, sort, and format. It should not be the only place where you track version, changes, and approval.

When Hammer can help

This is a good fit for a small Tool Forge or Mindset Forge engagement: map how presentations, training material, or customer decks are created today, then set up a lightweight routine for source, version, export, and approval.

If AI is already moving into your sales, training, or operations material, now is a good time to fix the workflow. Before the first wrong version gets forwarded.