How OpenAI Could Dethrone Google Docs
A speculative forecast
Last week the news leaked that OpenAI were working on a rival to Microsoft Office and Google Workspace. I wasn’t too optimistic in my initial reaction. The incumbents will have the advantage because I believe that the core paradigm of document suites won’t change dramatically with the introduction of Generative AI.
The big three pieces of software in any office suite - the word processor, the spreadsheet and the presentation - have been so foundational and unwavering since the late 1970s, that I find it hard to imagine Generative AI changing them enough to create space for a new entrant into the market.
But maybe that’s just a failure of imagination on my part! So in this post I want to push myself to imagine what it would look like if Open AI succeeds?
What Success Will Look Like
Let’s imagine that that year is 2030. OpenAI have managed to peel away 50 million of Microsoft’s 400 million paying Office customers. Or maybe they have peeled away 500 million of Google’s 3 billion free Workspace users.
This is far more than what others have done in the past. OpenAI haven’t peeled away a small, specific use case from one of the big 3 - like CV writing from Word, or KPI dashboards from Powerpoint. They will have built a suite of general purpose document software which is on track to dethrone the incumbents.
They’ve done this because not because their new software is better at traditional office software things, but because their new GenAI features are so good that it makes up for the fact that they’re not as good at the old stuff.
I think this can only happen if OpenAI have done one of two things:
With GenAI, they have changed the workflow around document creation so substantially that a better AI-first approach wins
GenAI has fundamentally changed one of the 3 atomic document types - the spreadsheet, the doc or the presentation.
Step 1: They Own The Start
The most difficult part of writing a Word Doc or a Powerpoint deck is often getting started. Staring at the dreaded blank page.
Microsoft and Google have traditionally tried to help users overcome this cold-start problem with templates. It’s clear that describing what you want to ChatGPT is very often a superior experience for starting a word document, and it’s not hard to imagine how the same will become true for presentations and spreadsheets too.
Will Microsoft and Google also do this? Sure, but people are already spending so much time in ChatGPT that it will become the most natural place to start. It will become the default, go-to app for hundreds of millions people when they want to ask a question, find a solution or start a document. It becomes muscle memory.
This is not too hard to imagine. I even do this quite often now, but then I inevitably need to copy/paste the ChatGPT draft into Google docs for editing. So they’ll need to break that behaviour next…
Step 2: Conversational Edits Will Be More Natural
In 2025, editing by conversation is a bit finicky. It lacks the precision that I want from a document editor. I ask it to update a list of bullet points, but it changes more than what I had intended, or not quite enough.
But I can imagine that in 2026 it gets better. And again in 2027. And 2028.
Eventually it starts to feel like the AI can read your mind. You can say “This is a great first draft, but I need it to feel more like something my boss would write”… and it works!
Sure Microsoft and Google can implement these features too, but they might be slower than OpenAI. They have huge legacy install bases that they have to support. Large enterprise clients that they can’t risk rushing an imperfect product to. Just look at Google Gemini in Google Sheets now, it can’t even make edits to a cell!
ChatGPT has none of this legacy problem. They can move fast.
And besides, for so many people they’ll already be starting by creating a document on ChatGPT, so why bother copy/paste it to the others?
It’s like the early days of search, when searching on a website was just as good as searching on Google, but Google was the universal starting point, so it became the default.
Step 3: The Workflow Will Be Transformed
All through the 90s and all through the 2000s, Microsoft Office ruled alone. Sure, Apple had iWork and new startups would come and go. Quip. Slideshare. Smartsheets. But even at their peak they barely made a dent in Office’s market share.
Until Google came along.
Google managed to break Microsoft’s dominance by doing something that couldn’t be done before the 2010s. They changed the collaborative workflow that happened around document creation.
The old way involved saving a document (“V1.1 - Peter’s edit - final3.docx”), then emailing it to your colleagues, both of whom would make some edits (and might track changes or might not) and send two separate files back to you. It was a nightmare.
By the time Google docs launched, browser technology had advanced to the point that you could all be working on the same version of the same document at the same time. Google changed the paradigm. The document wasn’t even a file any more, it was url!
Real time edits. Version history. Permissions management. Comment threads. It all worked so much better.
Google docs now has 3 billion users. (Zoho took the same approach at lower cost and has 130 million users)
So even though the other parts of the software weren’t as powerful (slides can’t do transitions like powerpoint and no-one in finance would be caught dead using something other excel), it didn’t matter, because the collaboration was so much better for so many people.
What might be the equivalent for documents in the AI era?
Maybe the new workflow looks like this: Someone one-shots a first draft of a presentation. ChatGPT creates 5 different potential versions, each with a different visual style. Their colleagues vote for the best overall approach. Another colleague wants to change slide 2, so they get ChatGPT to make 10 different variations. They trim this down to 3 and ask others for feedback.
When there is no cost to making options - no mental exertion or creative deliberation - the new workflow might be a constant expanding and trimming of the options. Document creation could take a double-diamond approach, rather than sequential line-by-line editing.
Another possibility could be that “shared context” becomes something like a mix of brand guidelines, standard operating procedures or employee onboarding manuals. Companies will want to set system wide prompts to describe the company’s style, or important rules and parameters that LLMs need to know to better make and edit documents. They will want the AI to have structured access to all of the company’s internal context.
I am not a product manager, so I don’t know what exactly it will look like, but if it happens it will likely involve using GenAI to unlock a workflow around document creation that is both incredibly powerful for a certain set of users AND very difficult for incumbents to quickly move to.
This would be enough to compensate for the fact that their software would likely have only the basic editing features. It would buy them time build the old stuff out.
Bonus Step: The Core Document Types Will be Rumbled
I think the three steps above would be enough to make OpenAI’s offering competitive with the free Google Docs. If they really nailed step 3 (in ways I find hard to envision, to be honest), then they could even put a significant dent in the paid Office Market.
The biggest upset would come from the most difficult challenge - could they use GenAI to not only re-structure workflows around the 3 core document types, but actually change them?
The word document, the presentation and the spreadsheet are the perfect blend of powerful and flexible. They are so foundational that they predate the computer.
Ebeneezer Scrooge had a spreadsheet-like table of numbers in his ledger book. Don Draper presented with a deck of slide-like cardboard cards. Every government department, large corporation and national army ran on paper memos.
The table of numbers, the page of words and the slide of images. These are foundational tools of organisation and communication.
Can GenAI rumble them?
I think Slack was a good example of a technology that came close. For a while there we all got a bit carried away and thought that messaging could be the new core format around which we organise a company’s communication.
But it turned out to be a bit lacking. It’s ephemeral and instant, where most organisations need canonical documentation, built with a bit of thought and effort, to communicate effectively and act as a story of the company’s knowledge.
I’m excited to see the GenAI-enabled attempts to dethrone the Kings.
As a thought experiment for what this could look like, we could think about the elements that often accompany these documents, but aren’t currently embedded within them. The documents often presented to teams, or used as a tool to talk teammates through a project, a budget or a proposal.
So what about a new “FAQ doc” format? An employee creates a project proposal, like they might do in Word or Powerpoint today. But instead of presenting it through a screen share on Zoom, they just share the link asynchronously, and colleagues can have a conversation with the document directly, asking for more information about the project background, the projected outcomes, asking to add other employees to the project team or subscribing to project updates.
What happens when the document shifts from being a tool to support conversation, to become the container for conversation? That could be the disruption (and the email killer) we’ve been hoping for since the dawn of the internet age.