Wedge or rebuild: the two ways to use AI

Two companies, same category, same model vendor underneath. Six months later one is cited as the category leader and the other is a footnote.

The difference isn't budget or team size. It's whether they wedged AI into the product they already had, or rebuilt the product from a blank page knowing AI exists.

That's the Red Ocean / Blue Ocean fork for AI. Almost everyone I talk to is on the wrong side of it.

Wedge plays commoditize on a calendar

Open the product pages of any 20 SaaS tools you used in 2023. Most of them now have a chat sidebar in the corner, a "summarize" button in the toolbar, a "generate" option in the editor. Same three features. Same vendors underneath.

This is the wedge play. Add an AI feature to the existing surface. Tick the box on the procurement RFP. It's rational and survival-grade. It's also, by definition, what every other vendor in your category ships this quarter.

The wedge has a ceiling because the underlying workflow stays the same. The user still opens 14 tabs and stitches outputs by hand. You can't out-feature your way to a moat when every competitor has the same shelf within 90 days.

Most organizations are stuck there

This isn't a niche failure mode. It's the modal state.

88% of organizations report using AI in some function. 6% say it has meaningful business impact. The 82-point gap is what wedge plays look like at scale: the chat sidebar shipped, the summarize button got clicks for two weeks, the workflow didn't change.

Rebuild plays look small until they don't

The Blue Ocean move is to start from a blank page with AI assumed.

Cursor didn't add an AI sidebar to VS Code. They rebuilt the editor with the AI as the primary interaction surface and the keyboard as the assist. $1B ARR faster than any SaaS company in history. Harvey didn't bolt a "summarize this clause" button onto existing legal software. They rebuilt the lawyer's research and drafting workflow with AI as the first reader. Top firms restructured their associate workflow around it inside 18 months.

The pattern: stop asking "how do we add AI to our product?" Start asking "what becomes possible that wasn't before, and is anyone serving that?"

The answer is rarely "the same thing, but with a chat sidebar." It's a different shape of product. Fewer screens. More agents. Less click-through, more direct path from intent to outcome.

The honest framing

Design at its core is reducing friction between intent and outcome. AI is the first tool in 30 years that can collapse the friction software itself added - the 12 screens you click through to do a thing the user described in one sentence.

The teams who notice this will pull away from the ones who don't. Not because of model access - everyone has the same access. Because of the question they decided to ask.

What I do differently in a sprint now

When I run a design sprint these days, the first question I ask is not "what should this product do."

It's "what would this product look like if it was being started this year, with AI assumed in every workflow." Drop everything else for an hour. Sketch the blank-page version. Compare it to the current roadmap.

The two diagrams are usually different enough that the existing roadmap stops making sense. Most of the value in a wedge play comes from the part of the workflow you keep. Most of the value in a rebuild comes from the part you're brave enough to drop.

The smaller test

I shipped a 90-second video on this exact thesis a few hours before writing this post. Most filmmakers I respect would have edited it the way they always edit: open the NLE, scrub takes, cut, grade, master, caption, export. Wedge play.

I built the editing pipeline from a blank page instead. It picks the takes from the raw footage, plans the motion-graphics beats against the spoken transcript, masters the audio, burns word-accurate captions, tracks my face through the 9:16 crop. The video at the top of this post was edited end-to-end by it. I didn't open Premiere.

Not a brag - the pipeline is rough in places. It's a small test of the thesis. If I'm telling clients to rebuild their workflows around AI, I have to be willing to rebuild my own first.

The actual question

Most of the AI strategy conversation right now is technical. Which model. Which agent framework. Which RAG setup. Real questions, with crisp answers if you're paying attention.

The harder one - the one that decides whether you wedge or rebuild - is older: what is your product actually for. If you can answer without referencing the screens it currently has, you can rebuild. If you can't, you're going to wedge.

The teams who win this cycle will be the ones who dared to dream and try different things. Not because dreaming is romantic. Because the wedge play is, by next quarter, table stakes.