How I Used AI to Pivot a Company in a Week

Five months of building the wrong product. Two weeks of revealing conversations. Six days to pivot, rebrand, rebuild, and launch.

What’s worth examining is which parts AI made faster and which parts it couldn’t touch at all.

The Wrong Wall

From October 2025 to January 2026 I worked on TeamScore, a performance intelligence tool for remote teams. The thesis was straightforward: Six years after the pandemic forced everyone remote, most managers still couldn’t really tell who was actually working and who was faking it. Surely they’d love to have that visibility.

They wouldn’t. Or more precisely, the ones who cared had already solved it, and the ones who hadn’t didn’t care enough to start now.

The signals were hard to ignore. Leaders kept asking the same question in demos: “Are you sure my team won’t know?” That question, repeated across enough calls, tells you something uncomfortable about the product you’ve built: if people are fearful to use it, you've got a real problem.

A friend was more direct. He took the product to his leadership team and they shot it down. When I dug in, he admitted that some of his people liked being able to hide and fake-work-from-home sometimes. The cost of having that fight wasn’t worth it to him; he’d rather accept slacking off as normal than have team conflict. Fair enough. If your champion doesn’t care enough to do what is obviously the smart thing, you don’t have product-market fit. You have a polite rejection.

So, I needed to pivot. But how could I do it without wasting even more time?

The Thinking AI Can’t Do

This is where founders get tempted to fire up ChatGPT and ask it to solve their existential crisis. Don’t.

I stopped listening to audiobooks for a month because I needed my subconscious to have empty cycles. The clarity about what to keep and what to jettison came during laundry, during sets at the gym, during the kind of unfocused thinking that AI will never replicate because it requires living with a problem long enough to feel the answer before you can articulate it.

I got on video calls with early users. While you’d think surveys, with their dozens of data points would be better, I’ve found surveys tell you what people think they should say. Video calls, however, let you watch someone’s face when you describe the pivot and know in three seconds whether they’re excited or polite. I outlined a mea culpa: here’s what failed, here’s what I think comes next. The body language told me more than the words.

As South Park brilliantly showed, AI doesn’t have judgment. Using it to get conviction on something you’re betting your future on is a recipe for disaster.

Where AI Earned Its Keep

Once the direction was clear, the acceleration from AI was real.

I downloaded my strategy notes from Google Docs as Markdown (the best format for giving an LLM structured context), told ChatGPT to interview me as a co-founder, and used that pressure-testing conversation to find gaps in my logic. Then I exported that chat, fed it into Gemini’s Deep Research (Google’s search index makes it stronger than ChatGPT for this), and brought the research back into ChatGPT for positioning work. Six hours over two days. The strategic foundation was pressure-tested and documented.

While I started using AI to come up with a name one Tuesday, the best approach came the old-fashioned way. I’d been banging my head against it all afternoon and AI was useless here, making all sorts of terrible suggestions. ChatGPT 5.2 Pro’s suggestions were vanilla. When I pushed harder, they got longer but not better. It was another example of where creativity remains a human skill. I gave up that Tuesday night and went to bed.

But again benefiting from the subconscious, on Wednesday morning, standing in the kitchen waiting for the kettle to boil, the new name hit me: WorkSights.ai.

Shockingly, it wasn’t already claimed. I added it to my GoDaddy cart by 7am and confirmed the trademark was clear on USPTO before taking the kids to school. Domain registration done at 8:59 AM and trademark filed for $350 by 10:10 AM. (ChatGPT was actually useful for clarifying a couple of filing questions. Give it a bounded, factual task and it performs.)

Then it was time to bring the product to life with the new brand. Cursor plus Claude to update the code and bring the entire platform to life with the new brand. Code live by 1:30pm. Loose ends tied up by 3pm. Under 6 hours on a Wednesday.

The last step mattered more than it sounds. I spent about 10 hours over the next two days turning all of that thinking into a detailed go-to-market plan: positioning, ICP, competitive landscape, messaging, brand voice, the full foundation. It felt like overkill before the website was even updated. But it wasn’t. 

That 30 page document - finished on Friday night - became the grounding context for every AI session that followed. Every blog post, every email, every script started from the same, well thought through strategic base. Without it, I’d have been re-explaining context from scratch in every chat, and the output would have been inconsistent. The investment in the plan is what makes the speed of everything that follows possible.

The Creativity Gap

Here’s what surprised me. I expected AI to struggle with judgment. It did. I expected it to be fast at execution. It was. But I didn’t expect it to be this bad at anything requiring creative taste.

The branding ideas were generic. The name suggestions were forgettable. Every attempt to push ChatGPT away from the grounding context of the old brand pulled it back toward safe, predictable territory. It couldn’t untether from what already existed to imagine something new.

This shouldn’t surprise anyone. Creativity and intelligence aren’t the same thing in people, either. I know plenty of brilliant people who couldn’t write a decent headline, and plenty of creative people who can’t structure an argument. AI in March 2026 is firmly in the brilliant-but-uncreative camp. I imagine the labs are working on this. For now, don’t expect your LLM to be your creative director.

The Real Timeline

A few weeks of thinking. A week of conversations. One day of naming (mostly staring at walls). Six hours of strategic planning with AI. Six hours of engineering with AI. Ten hours of detailed market planning with AI. Two days to launch.

The thinking took weeks. The execution took days.

AI didn’t make me smarter about what to build next. It made the distance between knowing and doing almost disappear. For founders facing a pivot, that acceleration matters. The longer execution drags, the more conviction erodes. The faster you get from “I know what’s next” to “it’s live,” the more of that hard-won clarity survives contact with reality.