Don’t Be the Accountant Who Chose Not to Adopt Excel.
There’s a lot of noise right now about AI making founders more productive. But what does that actually look like in practice?
We brought together Barry Peterson (CTO at Every, formerly Head of Engineering at Reflective) and Saadiq Rodgers-King (cofounder of Hot Potato, acquired by Facebook, now helping organizations implement AI at scale) to cut through the hype and show what they’re actually doing with AI tools every day.
TL;DR: The efficiency gains are real, but they require daily practice. File format choices matter more than you think. And when it comes to privacy concerns, history suggests convenience wins when the pain is high enough.
Scroll down to watch the full conversation (or keep scrolling for the highlights) 👇
This virtual event was presented by Every
Every is the modern back office for startups. We combine powerful software with expert support across payroll, benefits, HR, accounting, and taxes, all in one platform. Founders use Every to save time, stay compliant, and scale faster without the busywork. Whether you’re hiring your first employee or expanding globally, Every makes it easy to grow your team and run your company.
The 5-Hour Problem That Takes 5 Minutes
Barry shared a moment that captures the transformation happening right now:
“I was solving a problem yesterday, and I had this thought of, like, if I didn’t have this, this would have taken me five and a half hours. Like, no question. And five minutes later, I have all the context and all the, like, you know, awareness for me to be able to make a decision.”
The real leverage here is speed to context. Barry uses Claude Code to navigate Every’s codebase, and problems that would have meant hours of digging now take minutes.
For founders, that means collapsing the time between “I have a question” and “I have enough information to decide.”
Why Your File Formats Actually Matter
Saadiq dropped a surprisingly practical insight that most founders overlook: AI performs dramatically better with established file formats it’s seen before.
“And markdown files are incredibly efficient on a token basis for passing through AI.”
His approach: Keep documents as Markdown files. Export data as CSVs instead of spreadsheets. Stick with JSON for structured data. These formats aren’t just readable—they’re what AI has been trained on extensively.
“One of the benefits of using things that are established and known is that the AI knows how to interact with it. I’m not trying to teach it how to use the latest format.”
The temptation to chase more “efficient” novel formats often backfires when the AI doesn’t understand how to work with them.
“CSVs, JSON, all those things are good enough for me.”
The Plaid Lesson on AI Privacy
The conversation took an interesting turn when discussing how much data to share with AI tools. Barry drew a parallel that founders should think about:
“So Plaid, right, the tool that allows you to connect your bank account to so many various different services. At some point, everybody just became okay with giving somebody your username and password to your bank account and allowing them access to go grab your information. And that’s because the headwind, the frustration of connecting a bank account prior to that was so hard that we just made this concession.”
His point: When the friction is painful enough, people make exceptions they never thought they would.
“If the pain is high enough then people start making exceptions.”
This doesn’t mean blindly connect everything, especially if you’re handling customer PII. But it does suggest that the companies solving genuinely painful problems with AI integration will win users, even with privacy trade-offs that seem uncomfortable today.
Build Something Locally (Even If You’re Non-Technical)
Saadiq made a push he admitted he’s been hesitant to make before:
“But I do believe if you are a founder, technical or non, I do believe you should take a couple days to build an app locally on your machine that does something useful for you.”
Once you’ve built something small, even using Cursor, Claude Code, or Claude Desktop, you start seeing friction differently.
“It unlocks the ability to go, this is frustrating. How do I do this better to remove friction from your life?”
Saadiq shared that early in his consulting work, he wrote a script to hit Stripe’s API and generate draft invoices so he didn’t have to figure out Stripe’s interface. Small automation, real time saved.
The throughline here is clear: the founders getting the most out of AI aren’t waiting for perfect tools or perfect use cases. They’re using what’s available now, daily, and building intuition for where the leverage actually is.
Barry summed it up:
“You don’t wanna be the accountant that chose to not adopt Excel.”
The full conversation covered more ground—including how to set up a “second brain” with Obsidian, when to use Lovable vs. Claude for prototyping, and the privacy trade-offs of connecting all your tools. Watch the full discussion above.
