Disclaimer: we definitely think your therapist has a real job.

The most personal question on earth these days: How do you use AI? (more personal, perhaps, than what’s on your instagram “For You” page). 

I'll tell you how I use mine.

ChatGPT (better writing skills, less humorous) sometimes edits these newsletters and processes call notes (As a longtime proponent of the em dash, I will vouch for the fact that it's mostly me making these long-ass sentences). It gives me advice on how to sleep train my 3.5-month-old and enables my Chairish addiction / hypes me up on my ongoing design changes in my apartment. In a recent exchange, it called my taste “Cabana-esque,” which is how I know Sam Altman is dying for revenue and thinks incredibly well-aimed flattery will get him there.

Claude (more dry, tech-bro vibes, appreciates my shitposts) gives me feedback on product ideas, helps me through my existential spirals, writes me bad deck outlines (that I sometimes use, even though I know they're bad), and consistently fails to come up with a good way to read my Gmail and figure out who I need to follow up with.

Neither has really *revolutionized* the way I operate my business.

Because I'm mostly treating them like therapists, or friends I can text when my actual friends are busy and don't want to talk about decorative pillows or my company.

(Both Claude and ChatGPT thought this intro was too long.)

So: how are you using AI to revolutionize your business?

That was the promise, right?

People on Linkedin and the silicon valley set will have you thinking that everyone (but you) has already completely transformed their business using AI. In truth: most companies got an enterprise AI account, had everyone log in, and now employees use it to generate emails they’ve forgotten how to write themselves and talk to it like a work wife. Useful? Definitely. Revolutionary? I'm less convinced.

Because the really interesting version isn't just asking AI to write something for you.

Imagine asking it how your projects are progressing. What's running over? What's running under? Why?

Imagine updating your timecard by telling Claude what you did today instead of opening another piece of software.

Imagine uploading the messy scope your producer built in Excel and having it automatically turned into structured project data.

Imagine pricing a new scope based on similar projects you've done before — including the knowledge of where you fucked up the last three times.

That's what we think an actually useful version of AI looks like. And that's what we've built — and are building — with FOL-IO 2.0.

One reason AI isn't revolutionizing your business operations is that, in most companies, it doesn't actually understand how your business operates.

Your project data is in one place. Your budgets are somewhere else. Your timecards are somewhere else. Your scopes are PDFs and spreadsheets. Half the important context is sitting in Slack. And everyone has a slightly different name for everything.

AI is very good at working with information when it has clear context: what the data represents, how the pieces relate to each other, and where it should look for an answer. When asked to iterate repeatedly on messy data sets, AI has a high chance of hallucination. 

A great way to visualize this: when I feed AI a photo of my living room and ask it to render a decorative pillow, the first pass looks quite a bit like my living room. The more pillows I ask it to render (it’s a lot of pillows), and the more passes it takes over that same photo, the more the paintings on the walls warp and the more likely the couch is to lose a leg.

Your operating data has the same problem, except unfortunately the disappearing couch leg is the accurate read of your gross margin.

This is why we've spent so much time building the underlying data structure of FOL-IO. It acts like a rigid set of rails for the AI to run on top of, with clearly elucidated relationships between data points and and an understanding of which data point means what. (Increasingly, the infrastructure surrounding the model is now called a “harness”, which generates an amusing mental image).

After testing in beta, we've recently launched the FOL-IO MCP, which makes the data we capture in FOL-IO — hours, rates, projects, people, workstreams, scopes, budgets — legible and accessible to AI.

Give AI that structure and tools and suddenly you can ask:

What's running over budget, and why? Where do we have unsold capacity next month? Which types of projects are consistently under-scoped? What did we actually spend the last three times we made something like this?

And, importantly for those among us who have not made spreadsheets their life's calling, it can answer you in actual human (robot) words.

There is, of course, a catch.

Structured data is extremely useful.

BUT, after hundreds of phone calls, I can say with certainty that creative people generally hate entering structured data.

We've built a really robust system for creative operations and finance, but none of it matters if getting the data into the system creates another administrative job for already-busy teams.

So one of the biggest changes in FOL-IO 2.0 is using AI not just to analyze structured data, but to make creating it dramatically easier.

Scopes: We built our original scoping tool to create clean scopes with margin analysis built into the process. Revenue and margin were automatically calculated, the underlying data was captured, and versions were stored so you never had to figure out whether v.F or v.FF_FINAL was the one the client actually approved.

The problem: you had to scope our way.

Now, start your scope in whatever format your team already uses. Upload it into FOL-IO. We'll turn the mess into structured data and do the analysis from there.

Timecards: If you use the FOL-IO MCP, entering time can be as simple as telling the system what you worked on and roughly how you divided your time. It asks a few clarifying questions, translates that into structured entries, and you're done. No more “timecard entries for making timecard entries.”

And we're applying the same principle across the product:

Let humans communicate naturally. Let the software worry about structure.

Much has been made of the future in which “agents” take on work themselves.

While people in the media do a justifiable amount of hang-wringing about what this means for the future of human labor, at FOL-IO, we think about agents in much more practical terms: doing the work your teams are doing late, under extreme duress, or not at all.

And this is where the structured data really starts to matter. Just like a real employee, an agent can't make a useful decisions if it doesn't understand your business. Once it does, the possibilities get much more interesting.

A freelance assistant that checks capacity, pulls budgets, compares margins and tells you whether you can actually afford that freelancer before you book them.

A production assistant that takes information from Slack and updates the latest production budget so Operations isn't working from information that's three days old.

A bidding producer that takes a brief, looks at comparable historical scopes and actuals, and builds a starting budget based on what the work has actually cost you before.

None of these are particularly sexy robots, and none have a 10% chance of killing you (we think). They're just annoying jobs that currently require a smart person to go find information in six different places, reconcile it, and make a judgment - and that's exactly the kind of work I want AI doing.

As much as I love draining the world's water supplies to figure out the difference between Nobilis and Scalamandré tiger-print velvets (IYKYK), I do think AI has a higher use case. I believe we can use AI to learn from our past mistakes so we can plan better in the future. I believe it can take the enormous amount of operational knowledge already sitting inside creative companies and make it actually usable — massively expanding the capabilities of the teams people already have by freeing them from data digging and letting them spend their time where it actually matters.

And maybe most importantly, I believe AI can serve as a translation layer: helping right-brain people do left-brain shit.

That's the new FOL-IO.

Messy in. Structured out. Intelligence on top.

Reach out if you want your AI to do something a little more useful for your business.