One workflow, described in writing. What arrives, who touches it, where it waits, and which parts a machine should touch.
You describe one workflow in writing — what arrives, who touches it, where it waits. We read it closely and tell you where the work is waiting, what that is costing you, and what to do about it.
We don't sell software and we don't hand you a dashboard. Everything we find and everything we recommend arrives as a written document your team can act on without us in the room.
We work on one thing at a time, and we finish it before we start the next.
You describe one workflow in writing: how it enters the business, who handles it, where it stalls. Twenty or thirty unhurried minutes, and it does not need to be polished. No meeting required.
Jay reads for where the work is losing time. Andrew reads for where AI can carry part of it safely. You get one short written assessment, three to five pages, covering how the work moves today, where the time goes, what we would change, and what we would leave alone.
Specific, practical, in order. What to do first, then next, then later, and why. Written so your team can begin on Monday without calling us.
Optional, and genuinely optional. Owners who want a founder-level second opinion keep us on a monthly retainer. Owners who don't can walk away with something that works without us. That is the whole point of writing it down.
Almost every workflow we look at spends far more time waiting than working. Not because anyone is slow — because nobody owns the gap.
An eighteen-person HVAC company we wrote up takes about ten minutes of actual work to turn a finished service call into an invoice. The average gap between the two is eleven days. At their volume that leaves roughly $63,500 of finished work sitting unbilled at any moment. The owner thought it was two or three days until he pulled twenty tickets and counted.
The common mistake
Most owners assume the way to go faster is to do the work faster. It almost never is.
Waiting is the part a business cannot see from the inside, because it never lands on anyone's desk as their problem. A job that sat eleven days waiting on one piece of paper cost you eleven days, and nobody in the building experienced those days as a delay. There was no task for it. No one was late.
That is what we look for. It takes no software and no stopwatch. It takes attention, and someone who has spent a career noticing the difference between how work is supposed to happen and how it happens.
We use this every day. Not as a product we are selling you — as the way we work. It is in our drafting, our research, and how we run our own business. That is not a pitch. It is just where the tools are now, and it is why we can tell you what they actually do instead of what a vendor says they do.
Which means we will also tell you where they break. These systems are confidently wrong on a regular basis. They are good at the fourth draft of something repetitive and unreliable at anything requiring judgment. And they fail quietly — an automated step stops running and nobody notices until a deadline is missed. We have watched all three happen. The rules below are not hedging. They are what we learned.
So no, we do not think AI is going to run your business. We also do not think you can sit this one out. The useful position is the unglamorous one in the middle: find the work that genuinely repeats, hand that part over, keep a person on the end of it, and write down why.
Four rules we hold ourselves to
Where an AI tool drafts something that reaches a client, a customer, or a counterparty, someone on your team reads it first. We will not design a workflow where that step is optional, or where the reviewer is unnamed.
A reminder that quietly stops firing is invisible until it costs you something. So for every step that runs on its own, we write down who checks it and how often.
We tell you what to ask a vendor — what they retain, what they share, whether the system learns from what you put into it. Which product you buy is your call, and the right answer changes faster than any document can keep up with.
Every recommendation carries the argument behind it. If a regulator, an insurer, or a client ever asks why your business does something a particular way, the answer should already exist on paper rather than in someone's memory.
We never touch your client data, and we never log in to your systems.
Operations Architect
Jay has spent five decades inside operating businesses, most of them accountable for the numbers. He was Chief Financial Officer of a steel corporation, then its President through a turnaround, then an owner who bought a company and grew it from $10 million to $45 million. He has spent that career finding the gap between how work is supposed to happen and how it happens.
LinkedInAI Systems Partner
Andrew was selected into airborne cryptologic linguistics, a career field most candidates don't finish, and handed a language he didn't speak with a deadline to become fluent in it. He builds the systems a review calls for, and makes certain a person is standing in the right place when one of them runs on its own. MIT Professional Education, applied AI and machine learning.
LinkedInStart
Twenty or thirty minutes, in your own words, about one workflow. We read it and send back a short written assessment of where the time goes and what we would change. No meeting required.
PGP-AI, LLC · pgp-ai.com · © 2026