A doctor who sits on an AI board told me he can't keep up with the number of complex AI products he has to evaluate. What he really wants for himself is simpler. He wants someone to triage his inbox.

A patient who's dying and a patient who just needs a refill land in the same inbox. He reads every message himself to decide which one needs an answer in 30 minutes and which can wait seven days.

The vendors pitching him are selling something sophisticated when all he needs is better inbox management. I used to make the same mistake.

Before I knew where to start with a new client, I pitched big projects. I wanted to modernize how their whole business ran, and I made it so complex that it was hard to buy.

A buyer's first purchase has to be simple. It should fix one inefficiency they deal with every day. When I started leading with the simple fix, I landed more clients and showed value faster.

The operators I talk to are under pressure from two sides. They're told to do AI, then told to show results, and many of them take too big a swing, spend a lot, and end up with little to show for it. So the question I get all the time is where to start.

Below is what I tell them, why I recommend it, and what it looks like in five industries.

Every engagement we've run starts with one of eight AI workflows. The one I see implemented most often is the least exciting but often the most valuable. It's getting information out of emails, texts, calls, and documents and into the system the business already runs on, like a CRM, an ERP, or an underwriting model.

Across my clients, the biggest inefficiency I see is moving data from one place to another. Operators assume AI has to do something sophisticated to be worth paying for, while their teams spend hours a week on a manual process rekeying what someone already said or sent.

I recommend it first for three reasons. The return is easy to measure, because you can count the hours and labor a task takes before and after. There's no adoption needed, because it automates a step in a process your team already runs and gets them to the answer faster. And if your data doesn't exist in your systems of record in the right structure, nothing else you build with AI will work anyway.

Regardless of industry, the workflow has three basic parts:

Done this way, nobody on the team has to open the system to keep it current. That one simple workflow completely changed CRM adoption for one of our clients.

The client is a real estate brokerage. They had HubSpot for eight months, and the brokers never even logged in, which is a complaint I hear from COOs all the time about tools they want their employees to adopt.

The head of the brokerage wanted everything the brokers knew about their buyers in one place. When a new listing came in, the goal was to ask the system which buyers might want it instead of texting six brokers.

We set it up so the brokers never had to open HubSpot. They told Claude what happened with a client, and Claude logged it in HubSpot with the right structure because we trained it that way. When they needed to retrieve information, they asked Claude.

The brokers started doing what leadership had been asking of them for most of a year, because it was finally easy.

The same workflow applies across any industry. For each one below, you'll see what comes in, where it has to go, and what the time back is worth.

The inbox problem the doctor described to me shows up across primary care. A study of 1,275 primary care physicians in the Journal of the American Medical Informatics Association found they spend an average of 52 minutes on inbox management each workday, 19 of them outside work hours. That's about 4.3 hours a week per doctor, sorting urgent messages from routine ones by hand.

Triage reads every message as it comes in and sorts it by urgency. The doctor starts the day knowing which messages need an answer now, and the rest are grouped as ones that can wait.

Picture triage taking those 4.3 hours down to one hour a week. That's 3.3 hours back per doctor every week. In a practice with 50 primary care doctors, it's 165 hours a week the doctors can spend seeing more patients.

A commercial real estate company asked me if we could pull the numbers out of an offering memorandum and put them straight into their Excel underwriting model. Their analysts were doing it by hand.

Every OM is formatted differently, and some are just a list of stats in the body of a broker's email. An analyst reads it, keys each number into the spreadsheet, and hits run. The typing is the slow part, and it's where the mistakes show up.

Once the extraction is automated, the numbers land in the model as soon as the OM arrives, and the analyst's time goes to judging deals instead of retyping them. The team underwrites more deals each week with the same people. On a $20 million building, a better entry price of even half a percent is $100,000. If underwriting more OMs finds one better entry a year, that one deal covers the cost of the workflow many times over.

A loan officer at a large real estate lender walked me through what happens when a new client calls. Say someone needs a $200,000 loan for a fix-and-flip. He spends 20 to 30 minutes keying in everything they told him so his loan sizer can give them a rate.

While he was showing me, he stopped partway through, realized he had entered something wrong, and went back to fix it.

He told me he wants to grow his book and doesn't want to hire more people. At 20 to 30 minutes of typing per new client, he can't do both: the typing is the ceiling on how many clients he can take. Cut it to a couple of minutes and that time goes straight back into new clients. If he spends 10 hours a week on client intake today, clearing the typing is room for dozens more a month on the same schedule, which is the book growth he wanted without a single new hire.

The chief claims officer at a mid-market insurer told me he is not adding headcount. His claims keep increasing, and he sees AI as the only way his team keeps up.

Before anyone on his team can make a decision on a claim, they put together a claims package from the police report, medical records, and every other document tied to it.

Today an adjuster opens each document, pulls out the facts that matter, and assembles the package by hand before the real work starts. With the workflow in place, the package is built when the claim comes in, with the key facts from every document in one summary inside the claims system. The adjuster starts at the decision.

As an example, say assembling a package takes an adjuster 30 minutes and the team handles 200 claims a week. That's 100 hours a week spent before anyone decides anything. Cut it to 10 minutes per claim and the team gets about 67 hours back each week, more than one and a half adjusters' worth of time, without adding headcount.

One of our clients is a manufacturer. Quote requests come in by email or a web form, and an engineer keys the specs into the ERP by hand to build the fabrication package. It takes about 20 minutes each time, across 10 engineers, and they had done it that way for 35 years.

Across a full year, that came to about 10,000 hours. With the workflow in place, the specs are extracted and routed into the ERP as soon as a request comes in, and nobody on the floor has to retype them. They get more requests than they can handle today, so 10,000 hours back is 10,000 more hours to produce against that demand.

If your system of record is a mess, putting more data into it won't fix that. Clean up the fields that matter first, or you'll end up with more data in fields nobody trusts.

A person still approves what gets written. Check the first 20 records it creates against the original emails and documents before you let it run on its own.

Patient and client data needs a business agreement with your AI vendor and training on your data turned off. If you can't get that, de-identify the data first or stop here.

Time back only turns into revenue if there's more work waiting. The loan officer wants more clients and the claims team has more claims coming in, so their hours turn into volume. Without that demand, you're only saving cost.

This is usually the first workflow we build with a new client. It shows value fast, and everything we build after it runs on the data it puts in the system.

Zach

Founder & CEO, OutcomeCatalyst
Book 30 minutes with me →