Field Notes: Three Workflow Tools I Shipped Last Week

Last week I built for three very different jobs: an Alaska Airlines flight attendant trying to figure out her paycheck, a real estate agent drowning in follow-ups, and a flight attendant again, this time taking drink and meal orders in the aisle.

Different people, different work. But the same question sits behind all of it, and it is the only filter I run a build through before I write a line of code: does this remove a real step for a real person doing real work? If the answer is no, it does not ship. Here is what made it through last week.

A Pay Decoder: making airline pay legible

If you know anyone who works as a flight attendant, you have probably heard how hard their pay is to understand. The rules come from the contract between the union and the airline, so there are a lot of special cases that can move the number: premium time, holidays, delay pay when a trip runs long, and extra credit for working the lead position up front.

Everything is tracked in TFP, a value assigned to each trip. You can roughly estimate a trip by multiplying your TFP by your pay rate, which itself depends on how many years you have been flying. Then the variables kick in and the estimate stops being simple.

A Pay Decoder does that math for you. A flight attendant uploads their trip details, which are easy to pull from their work app, and the tool shows exactly what to expect for that trip. Upload the whole month and it totals the month, with a running tally as they work through it, so they can see where they stand instead of guessing.

Then it closes the loop. When the airline releases the final worksheet, which is reported in TFP rather than dollars, they can upload that too, and the app reconciles what they expected against what they actually earned. Upload the pay stub and it confirms everything matches, or it flags the discrepancies so they can contest or ask for clarification. Calling out the mismatches is the real value.

One design decision matters here more than any feature: none of that data leaves the device. The pay stub, the trip details, all of it is read and parsed locally on the phone or laptop. Nothing gets uploaded to a server, and the flight attendant can delete it at any time. Sensitive information stays where it belongs.

You can try it at apaydecode.com.

ListingStack: a week of CRM upgrades for real estate agents

ListingStack is the CRM I have been building into the most helpful tool it can be for real estate agents. Last week was a big one for it.

The piece I am most glad to have finished: automatic responses to leads. When someone replies to a campaign or sends an inquiry, the system answers them right away, in whatever format they chose, email or text, and handles the general questions. Anything specific that only the agent would know, like details about a particular property, gets handed back to the agent with a notification to follow up. The lead does not sit in silence, and the agent does not lose the thread.

I also built a buyer survey. One of the hard parts of working with buyers is remembering everything you need to gather. Now the agent clicks a button and sends a quick survey. The buyer opens the link and answers what they are looking for: single family, townhome, condo, or land; beds and bathrooms; the amenities that matter to them; whether they are pre-approved and what their budget is; and whether they are already working with another agent. All of it flows straight into the contact profile, so the agent starts with a head start instead of a blank page.

On the seller side, there is now a dashboard that shows sellers every marketing activity working for them: open houses held, how many people came, and the responses. Those visitor counts are captured electronically. ListingStack generates a one-click flyer with a QR code for the open house, and when visitors scan it to check in, their information lands directly in the CRM. No manual entry, and the agent can show the seller exactly who walked through.

Social posts got the same one-click treatment. The listing image generates as a shareable graphic sized for Instagram or Facebook, with just listed, open house, or just sold versions ready to go. Tap the network you want and it posts.

My personal favorite, though, is the simplest one: take a photo of a business card and it imports the contact for you. You glance at it, confirm it looks right, and you are done. Those stacks of cards in your drawer or your car become CRM contacts in seconds.

The Flight Steward: taking cabin orders without the paper

The Flight Steward is a small app I built for flight attendants to take drink and meal orders on their phones. Today the job runs on paper. You write down seat numbers, names, tier status, meals ordered, and drink orders, then carry that paper back to the galley to prepare everything. When the pre-printed template is not around, people end up scribbling on a blank sheet, and details like a passenger's name or frequent flyer status get lost.

So I built those pieces in. The flight attendant takes a photo of the passenger list, just first names, and the app drops each name into a seating chart along with the passenger's mileage or tier status. When it is time to take orders, they tap the seat and, with a few taps, capture the drink and meal. They can see their meal inventory at a glance: what is available, what has been ordered, and what is out. So they never have to tell someone a meal is gone that they thought they still had. Back in the galley, a prep list shows exactly what to make, and they check items off as they prepare and deliver them.

The whole thing runs locally on the phone, same principle as the Pay Decoder.

Something honest happened with this one last week. My partner showed me a social media post where someone had built essentially the same thing, and it got a warm response from a lot of people. My first reaction was to feel like an also-ran, since we are still making sure ours works exactly the way people would want before we go fully public. But I came around quickly. My mission is making sure technology makes people's lives easier, not harder, and if someone else is doing that too, good. That is the point.

The throughline

Three tools, three narrow jobs, and none of them is trying to be everything. A generic assistant cannot decode an airline pay contract, remember which amenities a specific buyer wants, or track a galley's meal inventory. A tool built for exactly that one workflow can.

Two things show up in all three. The first is that the work is specific. I would rather solve one real problem completely than solve ten problems halfway. The second is that when the data is sensitive, it stays on the person's own device by default. Where the data goes should be the first question you ask about a tool, not the last.

That is the week. If you run a business and there is a step your team repeats that a tool like this could remove, that is exactly the kind of thing I look for. The AI Audit is a good, low-risk place to start, or you can book a call and we will figure out whether it is worth solving.