The front desk used to dread the busy season. Now their software wish list gets built in hours, not weeks.

For businesses whose team works around its software, especially ones that take orders, bookings or appointments: heating oil, propane, dumpsters, HVAC, salons, clinics, contractors. One heating oil company had its software rebuilt around how it really works, and new ideas get built in hours.

Based on a real system. Names and client details changed; screens show sample data. The click count and the build times are my own rough figures, not measured. Results vary.

The business

It is a Monday in the busy season at a heating oil company. The phone has been ringing since 7:30.

At the front desk, someone logs in to another computer over remote desktop, just to open the office program. It looks like it comes from another decade.

An order came in online overnight, in a separate system, so it gets typed into this one. By hand.

The customer filled in the whole order, but the page could not show them a delivery date. The schedule is too customized for it to handle. So the office works out the date and calls to tell them.

The front desk is not looking forward to this season.

Same Monday, new system.

The customer ordered online and picked from the delivery dates that are actually available for their address. The delivery ticket prints on the printer the office already owns.

The office did not retype the order, and it did not have to call to explain the delivery date. The front desk is looking forward to this season.

Later that week, someone at the front desk says, “Wouldn’t it be great if...” The idea is built the same day, often within hours.

An illustrative scene, not a log of one day.

The short version. A heating oil company ran its office on an old program that only worked on one computer, and its online orders came in through a separate system, so the office retyped them by hand. The delivery schedule follows the place and the time of year, so customers could place an order but could not see their delivery date, and the office phoned them. We built one system, with AI, around how the company really works. Customers pick from the delivery dates available where they live, online and phone orders sit in one history, a typical office task went from about 15 clicks to a few, and the office’s new ideas get built in hours.

Sample data. A screenshot of the dashboard: revenue, gallons delivered, orders, payments collected, average price, new customers and outstanding balances, plus orders by ZIP code and by city for Boston, Worcester, Providence, Hartford, Quincy and Cambridge.
Sample data. A screenshot of the real system. The dashboard: revenue, orders, payments collected and where the orders come from.

This is for you if

  • Customers place the whole order, then someone has to call with the delivery date.
  • The busy season is something your team survives, not something they enjoy.
  • Your business has its own odd rules, and your software has no place for them.
  • You have a “wouldn’t it be great if...” list you stopped writing down.

The business

A heating oil delivery company in its busy season: orders arrive by phone and online, drivers are on the road, and the delivery dates available depend on where the customer is and the time of year.

The problem

What the office worked around

The office program was old and ran on one computer. The company’s version of “the cloud” was remote desktop, and it was hard to look after from anywhere else. It had many issues the office worked around, because it did not match how the business operates.

  • Online orders lived in a separate system. The program had no online ordering, so another system took web orders. The systems did not talk to each other, so the office typed those orders in again.
  • The delivery date was unclear. This company runs a custom schedule for spring, summer and winter, and which dates the trucks can deliver depends on where the customer lives. The online system had no way to hold rules like that, so a customer could fill in the whole order and still not know the delivery date.
  • The office did the phoning. The office worked out the delivery date and phoned the customer to say it. That did not work for the company.
  • Retyping meant matching. Entering the same order again and checking the copies against each other is where mistakes come from.

What we built

What was built: one system around the way they work

Some of the parts that matter most.

A calendar that already knows the schedule

Think of a calendar that changes with the time of year. The company has a custom schedule for spring, summer and winter. In each one, the trucks reach an area on certain dates, and each area follows its own pattern. The schedule is built into the system, so the office does not have to work it out by hand. The customer enters where they live and sees only the delivery dates that apply to them right now. Fewer phone calls and fewer upset customers. The order also offers the extras the company sells, like priming, which gets fuel flowing again after a tank runs empty, and emergency delivery, so adding one is easy.

Sample data. A screenshot of the online ordering page on two screens: choosing a delivery date from the dates available for the customer’s area, and choosing gallons, with the price, an online discount and an optional priming add-on.
Sample data. A screenshot of the real ordering page. The customer picks a delivery date from the dates available for their area, chooses gallons and adds extras.
Sample data. A screenshot of the online ordering page on two screens: the ZIP code check with the fuel and its price, and the order summary with the card payment form.
Sample data. A screenshot of the real ordering page. The customer enters a ZIP code, sees the fuel and the price, reviews the order and pays by card.

The wish list

The office keeps having ideas to save time, speed things up or improve things. With old-school software, an idea like that meant days or weeks of custom development, or it did not happen at all. Here, by my rough count, small requests are built the same day, often within hours. That gap is the difference between traditional development and building with AI.

Set up once, not typed again

The system already knows the company’s products, delivery areas and schedules, so staff do not enter them again. By my rough count, a typical office task that took about 15 clicks in the old program now takes a few.

One list, one history

Customers order online and can see their online order history and their full history, including orders the office placed over the phone. The office works from the same list: place a new order, look back through the whole history. Orders are entered once, so the checking and matching between separate systems is gone.

Sample data. A screenshot of the dispatch screen for one delivery date, with drivers and routes across the top. Each row is an order with its route, product, customer, address, city, ZIP code, gallons, total, extras, status and driver. A globe icon next to the order number marks an online order and a phone icon marks a phone order.
Sample data. A screenshot of the real system. One list for the deliveries on one date, split by driver and route: online orders (globe icon) and phone orders (phone icon) together.
Sample data. A screenshot of the office orders list with a bar of delivery dates across the top, and columns for customer, city, gallons, price, status, type, payment, driver and scheduled date.
Sample data. A screenshot of the real system. The office works from one orders list, one delivery date at a time.

Tickets and reports without the retyping

Delivery tickets print on the printer the office already owns, with no manual editing. The office builds whatever reports it needs and can look at them as pictures. Tax season is easier, and business decisions are easier too.

Sample data. A screenshot of the reports page: delivery, fuel purchase, sales and tax, special fuel tax, sales by product and accounts receivable aging reports.
Sample data. A screenshot of the real system. Reports for the office and for tax season.

Routes for the drivers, on a map

Dispatch puts each delivery date in front of the office, split by driver and route. A route shows on a map with the stops in order, the drive time and the distance, and the stops can be dragged into a new order. Driver tracking and delivery verification sit in the same system.

Sample data. A screenshot of the route map for one driver: a map of the Boston area with numbered stops joined by a driving route, the drive time and distance at the top, and a list of the stops that can be dragged into a new order.
Sample data. A screenshot of the real system. A driver’s route on a map, with the stops in order, the drive time and the distance.

The rest of the office sits in the same place

The same system also runs inventory, receivables and an audit log. Inventory keeps track of the fuel bought, the fuel delivered and what is left in the tanks. The competitor price tracker from Day 3 and the fraud check on online card orders from Day 8 are parts of it as well, and the price tracker shows each competitor’s latest price next to the company’s own. The office works in one place, not in a pile of separate tools.

Sample data. A screenshot of the inventory page: for each fuel, the opening balance, purchases, deliveries, net change and ending inventory for the month.
Sample data. A screenshot of the real system. Fuel bought, fuel delivered and what is left in the tanks.
Sample data. A screenshot of the competitor price page: the company’s own price, the lowest and average competitor price, and a table of competitors with their latest price, the change since the last check and the difference from the company’s price.
Sample data. A screenshot of the real system. Competitors’ latest prices next to the company’s own price.

Why not just buy software?

There used to be a choice. You bought cookie-cutter software and worked around it, or you hired a company to build it for you. By my estimate, a custom build took six months to a year and hundreds of thousands of dollars, so most businesses bent to the software. Now, in my experience, a custom system can go from start to launch in weeks, for a fraction of the cost. My opinion: a business that bends to its software loses ground to one whose software bends to it. It is the same argument as Day 12, a towing company, for a different business.

The result

About 15 clicks to a few

for a typical office task, by my rough count, with online and phone orders in one system

The result

BeforeNow
Online ordersTyped in again by handAlready in the system
Delivery dateThe customer ordered, then the office worked out the date and phonedThe customer sees the available delivery dates for their address while ordering
Order historyOnline and phone orders in different placesOne history, for the office and for the customer
A typical office taskAbout 15 clicksA few clicks
Mistakes from retypingOrders retyped and matched by hand, so mistakes could creep inBasically gone
New ideasDays or weeks of custom development, or not done at allBuilt the same day, often within hours
The busy seasonDreaded by the front deskLiked by the front desk

What stays human: the office decides what to ask for next.

What it changes for you

  • Customers see their delivery date while ordering instead of waiting for a call.
  • An order is entered once, and a typical office task takes a few clicks instead of about 15.
  • A wish list that gets built in hours, not weeks.
  • A busy season your team does not dread.

Now picture yours

The software does not have to be heating oil software. Wouldn’t it be great if...

  • the intake form filled in the patient’s chart before they sat down?
  • customers only saw appointment slots your shop can really keep?
  • the estimate, the crew calendar and the invoice were one record?
  • your clients’ documents sorted themselves in tax season?
  • customers saw only the delivery dates your trucks can make?

These are wishes, not systems we built.

Want this for your business? Book a 30-minute consult.

What do you wish your software just did? And what does your team do by hand until it does?

Steal this

Want to build this yourself? Here’s how it was built, in plain words. The idea fits businesses whose team works around its software.

The pieceWhat it does
Claude CodeBuilds the whole system and the rules behind it, and builds each new idea from the wish list
An online ordering siteWhere customers order, pick a delivery date and add extras like priming or emergency delivery
Customer accounts and sign-inLets each customer see their online history and their full history, phone orders included
A booking calendar that follows rulesShows only the delivery dates available for the customer’s location, following the schedule for the current season
Business settingsHold the products, delivery areas and seasonal schedules, so staff do not enter them again
A databaseOne customer list, with the orders placed online and by phone
A card payment serviceTakes card payments for online orders
A fraud checkGives each online card order a second look before a truck is sent (see Day 8)
Printer setupSends tickets straight to the printer the office already has, with no manual editing
A dispatch screenGives each driver their deliveries
Driver trackingShows where each driver is
Delivery verificationLets the office check off each delivery that was made
InventoryKeeps track of the fuel on hand
ReceivablesShows who owes what, and for how long
An AI agent that watches pricesChecks competitors’ prices and texts the owner when they change (see Day 3)
An audit logRecords who changed what, and when
Reports and chartsLets the office build its own, including the tax-season ones
A web serverRuns it all, so the office uses a web browser instead of remote desktop

Prefer not to build it? Have us do it for you.

Is your business in the same spot?

In 30 minutes we'll look at your version of this problem and how an AI system could handle it for you.

Book an AI opportunity audit, $249

Share this case study

A new case study every day. Follow on LinkedIn · X · Facebook · Instagram.