Find the Choke Point: What to Automate First

Kris Van Meter · July 2026 · Guide

Most owners pick their first automation the way most people pick a first gym exercise: whatever hurts the most. The task they hate. The inbox they dread. It feels productive, and it changes almost nothing, because the pain you feel loudest is rarely the constraint your business is queued behind.

Theory of constraints, in one paragraph. Every operation has a bottleneck: one step, often one person, where work stacks up and waits. The output of the whole system equals the output of that step. Speed up anything upstream of it and you have made the pile in front of the bottleneck taller. Speed up the bottleneck itself, or clear the queue around it, and the entire company moves.

Your first automation should attack the constraint. Nothing else earns the spot.

Map the week

Take one ordinary week and write down where work waits. Not where work happens. Where it waits.

1. Which approvals sit overnight, or over a weekend?
2. What does everyone say they are "waiting on" one specific person for?
3. Which step does a deal, a case, a work order, or a proposal pass through slowest?
4. What gets done in a Friday batch because nobody had time during the week?
5. Where does a customer wait on you?

Waiting is the tell. A task that takes four hours of honest work is a cost. A task that takes twenty minutes but waits four days for its turn is a constraint. The second one is worth more to fix, and it almost never announces itself, because the person doing the twenty minutes feels fine. The queue is invisible from inside the queue.

Price the wait

Attach dollars before you attach software. If proposals wait four days on you and one in five prospects goes quiet in that window, the constraint has a price: your close rate times your average engagement, bleeding weekly. If invoices go out in a month-end batch, the constraint's price is thirty days of float you are lending your own customers, interest free.

The estimate can be rough. A rough number this week beats a precise one that never gets calculated, and the exercise usually reorders the automation wish list on its own. The task everyone complained about drops to fifth. Some quiet handoff nobody mentioned moves to first.

Automate around the judgment, not through it

In an owner-operated business, the constraint is usually a person whose judgment cannot be delegated: the attorney on the argument, the estimator on the number, you on the final call. Good. Leave the judgment alone.

Automate the feeder and the verifier instead. The feeder is everything that prepares work before it reaches the judgment: the intake packet assembled, the file summarized, the comps pulled, the draft written. The verifier is everything after. Output checked against your rules, formatted, filed, sent. When the feeder and verifier run themselves, the constraint stops being a queue and becomes what it should have been all along, a decision that takes as long as the decision takes.

An estimator who opens a complete packet instead of chasing photos does the same judgment in a quarter of the elapsed time. Nothing about the estimating changed. The waiting died.

The wrong first picks

A few candidates look attractive and will waste your first month. The technically interesting task tops the list; it usually sits nowhere near the constraint, and you will spend three weekends automating something the business barely feels. The rare task is a different trap: a quarterly report might deserve automation eventually, but with four reps a year you get almost no chances to catch what breaks. And be honest about anything you cannot explain to a new hire in five minutes, because a process you cannot specify to a person cannot be specified to a machine. Automating confusion produces faster confusion.

The constraint moves

Clear the first choke point and a new one surfaces somewhere downstream. That is the system working. Remap the week, price the new wait, repeat. Businesses that run this loop quarterly compound; each automation raises the ceiling the next one pushes against. The owners who stall are the ones who automated the annoying thing, felt done, and never looked at the queue again.

Finding the constraint from inside the building is the hard part. You are standing in the queue you are trying to see, and everyone you ask describes their own waiting, not the company's. Getting that map drawn by someone outside the building is most of what our Discovery engagement is. Either way, draw the map before you buy the tool.