AI will not save your company. I build AI systems for a living, and the tool does not fix what is broken. It makes it worse, faster.
I spent decades as a software engineer. The platform I architected at Level 3 and Lumen went from $0 to $500M in four years, and I built its first architecture myself before leading the team that scaled it. That work taught me something most software vendors will not say out loud: automation does not create value. It multiplies whatever value is already there. Automate something broken and you own a broken thing that runs ten times faster.
A lever amplifies whatever you attach it to
A lever turns a small force into a large one. It has no opinion about what it moves. Attach it to the right point and you lift a boulder. Attach it to the wrong point and you launch yourself into a wall with remarkable efficiency.
AI is a lever, and the most powerful one most companies will ever touch. Right now the market is grabbing it and attaching it to whatever annoys people most. Faster meeting summaries for a meeting that should not exist. A chatbot over a knowledge base nobody maintains. AI-written content nobody reads and nobody trusts. That is busywork running at a higher clock speed. The machine has not changed, and the broken part is spinning faster.
The question nobody asks before they automate
Before you automate anything, you have to establish that the thing being automated should exist.
That discipline has no monthly fee, which is why vendors skip it. The demo is exciting: the assistant answers a question at 2am and everyone in the room lights up. Nobody asks whether the question should have been asked, or whether the person asking was ever going to buy.
Choosing where to apply the force is the entire job. I call it finding the leverage point. It is unglamorous, and it separates the companies that scale from the companies that get busier.
The five-question test
Take any process you are about to automate and answer these in order. Write the answers down. If you cannot answer one, that is your answer.
- Does this exist on purpose? Can you say in one sentence why the process is here and what it produces? "We have always done it this way" means you are about to automate a habit.
- If it vanished tomorrow, would a customer notice? Follow the process to the person at the end of it. If nobody is better off, you are automating internal theater. Delete it instead. Deletion is free and instant. Automation is neither.
- Is this the constraint, or is it just annoying? They are different. The thing that irritates leadership daily is rarely the thing holding back growth. The constraint is the one step where, if it got twice as good, everything downstream got better too.
- Would a simple fix beat the smart one? Many processes should be deleted, reassigned to one person, or turned off. A checklist beats an AI agent more often than anyone selling agents admits. Try the dull fix first.
- Can you measure it in 30 days? Name the number that moves if this works: renewal rate, time to quote, cost per resolved question. Without a number you cannot tell success from motion, and you pay for motion forever.
Anything that survives all five is worth automating. Almost nothing survives all five on the first pass, and that is the test doing its job.
What the right leverage point looks like
When the lever meets the right point, you get a different business instead of a faster mess.
On the Level 3 and Lumen platform, the constraint was how renewals, rerates and upsells got worked across more than 1,000 product lines. Faster email would not have helped. The leverage was in a system that put the right account, the right context and the right offer in front of the right person at the right time. The revenue followed.
In a more recent build, the constraint was volume of repetitive questions that needed a person to answer. We turned 7,519 historical questions into a governed knowledge system and validated it against 370 test questions with zero errors. AI now answers more than half the questions that used to need a human. The leverage was in choosing the right constraint and the right control: governed answers, tested before launch.
Neither project started with a tool. Both started with a question: what is the actual constraint here, and is it worth amplifying? The AI came last. It should come last for you.
What to do Monday morning
Do not call an AI vendor. Not even me.
Write down the three processes in your company you most want to automate. Run each through the five questions, and be honest on question two.
You will probably find that one should be deleted, one should go to a single person with a checklist, and the third is a real candidate. Fix or kill the first two this week. They cost almost nothing and free capacity immediately.
For the survivor, confirm it is the true constraint, name the number it must move, and set a 30-day check. Then reach for the lever.
The six-minute Leverage Diagnostic on this site ranks the constraint areas of your business by dollar impact, which is the shortest route I know to question three.
The tool is the easy part. Choosing where to point it is the work. Do that work first and the tool becomes a lever. Skip it and the tool becomes an expensive way to stay exactly where you are.