Stop Talking About AI. Give It a Job.

There is something strangely familiar about the way we are talking about artificial intelligence. Everywhere I go, people are discussing AI. They are discussing models, agents, benchmarks, reasoning, multimodal systems, autonomous coding, productivity, disruption, alignment, regulation, jobs, and the coming transformation of society. They are curious. They are excited. They are sometimes frightened. But after listening for a while, I keep wanting to ask a very simple question: What did you actually make it do? This isn’t a criticism of curiosity. Curiosity is where technology begins. But technology does not transform work because people talk enthusiastically about its possibilities. It transforms work when somebody takes the machine and gives it a job. I Have Seen This Movie Before The personal computer went through something similar. The arrival of the IBM PC, the XT, the AT, the 386 and the 486 was astonishing. Each generation brought dramatically more computing power, more storage, better software and new possibilities. People talked about what computers would eventually be capable of doing. And they were right. But none of that mattered to the bookkeeper sitting at a desk with a pencil, ledger and adding machine. The important question for that bookkeeper wasn’t: “How powerful is this computer?” It was: “Can this thing make my work easier?” The breakthrough came when the answer became yes. A spreadsheet wasn’t important because spreadsheets were technologically impressive. It was important because someone could stop doing arithmetic by hand. A database wasn’t important because databases were intellectually interesting. It was important because someone could find a customer record without searching through filing cabinets. The personal computer became revolutionary when it stopped being an object of fascination and became an appliance of work. AI Is Still an Object of Fascination Much of the current AI conversation hasn’t reached that stage. We talk about enormous models. We talk about context windows. We talk about agents. We talk about autonomous software engineers. We talk about artificial general intelligence. We talk about whether AI will replace programmers, lawyers, teachers, doctors, analysts and executives. But there is a danger in endlessly discussing the machine instead of the work. AI isn’t a profession. AI isn’t a strategy. AI isn’t a business plan. AI is a machine. Give it a job. The Real AI Question Suppose you run a company. You have somebody spending four hours every morning collecting information from different systems and putting it into a report. Don’t start by asking which frontier model you should buy. Ask: Can the machine do the job? Give it the files. Give it the APIs. Give it the database. Give it the rules. Give it a test dataset. Give it access to the tools it actually needs. Then make it produce the report. Now you have learned something. Maybe it works. Maybe it fails spectacularly. Either result is more valuable than another meeting about the future of AI. Because now you have an engineering problem. And engineering problems can be solved. The 90 Percent Trap Modern AI makes this even more interesting because it is often remarkably good at producing the first 90 percent of a solution. That creates a new temptation. Someone generates a beautiful report. Someone generates an application. Someone generates a deployment script. Someone generates an analysis. Everyone looks at it and says: “Wow.” But the last ten percent may contain everything that matters. Is the data correct? Did the system understand the exception? Did it preserve the business rule? Did it introduce a security vulnerability? Will it work next month? What happens when the network disappears? Who notices when it fails? Who is responsible? The last ten percent isn’t necessarily more typing. It is judgment. And judgment becomes more valuable when production becomes cheap. This Changes What We Should Hire For If AI makes competent-looking output abundant, then organizations have to change what they mean by competence. The valuable employee may not be the person who can produce the most text, code or slides. It may be the person who can recognize when the machine is wrong. The person who understands the customer. The person who knows how the old system actually works. The person who understands the network. The person who knows what can safely be automated. The person who can take an ambiguous problem and turn it into a working system. The person who can say: “No. That isn’t the problem we’re actually trying to solve.” That is not anti-AI. It is the opposite. It is recognizing what AI is actually good for. AI Should Be Given Work, Not Worship There is a strange social phenomenon developing around AI. Some people treat AI almost like a new civilization. Others treat it like a new employee. Others treat it like a dangerous alien intelligence. But the most productive relationship may be considerably less dramatic. Treat it like a very unusual machine. Give it a task. Give it tools. Give it constraints. Give it data. Test the result. Measure the result. Keep what works. Throw away what doesn’t. Then give it another job. That’s how technology becomes infrastructure. The Person Who Knows How to Use the Machine Wins The important worker of the AI era may therefore not be the person who knows the most about AI. It may be the person who already knows something worth doing. The network engineer who uses AI to automate infrastructure. The teacher who uses AI to create individualized exercises. The mechanic who uses AI to diagnose an intermittent fault. The accountant who uses AI to reconcile thousands of transactions. The security engineer who uses AI to examine enormous quantities of logs. The programmer who uses AI to eliminate the repetitive parts of programming and spend more time thinking about architecture. These people don’t need to spend their day talking about artificial intelligence. They need to use it. The Coming Divide I suspect the emerging divide won’t simply be between people who have AI and people who don’t. It will be between people who talk about AI and people who incorporate AI into useful work. The first group will have opinions about agents. The second group will have agents doing things. The first group will debate which model is best. The second group will have tested three models and know which one works for their particular job. The first group will discuss productivity. The second group will have eliminated six hours of repetitive work from their week. The first group will ask whether AI is going to change their industry. The second group will already be changing it. The Computer Didn’t Replace the Bookkeeper The PC did not simply arrive one morning and eliminate bookkeeping. It changed the economics and mechanics of bookkeeping. It automated arithmetic. It changed record keeping. It changed reporting. It changed auditing. It changed what a competent bookkeeper could accomplish. Eventually, the person who insisted on doing everything with a pencil and adding machine wasn’t demonstrating craftsmanship. They were operating outside the technological reality of their profession. AI may do something similar. It may not eliminate every profession. It may instead eliminate enormous amounts of work inside professions. And that distinction matters. The future may belong neither to humans alone nor to machines alone. It may belong to people who know how to construct effective combinations of the two. So Stop Asking What AI Can Do That question is becoming almost useless because the answer is becoming enormous. Ask something smaller. Ask: What job should I give it? Then give it one. Make it work. Measure it. Break it. Fix it. Give it another. Eventually you won’t be talking about AI anymore. You’ll simply be talking about getting the work done. And that is probably when the AI revolution will have actually arrived.