Office Red Tape Is Saving Your Job From AI. Here's For How Long.

What happened

Your office takes weeks to approve a new laptop. That is the reason AI hasn’t taken your job. So how long does that protection last? My answer: for the job you have today, about five years, not ten. For the job you want next, it’s already gone. That’s my estimate, not a measurement, so here is the evidence.

This started on the 9th of November. A founder with 318,000 followers posted: “It’s gonna take 10 plus years for AI to diffuse through most enterprises”. He listed 11 steps a big company goes through before one AI agent goes live. The last line of the post said that if you want help, call his company.

That doesn’t make him wrong. I am a data auditor. When the person with the answer also sells the fix, you check the numbers yourself. And the opposite camp exists too: the head of one AI lab warned that half of entry-level office jobs could go in one to five years.

So is the queue real? Yes. The numbers back him.

1. India. The Reserve Bank asked 612 banks and lenders whether they were using AI or building it. 127 said yes. One in five.
2. America. In the latest count, 24% of businesses used AI in any part of their work. In companies with 250 people or more, 44%. More than half of the big companies still said no.
3. The ones that said yes. Deloitte surveyed 3,200 business leaders. Only one in four had moved even 40% of their AI experiments into real use.
4. Headcount. Economists surveyed nearly 9,000 executives in seven countries. 78% said their firm uses AI. Asked what AI did to headcount in the last three years, the answer was essentially nothing.

He’s right about the queue. He’s wrong about what it protects.

Where the control failed

The queue only counts what the company approved. It doesn’t count what your colleagues did on Tuesday.

Verizon publishes a big report on data breaches every year. This year, 45% of employees regularly use AI on their work devices. And among the people using AI tools the company never approved, 67% do it through a personal account. MIT researchers found the same gap last year: 40% of companies had bought an AI subscription, but at more than 90% of them, staff were using personal AI tools for work.

Here’s the audit word: control. A control is a step that exists to stop something going wrong. Two signatures on a payment. A security review before new software. You call it red tape. An auditor calls it controls.

A lot of that AI use was never approved. That’s a control being bypassed. The official number says AI hasn’t arrived. The work says it has. Your company hasn’t adopted AI, your colleagues have, and nobody’s counting.

The evidence an auditor should ask for

Red tape slows the company. Does it protect you? Only the job you already have.

It doesn’t protect the job you want next. Stanford economists track payroll for millions of American workers. For people aged 22 to 25 in jobs like junior coding and customer support, employment is now 19% below where it would be if they’d kept pace with everyone else. The authors call that an early warning, not proof. The likeliest reason is simple: the job was never posted. In India, Zoho’s founder said it plainly in August: “We have not laid off people, but we are not creating new jobs either.”

It doesn’t protect your price. On its results call, Infosys said clients are asking for the productivity benefits of AI to be passed on. In plain words: you have AI now, so charge us less. Put two of its numbers side by side. AI revenue grew 29% in a single quarter. The forecast for the whole company this year: one to two percent. TCS tells the same story, with AI revenue at $3.1 billion a year. My reading: that is AI arriving as a discount.

To be fair, the CEO of TCS said its job cuts weren’t because of AI, and I’ll take him at his word. My point is different. The client didn’t deploy a single agent. Nobody passed a single control. And the pressure still lands on your billing rate, on your raise, on your team’s next 10 hires.

Red tape guards the door you’re already inside, but it doesn’t guard the door your junior needs, and it doesn’t guard your price.

So when does the queue break? I think we have around five years, and two or three more, no more, if a big security incident hits a large company. Why five and not ten: the share of firms using AI more than doubled in two years. Why security: in April, the cloud company Vercel was breached, and its own report says it started with a third-party AI tool used by one employee. In 2023, Samsung staff put internal data into ChatGPT and Samsung banned it on company devices. Press reports say it took three years to officially bring outside AI tools back.

Then I audited my own prediction. In 2019, data on about a hundred million people was taken from Capital One’s cloud. The next year, Amazon’s cloud business grew 30%. So I’m half right. A big incident freezes the company it hits, maybe for years, like Samsung. It doesn’t freeze everyone else.

So the honest answer isn’t one number. It depends on your company. One reply to the ten-year post argued that India will run at two speeds: a fast lane of global capability centres and fintechs, and a slow lane of traditional India Inc. That’s an argument, not a measurement.

The test to run this week

Which lane is your company in? Four questions. Count your yeses.

1. Is your work billed to a client by the hour or by the head? If yes, the discount reaches you first.
2. Has your team stopped hiring juniors? If yes, the door is already closing behind you.
3. If your company blocked every personal AI account tomorrow, would your team’s output drop? If yes, AI is already doing the work. The official number is just late.
4. Is your company outside the heavily regulated world? Not a bank, not a government, not defence. If yes, your queue is shorter.

Three or four yeses: don’t plan on ten years, plan on sooner than five. One or none: you probably have longer than my five. It’s a rule of thumb, not a forecast.

Every AI tool stuck in that queue is waiting for the same thing: someone who can show it’s safe, who can test it, check what it produces, and say what it’s allowed to touch. The AI doesn’t replace that person. That person is how the AI gets in.

So this week, take one task you already do with AI and write one page:

1. What data goes in.
2. What you check before you trust the answer.
3. What it must never touch.

That page is a control, and you’ve just become the person your company is waiting for. Don’t hide behind the red tape. Write it.