Nobody sat down and decided their company would “Google AI.” It just showed up. A support lead switched on an auto-tagging feature. Someone in finance noticed the accounting software could read invoices now. A marketer started pasting product notes into a chatbot at 5 pm on a Friday.
Give it a year and half the company is using AI, with no policy, no training, and no idea what it’s costing or saving. That’s how it’s really happening, and it’s a better starting point for this topic than any prediction about the future of work.
Support desks changed first
Customer service is the easiest place to see it. The chatbot on a retailer’s site now answers the dull stuff: where’s my parcel, how do I reset this, can I return it. If the customer is furious, or the issue is strange, a decent setup hands over to a person fast. A bad one traps you in a loop, and we’ve all been there.
Less obvious is what happens to the agents. Software reads each new ticket, sorts it, guesses the urgency, and drafts an answer. The agent checks it, fixes it, sends it. The job stops being copy-paste and becomes judgment. Some people like that change. Others find it stressful, because now the hard cases are all they see. Managers should notice this.
The dull departments gain the most
Finance is a good example. Invoice matching is tedious, and software is good at tedious. It reads the invoice, checks it against the purchase order, and only bothers a human when something doesn’t fit.
Forecasting is where the real money hides. Order too much stock and cash sits on a shelf. Order too little and customers go elsewhere. Amazon’s warehouses, with their little orange drive-unit robots carrying shelves to packers, only work because software predicts what will sell and where to keep it. Smaller firms can now buy a simplified version of that.
Marketing: fast is easy, good is not
Everyone in marketing uses AI to write. You can tell. Product pages start sounding identical: same rhythm, same cheerful vagueness. Customers scroll past.
The teams doing it well use the tool for the boring first draft, then a person rewrites it with the stuff only they know: which customers complain about what, what the product is bad at, what’s actually funny about the industry. [YOUR STORY: a time a generic AI draft fell flat, or one you fixed]
Hiring is the risky one
Recruiters use AI to sort CVs and book interviews. It saves hours. It also has a nasty failure mode: if the tool learns from past hires, and past hires were biased, the bias gets baked in and hidden behind a score. Amazon reportedly scrapped an experimental recruiting tool a few years back for this reason. If AI touches who gets interviewed, test it, and let a human make the final call.
What it can’t do
It’s confidently wrong sometimes. It doesn’t know your business unless you tell it. It won’t calm a client, close a tricky deal, or take the blame when things go sideways.
If you want to try it, Pick one irritating task. Just one. Something that eats hours and is easy to count, like tagging tickets or typing invoice data. Run it with a single team for a month. Check where your data goes before you paste anything private into a tool. And count something real afterward: hours saved, errors avoided. If the number didn’t move, stop.
There’s no award for using AI. There’s only the problem you had, and whether it’s smaller now.
FAQs
Is AI only for big companies?
No. A five-person shop can get value from scheduling and email sorting tools that cost very little.
Will it take my job?
Some tasks, yes. Whole jobs that depend on judgment and trust, much less so.
