There's a funny video doing the rounds on Instagram right now by the "Corporate Bro" having a laugh that every keynote, every panel, and every LinkedIn post in 2026 seems to be built around a single word: "AGENTIC." Say it enough times and it stops sounding like English.

I'll hold my hands up here. I'm part of the problem. I'm building an AI agent team for coaches as I type, so yeah I've been sucked in right alongside everyone else.

But there's the thing about words that turn into punchlines. The punchline comes from the ubiquity. And the ubiquity comes because the thing works and is popular.

This week's stories are the proof. Agentic AI is not just the buzz word people talk about on stage. It's now the thing that trades your money, builds your marketing, and sits in on your sales calls. The twist...not everything is as it seems. The ying of great products comes with the yang of mass layoffs. Some of those layoffs are genuine. Some are political cover. Telling the two apart is the whole game.

From giving you answers to doing the thing

For about two years, AI mostly handed you information. You asked, it answered, and then you did the actual work. That arrangement just broke.

This past week, Robinhood gave its 27 million customers the ability to let an AI agent trade stocks and spend on a credit card on their behalf. Not suggest trades. Make them. You can tell an agent to rebalance your portfolio, or send it off to buy the cheapest flight available, and it goes and does it. Maybe even while you sleep.

To their credit, Robinhood screwed on some sensible stuff. The agent can only touch money you specifically deposit into it, you can set spending caps, and there's a one-tap kill switch if it starts misbehaving. However, they also say plainly that agents make mistakes, and that keeping an eye on the account is your job, not theirs.

Yeah...a regulated brokerage just handed an AI agent a credit card and effectively said, "good luck, keep an eye on it."

That's the shift. We've moved from AI as a clever assistant that informs you, to AI as an actor that does things and then leaves you holding the consequences. And it isn't confined to one corner of the economy. It's in e-commerce, where Shopify and Google have built a shared standard so brands can sell directly inside AI chats. It's in finance. It's in sales. Like I said a while back...once your customer, your trader, and your shopper can all be a bot, every playbook ever written for humans needs a rewrite.

The small business land grab (and why I still think Google wins)

Google quietly expanded Pomelli. For those of you not familiar, Pomelli is a marketing tool out of Google Labs and DeepMind. Now a small business can hand it (Pomelli) a website or a few prompts and get back its entire "Business DNA." Tone of voice, colour palette, typography, the works. Then it builds the campaigns, the brand book, and the website off the back of it.

Think about what that quietly removes. A decent brand identity used to mean hiring a designer, a copywriter, and finding a few thousand pounds SME owners probably didn't have spare. Pomelli turns that whole process into a conversation.

This is why I keep saying Google is best placed to win the AI wars, and it has very little to do with who builds the cleverest model. 99% of businesses are SME's. A lot of them have been priced out of great marketing tools for decades. Whoever knocks down that cost barrier captures the single biggest pool of economic activity on the planet.

Now look at what Google already owns. Search, where people find things. Maps, where they make local decisions. The advertising rails, where the money changes hands. And now the brand-creation layer right at the very start of the journey.

The loop closes on itself.

Lower the cost of looking professional, and millions of small firms can suddenly build distribution and tap into a revenue pool they simply couldn't reach before. That's the part that genuinely moves productivity and GDP. When the agentic dust eventually settles, the company sitting underneath the most small businesses is the one I'd put my money on.

The trust tax nobody put on the keynote slide

Now for the less glamorous half. Because handing agents the keys creates a bill, and somebody has to pay it.

Behind the slick shopping demos, payment processors and banks are watching chargebacks, disputes, and fraud alerts climb, according to a piece on Quasa. An over-eager agent buys the wrong thing, or a customer regrets the "autopilot" purchase, and the bank's risk engine lights up. The rails are tightening: stricter limits, more manual reviews, and higher fees on anything that looks risky.

The lesson buried in there is a good one. The future of agentic commerce won't be the quickest shopper. It'll be AI agents that understand the rules, read the policies, and can work out whether a transaction is actually safe for everyone involved. Future agents will behave more like cautious compliance officers, rather than trigger-happy personal shoppers.

We've all seen this film before, by the way. Early e-commerce was a fraudfest until the industry added trust layers. Things like card verification and chargeback protection. Agentic commerce is in that same awkward teenage phase right now, growing faster than its judgement can keep up with.

Progeress used to be raw capability. Now it's trust. The winning agents of the future will be the ones that are genuinely accountable for their choices

The AI washing problem, and the bit about your job

Which brings me to the part most people tiptoe around. A lot of what's being sold this year as "agentic transformation" is just "AI Washing". We need our AI agents to be accountable, but our corporate execs struggle with the concept for themselves.

ClickUp cut 22% of its staff, rolled out roughly 3,000 internal AI agents, and framed the whole thing as a bold leap toward a "100x org" rather than a cost cut. Maybe it is. But Gartner found that while around 80% of firms using autonomous tech have cut jobs, the financial returns mostly aren't showing up yet. On top of that, more than 40% of agent projects are expected to fail by 2027, and 88% of pilots never make it into production, per figures cited by TheStreet.

A good chunk of these layoffs aren't really productivity stories. They're ordinary bad decisions wearing an AI costume, because "we restructured around AI" plays a lot better in a boardroom than "we overhired and misread the market."

So that's the cynical truth. Here's the uncomfortable one sitting right next to it.

The capability underneath the spin is real, and it's aimed squarely at your desk work. According to data in TheStreet, 87% of organisations already lean on AI for prospecting and forecasting, and somewhere between 60 and 70% of entry-level sales-rep tasks could be automated within 12 to 24 months. Some firms are now putting AI on live calls as a visible participant, fielding technical questions in real time. Gartner expects 40% of enterprise apps to carry task-specific agents by the end of this year, up from under 5%. There's even a one-person startup, Polsia, that raised $30 million at a $250 million valuation by running almost everything on automation.

So both things can be true at once. Your employer might be using AI as a convenient excuse, and an agent might genuinely be able to do real slices of your role. If your work happens on a computer, in part or in full, the safe assumption is that an agent is being pointed at it.

That sounds bleak. It could be. Depends on your perspective...and you get to choose which side you stand on. You can be the person directing the agents and checking their work, or the person whose tasks get quietly handed over to them. If you want to be part of the first group the data is in your favour. Research shows companies that use AI to support their people consistently outperform the ones treating it as a straight swap for headcount.

Where this leaves you

Agents have crossed the line from talking to doing, and they've done it across e-commerce, finance, marketing, and much more.

Some of that is real, life-changing value, especially the kind that finally hands small businesses tools they could never afford. Some of it is theatre. Your job, as someone trying to stay useful and employed, is to stop treating "agentic" as a buzzword you can wait out, and start treating it as a skill you build.

So go ahead and laugh at the meme. Then go and learn to drive one of these things, before someone decides you're cheaper to replace than to retrain.

If you remember nothing else: Agentic AI isn't a panel-talk abstraction anymore; it's trading, selling, marketing, and quietly absorbing desk work as you read this. The only question that really matters now is whether you'll be directing the agents or replaced by them, so pick one task you do on a computer every single week, hand it to an AI agent before next Monday, and start building that instinct while it's still a choice rather than a redundancy letter.

Thanks for reading this far, see you next week.