Another week in AI and the pattern is getting hard to ignore. The noise around new models and tools is hiding a bigger shift underneath. Workflows are breaking apart. Creative pipelines are shrinking from teams to prompts. The economy is leaning on AI more than anyone expected. Enterprise software is finally handing real workloads to agents. And at the platform level, the balance of power is shifting again. None of these trends stand alone. They all point to the same destination: the collapse of the old work model and the rise of systems built around AI agents instead of human throughput. The first signal is the one most people feel every day but rarely say out loud. AI is not replacing jobs, it’s exposing the parts of work nobody wanted.
The Work No One Wanted
Everyone’s worried AI will automate their job. We should be more worried it will expose how much of our job was fake work. I just read a Linkedin post mocking AI email ... everywhere people blame AI for “ruining” hiring, sales funnels, spamming customers. But AI didn’t invent these bad workflows. It just put them under a spotlight. For 20+ years we optimized knowledge work for activity, volume and tracking, not outcomes. Now we’ve handed that system a tool built for infinite speed and scale. Of course it snapped.
Hiring broke first We turned talent into an SEO game:
- Candidates blast 500 applications a day
- Recruiters rely on filters that judge formatting over skill.
Add AI and suddenly you get infinite polished resumes and a totally clogged funnel. That’s not an AI problem. That’s a workflow problem.
Sales and support were next We told reps the job was “log everything.” AI shows up and auto-logs, auto-emails, auto-sequences, Now the CRM is full… and nobody trusts what’s in it. Again, workflow problem. Same story in support and marketing. If your strategy was “more volume,” AI just put it on steroids.
Here’s the pattern:
- If a process can be abused by pushing a button 1,000 times, it was never measuring real work.
- If AI junk slips through, your system was never measuring quality, only compliance.
AI is not the villain. AI is the stress test. It ripped off the duct tape and showed us what was underneath. The next model of work isn’t “everyone use AI tools.” It’s agents and Cloud Employees doing the parts of the job humans were never built for at this scale:
- Screening, qualifying, summarizing, routing in real time
- Reading signals across systems and surfacing the 3 things that matter, not the 300 that just happened
Not to replace judgment ... but to clear out the fake work that prevented humans from using it. The real fight isn’t AI vs jobs. It’s broken workflows vs rebuilt ones. Once AI exposes fake work, the next question is which functions fall first. Creative work turned out to be the first domain where the entire pipeline collapsed overnight.
The Creative Pipeline Is Shrinking
I played with Nano Banana Pro this weekend and yeah, it’s an amazing creative tool. But that framing misses the real point by a mile. This isn’t a tool. This is Google hiring your next creative employee for you. Look past the demos. The WIRED review hints at the real story. It’s not the images that matter, it’s the fact that it does the work creative teams usually grind through, and it now sits inside Ads, Slides, Workspace, and Search. That is not a feature upgrade, it more like a platform taking over the work. For the last twenty years, enterprises, myself included, built entire teams around production.
- Designers
- Coordinators
- Agencies
- Revisions
- Approvals
Nano Banana Pro might just collapse that entire pipeline into a prompt. And if that’s true, the competitive gap is about to move faster than most teams are ready for. It had me thinking. Maybe the winners in the next cycle won’t be the ones with the “best” creative AI. Maybe the advantage shifts to the companies that accept the deeper truth, that creative capacity is now effectively infinite, and reorganize their teams around that reality. Because once the platforms start embedding agents directly into your workflows, the job has to change. You stop managing execution and you start orchestrating outcomes. The era of tools is ending. The era of autonomous creative employees has begun. And the creative collapse is just the micro version of a larger truth. The same force breaking workflows is now propping up the entire economy.
AI Is Holding Up the Economy
American's are arguing about AI like we have a backup plan. We don’t. Every debate about AI has turned into a belief system.
- You’re either convinced AI is a bubble that will crash everything
- Or you believe it’s the engine that will save the economy
And everyone defends their position with the same intensity people used to reserve for religion or politics. But strip out the emotion and look at the data sitting in front of us. Seriously.
- Credit delinquencies are climbing
- Consumer sentiment is scraping 15 year lows
- Views of personal finance were the lowest since 2009
- 4 year college grads can’t get hired (25% unemployment)
- Auto loan delinquencies at highest level on record (since 1994)
- Outside of government, job growth is shrinking
Honestly, you could basically pick any economic indicator you wanted and it would not be good. I don't think this is noise. It’s the early innings of a recession. Now look at the other side of the ledger.
- Trillions of dollars are flooding into the AI Economy ($3–7T in AI infrastructure by 2030 alone)
- Data centers, power grids, chip fabs, sovereign AI deals, federal contracts
- Amazon just announced a fifty-billion-dollar buildout for government AI infrastructure
- Nvidia, Alphabet, Microsoft, Meta, Oracle, Amazon, Tesla, and half the Fortune 500 are spending at historic levels
Here’s the part everyone keeps skipping over. If you remove AI, the U.S. economy is already underwater. AI isn’t inflating the economy. It’s propping it up.
And this is where the poverty math matters. When Michael Green recently recalculated the real cost of simply participating in American life, the floor wasn’t $31K. It was closer to $140K. The median household makes $80K. Which means half the country is already below the poverty line before AI even enters the conversation. So here’s the paradox.
Many of the most skeptical of AI are the same people most exposed to a collapsing cost structure. And the technology they distrust happens to be the only force with enough scale to bend that cost structure back down.
- Better productivity.
- Smaller companies with bigger capability.
- Lower admin burden in healthcare and education.
- More entrepreneurship.
- More meaningful work per person.
- Less overhead per job.
AI is not a magic cure. But when every non-AI indicator is flashing recession, and AI is the only sector at this scale generating investment, hiring, and output, you have to ask ourselves... If AI fails, what exactly is left to catch us? That’s the wager. Not whether AI is perfect. But whether the alternative is even survivable. If AI is holding up the economy at the macro level, enterprises are now proving it at the operational level. Sierra is the first loud example.
Sierra Proves Enterprises Are Ready for Agents at Scale
AI finally just hit the enterprise at real scale with a $10B valuation, following a $350M round. Bret Taylor's (former Salesforce Co-CEO) Sierra hitting $100M ARR is the loudest signal yet that enterprise AI agents are not a toy market. They’re real. They’re funded. And the Fortune 500 is buying. Look at the AI evolution stack right now.
- AI GPU Chips (winning): Nvidia had one of the most explosive quarters in tech history. Google is pushing TPUs into the wild with Anthropic. Meta is following.
- AI Data centers (winning): CoreWeave, Oracle, Amazon committing $50B for new government capacity.
- LLM Models (winning): OpenAI and Anthropic defined the category, and Google just vaulted into first with Gemini 3, Opus 4.5 groundbreaking in coding capabilities.
- Developer tools (winning): Copilot, Cursor at a $29B valuation, Lovable, all eating the software IDE.
All that momentum has been below the enterprise workflow. Infrastructure. Models. Dev tools. Sierra is the first meaningful example of AI breaking through one layer higher. Not productivity or copilots but actual back-office, customer-facing, go-to-market work at scale.
That’s why this $100M number matters. It proves that the enterprise is finally comfortable letting agents handle real operating workloads. Now, if you look under the hood, the magic fades a bit. A lot of this is still custom AI builds, heavy consulting, heavy integration. Impressive traction, but not a truly scalable stack yet.
But it’s the same pattern we saw in early cloud. Massive early wins built through custom work and consulting, then real platforms emerged. This is why I love seeing Sierra’s rise even though they’re technically competitors. It validates the category. It confirms enterprise demand. And it exposes the next opening. If I’m honest, I think Atonom has already crossed that line :)
And once enterprises lean in, the real question becomes: who controls the platform they’re building on? That answer just changed.
Google Becomes the AI Stack
Google wasn’t supposed to win the AI race. It was supposed to be the dinosaur that missed it. Instead, sometime between spring and now, Google turned into AI’s landlord — and everyone else started paying rent. Alphabet is up ~70% at $3.8T passing Microsoft. The “did Google miss it?” debate is over. The question now is: who just lost pricing power because Google flipped its stack on?
Here’s what changed:
- They wired AI into the cash machine, not around it Gemini 3 isn’t a side project. It’s wired into Search, Android, and Workspace. AI Overviews don’t kill search, they drive more of it and keep you inside Google.
- They built a parallel Nvidia inside the company TPUs were a science project. Now they’re product. Google trains on its own chips, then rents those same chips through Cloud while a lot of “AI infra” startups just resell compute.
- The antitrust bomb fizzled Worst case was: break up Search and nuke default deals. Instead, Google got a slap on the wrist and kept the engine. In an AI world, that distribution is lethal.
- Buffett called it “real” Berkshire Hathaway didn’t buy AI hype. It bought Alphabet. That’s conservative capital saying: this is core infra with a cash hose attached.
- “AI risk” turned into “AI gravity” AI was supposed to wreck search economics. Instead, it’s making it harder for anyone else to peel off users or ad dollars while Google funds the whole experiment with its own profits.
Now this runs straight into Black Friday. Tomorrow a big chunk of shoppers won’t start on a retailer site or even a search bar, they’ll start by prompting an AI. And in a lot of cases, that AI is still Google. The interface changed. The landlord didn’t. The bigger picture:
- SaaS is getting de-rated into a slower, utility-like asset class.
- Google is getting re-rated as the full AI stack: Chips → Cloud → Models→ Platform → Distribution → Cash flow.
When Google becomes the full stack, the constraint shifts. Compute isn’t the bottleneck anymore. Work is. McKinsey just quantified that shift.
McKinsey Confirms the New Workforce
McKinsey & Company's new study shows 57% of today’s work hours are technically automatable by AI and robots.
Most people see that and jump straight to, “Half the jobs are gone.” I don’t think that’s what’s happening. This isn’t a layoff story, it’s a workflow story.
McKinsey is basically telling us:
- Almost half of what we do is structured, repeatable, rules-based.
- Agents and robots can do that work today with current tech.
The real question isn’t if they will… it’s who chooses to use them first.
The interesting part: the skills don’t disappear. Something like 70%+ of today’s in-demand skills show up in both automatable and non-automatable work. Translation: skills are being reassigned between humans and AI, not deleted. That’s how I think about AI Agents/Cloud Employees. You’re not replacing your team. You’re stripping out the 57% of work that never needed a human in the first place:
- The copy-paste admin.
- The “check three systems and update the fourth” work.
- The rote follow-ups nobody has time for.
What’s left for humans? Judgment. Trust. Strategy. Ambiguity. Relationships. If McKinsey is right, the winners won’t be the companies that “use AI.” It’ll be the ones that rewrite jobs so humans and AI Agents/Cloud Employees are working the same role from two different sides.
If 57% of our team’s work can be done by AI today, which parts do we start with?
Zoom Out
These are not isolated shifts. They are one movement up the stack. For decades, work relied on coordination, formatting, reviews, approvals, and handoffs. AI didn’t replace those layers. It revealed how brittle they always were.
The next era won’t be defined by tools or screens. It will be defined by environments where digital employees operate alongside human ones. The advantage will belong to the companies that rebuild their systems around that reality instead of trying to bolt AI onto the old model.
This is not the future of work. This is the reset that makes the future possible. And the winners won’t be the teams that adopt AI tools. They’ll be the ones that redesign their work around digital employees from the ground up.
