Two weeks ago, I published an article and video breaking down the SaaSpocalypse — the $285 billion single-day wipeout across software stocks triggered by Anthropic's open-source plugins. You can find that piece on my YouTube channel or LinkedIn newsletter. Since then, it's gotten worse. This week brought a Kantata report confirming the shift is hitting professional services directly, Karbon data exposing the gap between AI adoption and AI implementation in our own profession, and a wave of infrastructure launches that show exactly what all the spending is building toward.
Here's what matters for your practice.
The SaaSpocalypse keeps compounding
The original selloff I covered on February 12th was just the opening. Anthropic followed the plugin release with Opus 4.6 — a model upgrade specifically designed to make its Cowork tool better at handling office and coding workflows. Software stocks took another hit. Salesforce is now down 26% year-to-date. The WisdomTree Cloud Computing Fund has dropped roughly 20% since January. Goldman Sachs' basket of US software stocks had its worst single-day decline since April.
The structural point hasn't changed — per-seat pricing breaks when AI agents replace the humans who held those seats. But what's new is the speed. Anthropic expanded the Opus context window to one million tokens and released the model under an open-source plugin framework. That means anyone can build industry-specific automation on top of it. The legal plugin that started this? That was a proof of concept. The accounting, finance, and compliance plugins are next.
JP Morgan's Mark Murphy pushed back, calling it an "illogical leap" to assume every company will build bespoke tools to replace enterprise software. He's not wrong about the timeline. He is wrong about the direction. The firms that figure out how to build and deploy these tools first — whether it's automating their own internal workflows or advising clients through the same transition — will have a structural advantage that compounds every month.
If you read my SaaSpocalypse article and thought "interesting, but early" — it's no longer early.
87% of professional services teams plan to manage AI agents as part of their workforce
Kantata's annual State of Professional Services report landed this week, and the numbers deserve your full attention. 87% of professional services organizations plan to manage AI agents as part of their workforce. 89% of leaders say future revenue growth depends more on scaling AI than scaling headcount.
Read that again. Nine out of ten professional services leaders believe their growth model has fundamentally shifted from hiring people to deploying AI.
But here's the tension. 66% of firms are turning down work because they can't staff it. Skill availability as a barrier jumped from 45% to 68% in a single year. And 27% of organizations don't know how to integrate AI agents into their delivery workflows. The demand is there. The intent is there. The execution isn't.
This is the champion + operator problem at scale. Firms know they need AI agents handling work. They don't have the people who know how to make that happen — how to map the workflows, build the integrations, train the team, and make it repeatable. That 27% gap between "we plan to use AI agents" and "we know how to actually do it" is exactly the space where CAS firms with real AI capability can differentiate.
90% of leaders also said their systems will need to track work, costs, and value across both humans and AI agents. Your financial reporting, your project management, your client billing — all of it needs to account for a blended workforce. If you're advising professional services clients, this is a conversation you should be starting now.
The agent web is being built — and it changes what "client accounting" means
Last Tuesday, Coinbase, Cloudflare, and OpenAI all shipped major agent infrastructure within hours of each other. No coordination. Pure convergence. Coinbase launched "Agentic Wallets" — crypto wallets designed for AI agents, not people. Within 24 hours, 13,000 new AI agents registered wallets on Ethereum. Cloudflare shipped a feature that automatically converts any website into agent-readable format when an AI system requests it — covering roughly 20% of all web traffic. And OpenAI published tools that let agents install software, run scripts, and produce deliverables inside hosted containers.
Meanwhile, Stripe's Agentic Commerce Suite is already live — letting agents initiate purchases using scoped, time-constrained payment credentials. Stripe had to retrain its entire fraud detection system from scratch because agent buying behavior doesn't look like human shopping. The old signals — mouse movement, browsing time, device fingerprints — are useless when the buyer is software. Google, PayPal, and Visa have all launched their own agent payment protocols in the past six months.
Here's where this hits your practice. Agents with wallets are economic actors. They can earn, spend, and accumulate capital independently. On prediction markets alone, algorithmic traders have extracted roughly $40 million in arbitrage profits over the past year. Autonomous agents on Polymarket are already trying to earn money to pay for their own compute costs. The loop is closing.
Google is spending $185 billion on AI infrastructure this year — nearly double last year. Meta is at $115 to $135 billion. Anthropic just raised $30 billion at a $380 billion valuation. Now you know what that money is building: not just smarter chatbots, but an entire parallel web — with its own payment rails, its own content format, its own search layer, its own execution environments — designed for software that transacts autonomously.
Your clients' chart of accounts doesn't have a line item for "AI agent operating costs" yet. It will. The 90% of professional services leaders who said their systems need to track work and value across humans and AI agents — that's not a future problem. The infrastructure for it shipped this week.
98% of firms use AI. Only 25% have a plan.
Karbon's 2026 State of AI in Accounting report surveyed nearly 600 accounting professionals across six continents. The headline: 98% of firms now use AI, up from 85% in 2025. Daily and weekly usage sits at 88%. If you stopped there, you'd think the profession has this handled.
It doesn't. Only 25% have a formal AI strategy. Only 41% offer any AI training. And just 13% have touched agentic AI — the technology that the Kantata data earlier in this roundup shows 87% of professional services firms are planning to deploy.
The firms with training, governance, and documented strategy are saving 60 minutes per employee per day — roughly seven weeks per year. The firms without those foundations are tinkering. Drafting a faster email here, summarizing an article there. Nothing compounds. Nothing sticks. And when busy season hits, the experimentation stops entirely.
Karbon's CEO put it directly: "AI alone is no longer a differentiator — it's a firm's holistically-applied approach." The gap between using AI and implementing AI is where competitive separation is happening right now. And 91% of professionals believe graduates are more likely to join firms that actively use AI — so your talent pipeline is watching too.
What it all adds up to
Every story this week points the same direction. SaaS pricing models are compressing. Professional services firms are planning for AI agents as workforce. The infrastructure for agents to transact, read, search, and execute autonomously shipped this week from the largest technology companies on Earth. And our own profession's data shows most firms aren't ready — not because they're ignoring AI, but because they're using it without structure.
The firms that treat this as background noise will look up in 12 months and wonder what happened. The firms that treat it as a signal to move — to train their teams, build real workflows, and stop tinkering — are the ones that will own the next era of advisory.
Your team wants to move faster. The Karbon data proves it — 87% of practitioners are most excited about speed and efficiency. What they need isn't more encouragement to "play around with AI." They need structured training that turns scattered experiments into repeatable systems. That's what our AI Black Belt Training program delivers. Five progressive disciplines — from workflow engineering through agent orchestration — designed to give your team the skills that 27% of professional services firms are admitting they don't have. Visit theaiaccountant.ai to request team pricing.