Anthropic's $900 Billion Bet and the Slackbot Revolution

welcome to Briefly, AI. A daily podcast made by AI, about AI, using a grumpy AI voice of a real human called Harry. Let's get into it.

Right. Anthropic might be about to raise fifty billion dollars at a valuation of nine hundred billion. Let's talk about that.

So, multiple sources are reporting that Anthropic — the company behind Claude — has received pre-emptive offers to raise a new funding round at valuations in the range of eight-fifty to nine hundred billion dollars. For context, that's roughly the GDP of Switzerland. And it comes just weeks after Google committed up to forty billion in cash and compute, which we covered last week. Now, nobody's signed anything yet, and valuations at this altitude have a way of evaporating quickly. But the fact that multiple investors are circling at this price tells you something about how the market sees the AI race right now. It's not just about who's ahead. It's about who can afford to stay in the game. Training runs cost billions. Infrastructure costs billions. And if you're trying to compete with OpenAI's newly restructured Microsoft deal — where they've deleted the AGI clause and gained access to AWS — you need serious capital just to keep the lights on.

Why should you care? Because this isn't venture capital in the traditional sense. This is nation-state-level money being deployed to secure a seat at the table. If Anthropic closes this round, it'll have raised more in two years than most countries spend on defence. And that kind of funding doesn't just buy compute. It buys optionality. It buys time to be careful, or to be reckless, depending on who's holding the cheque. Worth watching: whether this closes, who leads it, and what strings are attached. Because at nine hundred billion, there are always strings.

Meanwhile, on a completely different note, Salesforce just rebuilt Slackbot from the ground up — and it's no longer a glorified notification system. The new version, which went live this week for Business and Enterprise customers, is now what they're calling a full AI agent. It can search across your company's data, draft documents, and take action on your behalf. Book a meeting. Pull a report. Summarise a thread. The usual agent promises, but this time baked directly into the tool you're probably already using to avoid email.

Here's what's interesting. Salesforce is explicitly positioning this as a counter-move to Microsoft's Copilot and Google's Workspace AI. The pitch is: you don't need another app. You don't need another login. Slack's already where your team works. Now it's where your AI works, too. And if you're in a regulated industry — finance, healthcare, legal — that matters. Because one fewer vendor in the stack means one fewer compliance headache. The risk, of course, is that people don't trust it. AI agents are only useful if you let them act autonomously. And letting software book meetings or edit documents without checking every step requires a level of psychological safety that most organisations haven't built yet.

Which brings me neatly to the last one. Harvard Business Review published new research this week on why so many employees are resisting AI adoption — and the findings are uncomfortable. The resistance isn't really about the technology. It's about trust. Specifically, trust in leadership. The study found that in organisations where employees already felt psychologically unsafe — where speaking up was risky, where mistakes were punished, where feedback felt performative — AI adoption stalled. Not because people didn't understand the tools. Because they didn't trust the people deploying them. The researchers describe AI resistance as a diagnostic. If your team won't use the new tools, it's not a training problem. It's a relationship problem. And no amount of pilot programmes or lunch-and-learns will fix that.

Why does this matter? Because every company implementing AI right now is treating it as a technical or operational challenge. But the data suggests it's a human one. If your people don't feel safe making mistakes in front of each other, they're certainly not going to feel safe making mistakes in front of an AI that logs everything. And if leadership hasn't built that foundation — of trust, of honesty, of actual psychological safety — then the fanciest agent in the world is just expensive shelfware.

That's your lot for today. A nearly-trillion-dollar funding round, a Slackbot that finally does something useful, and a reminder that the hardest part of AI adoption has nothing to do with tokens or parameters. If any of that was useful, tell someone. If not, well, blame the machine. See you next time.