I've been sitting on these updates for a few days. It feels like every week we get a fresh wave of "revolutionary" releases, and frankly, it can be exhausting to keep up.
I’ve spent the last few days looking at the latest announcements from Google, Anthropic, and some of the bigger enterprise players. The thing nobody mentions about these weekly cycles is that the version numbers matter far less than how these tools actually change our daily work.
If you’re feeling a bit overwhelmed by the pace, you’re not alone. I’m still trying to figure out which of these tools actually help me solve problems for clients and which ones are just shiny new toys.
Gemini 3.1 Pro and the Rise of Agentic Workflows
Google just released Gemini 3.1 Pro in preview. On the surface, it’s another model update, but the language they’re using is very specific. They are focusing heavily on "agentic workflows."
In my experience, most businesses use AI as a better search engine or a high-speed copywriter. We’ve been stuck in a "prompt and response" loop. What Google is pushing with 3.1 Pro is the idea that the model shouldn't just talk to you; it should do things for you.
Think about a complex task like reconciling an entire month of messy e-commerce data. Usually, you have to prompt the AI for every step. An agentic model aims to take the goal, figure out the five steps required, and execute them across different platforms. It's a shift from "tell me how" to "go do it."
For agencies, this is a massive opportunity to move away from hourly tasks. If a model can handle the heavy lifting of multi-step reasoning, your value shifts toward being the person who knows which problems are actually worth solving. You can find more articles on AI that touch on this shift in my recent archives.
Claude Sonnet 4.6: Better at Using Computers
Anthropic also dropped Claude Sonnet 4.6 this week. Again, the focus isn't just on "it writes better poetry." They are doubling down on computer use and coding.
I've found Claude to be consistently more "human" in its reasoning than its competitors, but this latest version is targeted directly at design and knowledge work. It’s becoming much more capable at navigating a screen and executing designs.
The interesting part here is how this affects the "free" and "pro" tiers. By making 4.6 the default, Anthropic is trying to lower the barrier for people to build actual software or tools using nothing but natural language.
I’ve had many conversations with business owners who are scared of the technical debt of building custom tech. Tools like Sonnet 4.6 make me think that we’re getting closer to a world where "non-technical" founders can build quite sophisticated internal systems without needing a massive dev team.
The Reality of AI Security and Compliance
As these models get more powerful, the risks get messier. The US Treasury recently released resources to guide AI use in financial services, and Security Compass launched a specific solution for managing agentic AI security.
It’s easy to dismiss this as "boring corporate stuff," but I think we’ve got this whole AI thing backwards if we ignore the safety side. If you’re building agents that can actually do things, you’re essentially giving a piece of software the keys to your house.
I often see startups rush to integrate the latest API without thinking about where that data is going or who is responsible when an "agent" makes a bad decision. These new frameworks from the Treasury and private security firms are a sign that the honeymoon phase of "just try it and see" is ending. We’re moving into a phase of professional, responsible deployment.
Mandatory AI Use and the Culture Shift
One of the more controversial bits of news this week was reports that Accenture is linking staff promotions to the use of their internal AI tools.
This is a move that makes me a bit uneasy. On one hand, you want your team to be efficient. On the other hand, forcing the use of a specific tool can lead to "performative AI usage"—where people use the tool just to check a box, rather than using it to actually do better work.
In my experience, the best AI adoption happens when the team sees a clear benefit to their own day-to-day life. When you force it, you often lose the nuance and the critical thinking that makes a human professional valuable in the first place.
If you’re leading a team and wondering how to encourage this without being heavy-handed, it might be worth a chat. I’ve seen what works (and what definitely doesn't) when rolling out these tools. Feel free to book a consultation if you want to swap notes on team adoption.
Practical Takeaways for Your Week
It’s a lot to digest. If you’re looking for a few things to actually act on, here is what I’m paying attention to:
- Test the "Agentic" Capabilities: Instead of asking Gemini or Claude to write a draft, ask it to "Plan and execute a research project on [Topic] and format the results into a table with sources." See how well it handles the multi-step logic.
- Review Your Permissions: If you are using tools that have "computer use" features, be careful about what windows or data they have access to.
- Look at E-commerce Integration: With Google and Sea (Shopee) teaming up, expect AI-driven shopping and seller tools to get much more intuitive very quickly.
- Focus on Reasoning Over Speed: The new models from China (like Doubao 2.0) and the US are all competing on "complex reasoning." Stop using AI for the easy stuff and start testing it on the things that actually make your head hurt.
Since it's Monday, it's also worth checking out how automation is hitting the sales side of things. I've been keeping an eye on SalesM8 for how they handle the actual execution of these types of workflows in the real world.
The bit most people miss is that AI isn't a single "event" that happens to your business. It's a series of small, iterative changes in how you think about work. This week’s releases are just more evidence that the "thinking" part of AI is getting a lot better, very fast.
Let me know what you're testing this week. I'm always curious to hear which of these updates actually makes a difference when the rubber hits the road.
I cover the latest AI releases and tools every Friday at steventann.com. If you found this useful, there's plenty more where this came from.
About the Author
Steven Tann is an AI consultant, author of "You're Selling AI Wrong", and founder of SalesM8. He writes about AI, sales, and running a business from a narrowboat on the English canals. Connect with him at steventann.com.
