I keep noticing a pattern with the agencies that seem happiest right now. They aren't the ones jumping on every single "AI agent" wrapper that pops up on Twitter. Instead, they are the ones quietly testing how new model capabilities can shave three hours off a boring Friday task.
It has been a particularly loud week in the world of artificial intelligence. We’ve seen major updates from the big three—OpenAI, Anthropic, and Google. If you feel like you're falling behind because you didn't spend your weekend reading technical whitepapers, don't worry. Most of it is just incremental polish.
However, there are a few shifts that I think are worth sitting with for a moment. Here is my take on what actually landed this week and what it means for those of us trying to run a business.
OpenAI o1 and the shift toward "reasoning"
The biggest talking point has been OpenAI's "o1" model (formerly known as Strawberry). For a few months, the rumour mill was spinning about a model that could "think" before it spoke. Now that it is here, the reality is a bit more grounded.
In my experience, the standard GPT-4o model is great for creative heavy lifting or quick summaries. But it can be impulsive. It wants to give you an answer immediately, which often leads to silly logic errors in complex tasks.
The o1 model is different because it uses "chain of thought" processing. It literally takes a few extra seconds to check its own work before showing you the result.
What this means for your business: If you are using AI for basic email drafting, o1 is overkill and too slow. But if you have complex logic problems—think complicated Excel formulas, legal contract analysis, or architectural planning—this is a genuine step up.
It's less of a "chatbot" and more of a "reasoning partner." I've found it particularly useful for troubleshooting code where I used to get stuck in a loop of errors.
Claude’s new "Analysis Tool" is a quiet game changer
While OpenAI grabbed the headlines, Anthropic released something for Claude that might actually be more useful for the average agency owner. They’ve introduced a built-in analysis tool that allows Claude to write and run JavaScript code to process data.
Previously, if you gave Claude a massive CSV file, it would try to "read" it like a person. This often led to hallucinations or missing rows. Now, Claude can build a custom script to crunch the numbers exactly like a data scientist would.
The practical implication: Think about your monthly lead reports or client performance data. You can now drop a raw export into Claude and ask, "Which of these leads had the highest conversion rate but the lowest cost per click?"
It doesn't just guess; it calculates. It’s essentially a bridge between a creative assistant and a data analyst. If you've been hesitant to trust AI with your numbers, this new approach should give you a bit more confidence. You can find more more articles on AI where I discuss how to integrate these tools into your workflows.
Google Gemini and the "NotebookLM" craze
If you’ve been on LinkedIn lately, you might have heard AI-generated podcasts that sound eerily human. This comes from Google’s NotebookLM.
You feed it a bunch of documents—PDFs, meeting notes, or blog posts—and it generates a "Deep Dive" audio conversation between two AI hosts. They banter, they laugh, and they summarise your content in a way that feels surprisingly natural.
Is it hype or helpful? Honestly, it’s a bit of both. The technology is incredible, but the use case for a business is still maturing.
- The Hype: Using it to turn your boring company handbook into a podcast. (Nobody is going to listen to it).
- The Helpful: Using it yourself to "listen" to a long, complex industry report while you’re at the gym or driving.
It is a brilliant tool for personal consumption and learning. As a marketing tool, I think people will eventually get tired of the "AI voice" tone, but for internal knowledge sharing, it’s a fantastic way to digest information quickly.
How to decide which tools to adopt
With something new coming out every Tuesday, it is easy to feel overwhelmed. I’ve made the mistake of trying to build a workflow around a tool that disappeared three months later.
Here is a simple framework I use to decide if a new AI feature is worth my time:
- The 10x Rule: Does this tool make a specific task 10 times faster, or is it just 10% better? If it's just a slight improvement, keep your current process.
- The Integration Check: Does it work where I already work? A tool that requires me to open a new tab and log in every time is a tool I will eventually stop using.
- The Data Privacy Factor: For agencies, this is huge. Is the client data I'm putting into this model being used for training? Always check the settings.
The bit most people miss is that the best AI strategy isn't about having the most tools. It's about having the most refined prompts and processes within the tools you already use. If you need help figuring out which part of your sales process to automate first, feel free to book a consultation.
The common thread: Moving from "Search" to "Action"
The theme of this week’s releases is clear: AI is moving away from just being a fancy version of Google Search. It is becoming an engine that performs tasks.
Whether it’s OpenAI o1 solving a math problem, Claude running a data script, or Gemini summarising an entire library of documents, the focus is on output.
I keep coming back to the idea that we shouldn't ask "What can AI do?" but rather "What am I doing today that I shouldn't be?" Usually, the answer to that question will point you exactly toward the tool you actually need.
If you’re looking to streamline your sales outreach specifically, you might want to check out SalesM8 to see how we’re putting some of these principles into practice for actual lead generation.
The pace isn't going to slow down, but your response to it can be measured. You don't need to be first; you just need to be effective.
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.
