Hi {{first_name}}!
This marks my 11th weekly newsletter and I’m getting some great feedback so thank you all! I truly want to hear from YOU so please reply with feedback if you have a couple minutes. I read every email and take every idea into consideration when I write these!
This week is packed with practical updates you can use right away. Google is pushing hard with new reasoning capabilities, and OpenAI is ramping up as it faces increased competition.
I’m also sharing a specific approach we use to cut our automation infrastructure costs. Plus, I want to walk you through a framework that helps organizations move past the pilot stage.
If you have been feeling stuck testing individual tools rather than thinking about the big picture, this one is for you.
Here is what we are covering:
New AI developments from Google, OpenAI, Meta, and Anthropic
How to cut automation costs by 75% by self-hosting N8N
The four-phase framework for scaling AI across your company
OK, let's get into it!
New and Noteworthy
OpenAI Releases GPT-5.2 Model Series: OpenAI just rolled out GPT-5.2 to everyone yesterday with three distinct models: Instant for everyday work, Thinking for professional tasks with major spreadsheet and presentation improvements, and Pro for complex programming and scientific work. All three feature a massive 400,000-token context window (roughly 300,000 words) so you can process entire codebases or document sets in one conversation. The models are designed to unlock more economic value for businesses by handling complex multi-step projects better than any previous model. Available now in ChatGPT for all users and in the OpenAI API for developers.
Disney Bets $1 Billion on OpenAI: Disney became OpenAI's first major Sora partner with a $1 billion investment, using Sora to let fans generate videos with 200+ Disney characters for Disney+. They're also deploying ChatGPT company-wide and building new Disney+ products using OpenAI's APIs. This signals Fortune 500 companies are moving from AI pilots to full-scale integration.
Google's Gemini 3 Deep Think Mode: Google launched Deep Think for Ultra subscribers (currently at the $200/mo level). This is a new reasoning mode that spends more time working through complex problems using parallel reasoning. If your business deals with complicated data analysis, strategic planning, or technical challenges, and you’re a power user of AI I’d consider testing this to see if the upgrade produces an ROI.
Claude Comes to Slack: Anthropic rolled out a beta integration letting developers delegate coding tasks to Claude Code directly inside Slack. This turns everyday team conversations into automated development workflows right where you're already working.
Google and Replit Partner on Coding Tools: Google and Replit are teaming up to build new coding tools competing with Claude Code and Cursor. The AI coding space is heating up fast, which means better tools and more options for businesses building internal systems and apps.
NotebookLM Expands Mobile Features: Google added visual uploads, slide support, and synced audio progress to NotebookLM's mobile app. Now you can capture research on the go and pick up your audio overviews right where you left off.
Meta Signs AI Training Deals: Meta locked in licensing agreements with CNN, Fox News, and USA Today to train its AI models on news content. This signals AI companies are moving toward legitimate content partnerships, which should improve accuracy across AI tools.
OpenAI's "Code Red": Sam Altman paused side projects for eight weeks to focus entirely on improving ChatGPT, with a new model dropping this week and another in January featuring better images and more natural personality. The AI model wars are accelerating, and businesses using these tools will benefit from rapid improvements.
Tool Spotlight: Self-Hosting N8N
Most automation platforms charge based on executions. Zapier bills per step in a workflow, Make.com and N8N.com charge on execution volume. This is a great place to start if you’re just beginning to explore automation within your business, but once you build a few automations, you’ll realize these costs can start to add up fast!
The beauty of N8N (besides the fact that it’s an amazingly powerful automation building tool) is that it’s open source and can be self-hosted. By self-hosting you only pay a small flat monthly hosting fee for unlimited workflows with no execution caps.
We’ve tested a few hosting options and Hostinger has become our go-to solution. They also happen to be running a cyberweek promo right now so for just $6.49/month you can sign up for their “KVM2” plan (our recommendation) and run unlimited workflows. If you plan on building agentic workflows as part of your growth strategy in 2026 I’d jump on this deal.
They offer one-click N8N installation, automatic daily backups (add $6/month), and you can scale server resources as your needs grow. We typically end up setting this up for clients as part of our FlexFlow offering and it only takes about 15-20 minutes.
We recently helped a client build some agentic workflows, and while we were at it we moved a bunch of active automations they had from a “cloud” account to their new self-hosted account saving them hundreds of dollars a month.
In short, if you're just starting with 2-3 simple workflows, cloud platforms work fine and they are a little bit easier to get up and going. But once automations are helping to run core business processes and you're hitting execution limits or spending $100+ monthly, owning your infrastructure pays for itself quickly.
If you have questions about this or need some guidance I’m happy to help!
OpenAI's Framework for Getting Your Entire Organization Using AI
OpenAI just released a practical framework for organizations ready to move beyond the pilot phase. The timing is interesting because McKinsey found that 62% of companies are still experimenting or running small pilots with AI.
If that's you, there's a real opportunity to get ahead while others are still figuring out where to start.
Most businesses treat it like software from the past. Test a tool. Run a pilot. Wait for results. Then scale. But AI moves faster than that. Capabilities evolve every few weeks. Success depends less on which tool you pick and more on the systems you build around it.
OpenAI's framework breaks it into four phases:
Set Foundations - Get executive alignment, create governance that evolves with the technology, and open up data access. Organizations with formal AI governance programs score significantly higher on readiness than those winging it.
Create AI Fluency - Build internal champion networks. Your best AI teachers aren't just external consultants. They're the people in your organization who've already figured out how to use these tools in their actual daily work.
Scope and Prioritize - Create simple ways for ideas to surface from anywhere in your company. Some of the best use cases come from people closest to the work, regardless of title or department.
Build and Scale Products - Move from individual productivity wins to actual revenue generation. This is where ROI compounds.
Companies thinking systematically are pulling ahead of those stuck on tool-by-tool decisions. One pattern we see constantly is teams being too busy to learn the thing that would actually save them time. That's why mandates without allocated learning time don't work.
If you're ready to move your organization from scattered experiments to systematic adoption, happy to walk through what that could look like for your team.
BTW, the infographic pictured above is something I created by uploading the 25-page OpenAI PDF report, putting it in NotebookLM, and using the new infographic creation tool! It took about 2 minutes! 😂
That is it for this week.
If there is a specific challenge you are facing that you want me to cover next week, I’m all ears.
Julien
PS: It would mean a lot to me if you’d forward or share this newsletter with any that might find value! They can subscribe at www.ampra.ai/join-our-newsletter.