Hi {{first_name}}!
Hope your week is going well! I've been thinking a lot about automation lately, trying to help people discern between when they might need an AI agent versus an automation workflow. Everyone's rushing to deploy agents because they're the hot new thing, but sometimes the “boring” reliable automation is exactly what your business needs most.
This week I'm sharing that perspective along with some important AI updates worth knowing about.
This week we're covering:
New AI developments from Google, OpenAI, and others
When workflows beat agents - and how to know which one you actually need
OK, let's get into it!
New and Noteworthy
Personal Intelligence from Gemini: Google just launched something genuinely useful: Gemini can now connect across your Google ecosystem (think Gmail, Photos, and YouTube) to answer personalized questions. Gemini will reason across all your personal Google data to give you contextual answers that actually connect the pieces. You can check out the linked article for some really cool examples and use cases. Right now it's only available for personal Google accounts (not Workspace business users yet), but once this rolls out to business accounts, it'll be incredibly powerful. Worth keeping on your radar if you're deep in the Google ecosystem.
OpenAI's AI Wearable: Leaked details suggest Sam Altman and Jony Ive's first hardware device is actually going to be an audio-centric wearable worn behind the ears, codenamed "Sweetpea," targeting a September 2026 release. The other device that is rumored to be in development is a digital pen.
HeyGen for Business: HeyGen launched an enterprise offering for AI video avatars that lets businesses create professional video content at scale. Useful for training videos, product demos, or personalized sales outreach without coordinating schedules for actual filming.
Google's Agent Commerce Protocol: Google announced a new protocol to facilitate commerce using AI agents, built alongside Shopify, Walmart, Target, and Etsy. This could fundamentally change how AI-powered customer experiences work in e-commerce.
Veo 3.1 Upgrades: Google announced improvements to its video generation model, bringing more consistency, creativity, and control. Paired with platforms like Canva that integrate Veo, this makes professional video content creation more accessible to businesses without dedicated video teams. Personally I find it great for B-roll and for turning images into videos!
TranslateGemma: Google released a new open-source family of translation models designed for efficient multilingual text translation. Particularly useful for businesses working with international clients or teams who need reliable translation without the cost of premium services.
ChatGPT Memory Improvements: ChatGPT's memory feature received significant upgrades, making it better at remembering your preferences, work context, and ongoing projects across conversations.
ChatGPT Ads Coming: ChatGPT is rolling out ads, starting with US users. This only affects free and Go ($8/month) plan customers. Ads will appear at the bottom of answers, they will not affect the actual responses, and no conversation or personal data is shared with advertisers. Business and Enterprise plans remain ad-free. As much as this negatively impacts the user experience, this also represents an opportunity for advertisers to jump into a new ad platform.
Meta's Infrastructure Push: Mark Zuckerberg announced Meta is launching its own AI infrastructure initiative, signaling the company's commitment to competing in the enterprise AI space beyond just consumer products.
Claude Cowork Is Evolving Fast: Last week I mentioned Claude Cowork launched for Max subscribers. This week we learned Anthropic built the entire thing in just over a week, mostly using Claude Code to build itself. It's designed specifically for non-technical users to handle everyday work: organizing files, creating documents, turning receipts into spreadsheets. Anthropic has been transparent about security risks around prompt injection attacks, so if you're testing it, start with non-sensitive work and limit access to trusted sources while they continue developing safeguards.
Workflows vs Agents: Which One Does Your Business Actually Need?
I listened to a podcast this week with Darren Patterson from Make.com and had a realization I wanted to share. Make just launched AI agents alongside their traditional automation tools, but they’re not telling customers to replace everything with agents, and that restraint matters.
Everyone’s talking about AI agents right now. It’s a common request we get. And to be fair, we’ve built plenty of them. Agents can be incredibly powerful when used in the right way.
What I see most often at Ampra, though, is this: teams come to us asking for agents when what they actually need first is clarity. Clear processes. Documented SOPs. Clarity on which processes even make sense to automate. Agreement on how things should work before anything gets automated or handed off to an AI agent.
Tools like n8n, Zapier, and Make all now include agent builders, but that doesn’t mean everything inside those tools should be an agent. In practice, most of the automations we build are still pure workflows, sometimes with small, intentional AI steps embedded where reasoning is actually needed. From my perspective, that’s how you get the best balance of reliability and flexibility.
Automation workflows are deterministic. You define the steps and get the same outcome every time. Agents are non-deterministic. You give them a goal and context and let them decide the path. Both are useful, but neither works well without documented systems underneath them. Ultimately, you can’t automate what you haven’t yet defined.
The real skill isn’t choosing workflows or agents. It’s knowing when predictability matters more than flexibility, when judgment actually adds value, where a human-in-the-loop is necessary, and how to combine the two without overcomplicating things.
I’m just finishing a deeper-dive white paper on this exact topic because I think it’s critical for businesses to understand how to think about automation and agents in 2026. It breaks down how workflows and agents differ, how to decide what your business actually needs, and includes real examples and case studies from teams doing this well.
If this is something you’re navigating right now, reply and I’ll send you the white paper.
If you spotted a quick win here, or if you're just wondering where to start with automation, feel free to reach out. I'm always happy to share insights and help you figure out the best first step.
See you next week,
Julien
PS: If this was valuable, please forward it to a friend or colleague. They can subscribe and catch up on previous editions at www.ampra.ai/join-our-newsletter.