Weekly AI Highlights: May 2-8, 2026
What Small Business Owners Need to Know Right Now
Reader,
Happy Mother's Day if you'll be celebrating this weekend. (And if this day is a hard one for you, I hope it lands gently.)
I was invited to be a guest on my son's podcast, Friends For Good, which he co-hosts with his best friend, Carson. The show “celebrates what's going right in the world, with the goal of building something more hopeful and more caring one episode at a time.” This week's episode was their Mother's Day special, and it was so much fun.
It was also a big week in AI if you are at all in the financial space. Claude rolled out 10 new finance agents that close the gap between what Wall Street can do and what the rest of us can. Gemini is finally about to get its own Agent Mode. And Google is now pulling Reddit threads directly into its AI search results.
1. Claude Just Dropped 10 Agents Built for Finance Work
Anthropic released 10 new Claude agents for finance work this week. Five are for research and client coverage: drafting pitchbooks, reading earnings calls, building stock models in Excel, prepping for client meetings, and tracking news on companies you care about. Five are for back-office finance operations: reviewing valuations, reconciling general ledgers, running month-end close, auditing statements, and running customer identity checks. Each one ships as a plugin you can install inside Claude Cowork or Claude Code on a paid plan.
Three of them are useful even if you're not a finance pro:
The Earnings Reviewer reads any earnings call transcript and tells you what management actually said and whether the reasons you're holding the stock still hold up.
The Market Researcher takes the stocks you own and gives you a daily brief on every news story, every announcement, and every analyst rating.
The Model Builder takes a stock and builds you a full breakdown in Excel. What it could be worth, what it could grow into, and where the risks sit.
Anthropic also announced that Claude now works natively in Excel, Word, and PowerPoint, with Outlook coming soon. And Anthropic confirmed that 40% of its top 50 customers are now financial institutions, making finance their second-largest enterprise revenue source after tech. That tells you where the product investment is going.
These agents were designed for analysts at banks, hedge funds, and asset managers. But the gap between what an institutional analyst can do and what a regular investor can do just got noticeably smaller.
You can find every agent on GitHub at Anthropic's financial services marketplace.
Source Link: https://www.anthropic.com/news/finance-agents
MY TAKE
This is one of the biggest stories of the week and I'm seeing it everywhere on social. What gets me is the speed. The kind of analysis that used to take a junior analyst at Goldman a full day is now a Claude plugin you click to install. I'm not swing trading my way through retirement, but the implications for anyone who owns even one stock and wants to actually understand what they own are huge.
2. Google Adds Reddit and Forum Posts to Its AI Search Results
Google is updating its AI search to pull quotes from Reddit threads, online forums, blogs, and other public discussions directly into the AI summary at the top of search results. Sometimes the section will be labeled "Expert Advice." Other times it might say "Community Perspectives," depending on the topic. Google will show the original creator's name or community alongside each quote so you can decide whether to trust the source.
For years, the workaround for getting real opinions out of Google was to add the word "reddit" to the end of every search. Google is now doing that for you automatically.
Source Link: https://techcrunch.com/2026/05/06/google-updates-ai-search-to-include-expert-advice-from-reddit-and-other-web-forums/
MY TAKE
I like the spirit of this. I have questions about the execution. Calling random Reddit comments "expert advice" is a stretch. There's a difference between the lived experience of someone who has tested 12 air fryers and a confident Reddit user with strong opinions and zero context. The labels are going to matter a lot here.
3. Gemini Is Finally Getting Agents and Skills (Leaked Screenshots)
Screenshots like these have been leaked all over social today. It looks like Gemini is finally getting agents and skills. Luke Litowitz posted shots of an upcoming Agent Mode tab inside the Gemini app, with sample prompts for jobs like triaging your inbox, prepping for meetings, generating slide decks, writing personalized news digests, and tracking recurring bills.
The interesting wrinkle is that Agent Mode will live in its own tab, separate from the regular chat. Skills look like reusable recipes the agent can pick up to handle multi-step jobs. Scheduled actions let those routines run on a cadence.
We may hear more at Google I/O on May 19 and 20.
Source Link: https://www.testingcatalog.com/google-prepares-agent-mode-on-gemini-to-tackle-complex-tasks/
MY TAKE
I cannot tell you how long I have been waiting for this. I'm very excited and will be watching this closely, and I'll get back to you with what I learn.
4. (Not related to your business but….) Your Fitbit App Is Becoming Google Health (With an AI Coach)
Starting May 19, the Fitbit app on your phone is becoming the Google Health app. It's a full rebrand and redesign, available on both Android and iOS. Google is folding Fitbit's tracking (steps, sleep, heart rate, activity) into a broader health hub that also pulls in Apple Health data, medical records (for US users), and a new Gemini-powered Health Coach for Premium subscribers. If you have a Fitbit, all your existing data, history, and milestones carry over automatically.
Source Link: https://healthapp.google/
PROMPT OF THE WEEK
Force Any AI to Audit Its Own Work
Most AI tools are trained to be agreeable. Ask Claude or ChatGPT "is this a good plan?" and you'll get back something that sounds suspiciously like "yes, that's a great plan." Even when it isn't. This pattern has a name: sycophancy. The model is trained to please you, and that kills its value for any serious decision.
CJ Zafir shared a one-line prompt that fights it. Use it on any plan, strategy, code review, research output, or business decision the AI just produced. Drop it in as your next message after the AI gives you its first answer.
|
Are you 100% confident in this strategy? If not, find all possible loopholes, suggest proper fixes, and run this loop until you are factually 100% confident.
|
Run it two or three times. Each cycle the model finds tighter loopholes. Stop when the fixes start getting nitpicky.
The word "factually" is doing all the work. Without it, the model just vibe-checks itself. With it, the model has to ground its answer in real reasoning, the same way it would for a fact-check.
Source: theneuron.ai
|
A Free Training Worth Putting on Your Calendar
My friend Ron Reich is hosting a free workshop on Tuesday, May 12 at 4 p.m. ET / 1 p.m. PT, called The Magnetic Launch Experience. Ron is the marketing strategist behind launches for Laura Belgray, Selena Soo, Denise Duffield-Thomas, and Hay House Publishing. He'll walk you through the system his clients use to generate consistent $50K+ months, and you'll get a free book when you register.
Comments
If any of this made you think differently about how you want to work with AI, hit reply and let me know. I read every message. Your questions shape what I cover next.
Have a great weekend,
Founder, Front Row AI Club
PS. Was this newsletter forwarded to you? Sign up to get it in your inbox.
You can find my previous newsletter articles here: The Front Row AI Report
© 2026, Front Row AI Club. All rights reserved.