• The Neuron
  • Posts
  • 😺 Microsoft thinks we all need to become an "AI Boss"

😺 Microsoft thinks we all need to become an "AI Boss"

PLUS: This Google court case is WILD...

Welcome, humans.

Required reading time! So, y'know how there's a backlash against ā€œjust putting AI everywhereā€ in products we use daily? YC Partner Pete Koomen just nailed why.

He says most current AI features are ā€œhorseless carriagesā€ā€”they jam AI into old UIs instead of rebuilding for an AI-first world. Using Gmail's clunky AI draft feature as exhibit A, Koomen shows how it takes MORE time than just writing the email yourself.

His key insight: YOU should control the System Prompt for any AI acting on your behalf. Most AI apps should be "agent builders," not pre-configured agents. Example:

We'll take it further: everyone's too lazy to rebuild their codebase for AI—but that's exactly what's needed. For AI to be actually useful, it needs to be able to actually do things for us. That’s why you’re seeing such an emphasis on ā€œComputer useā€ agents right now.

We’ll literally teach AI how to use computers like humans do before we’ll give humans the tools to create actually useful AI apps.

The piece is packed with interactive demos showing how much better user-controlled AI could be. Definitely worth the full read (and great discussion on Hacker News).

Here’s what you need to know about AI today:

  • Microsoft released a new Work Trend Index report.

  • OpenAI projected $174B in revenue by 2030.

  • Google's antitrust case revealed some wild facts.

  • The US government ordered AI in K-12 schools.

Microsoft says we're in ā€œThe Year of the Frontier Firmā€... but what does that ACTUALLY mean?

Is your company hiring AI agents as employees? According to Microsoft's new Work Trend Index report, that's exactly what leading companies are doing in 2025.

Microsoft analyzed data from 31K workers across 31 countries and found the emergence of a new kind of organization they're calling the ā€œFrontier Firmā€ā€”companies powered by intelligence on tap and hybrid teams of humans and AI agents.

In a one-on-one chat with Microsoft’s Chief Marketing Officer, AI at Work Jared Spataro, he told us that after studying AI-native firms for a year and a half, they've observed a significant shift from employees caring about AI to employers driving AI adoption.

The research highlights three major shifts happening right now:

  1. Intelligence is becoming a utility you can purchase. With 82% of leaders planning to use AI agents in the next 12-18 months, ā€œdigital laborā€ is becoming as essential as electricity. 

    1. The need is clear: 53% of leaders say productivity must increase, but 80% of workers lack the time or energy to deliver more.

  2. Human-agent teams are replacing traditional org charts. Nearly half of leaders say they're already using agents to fully automate workflows.

    1. This is creating what Microsoft calls a ā€œWork Chartā€ā€”where teams form around goals rather than functions.

  3. Everyone's becoming an ā€œagent boss.ā€ The most valuable skill now is managing AI agents, not just using them.

    1. Leaders are ahead of employees here–67% of leaders are familiar with agents versus just 40% of employees.

The report shares some impressive real-world results, too: 

  • Wells Fargo built an agent for 35K bankers, cutting query time from 10 minutes to 30 seconds.

  • Dow created agents to find hidden shipping losses, expecting to save millions.

  • Bayer's researchers each save 6 hours weekly using agents.

The secret to success? Spataro revealed that high-quality input is crucial: ā€œOne little-known secret is grounding agents on really high-quality input content.ā€ He advises creating really good briefs for projects and using the agents to scale out. 

This caused us to ask: What makes a good brief? It turns out, the same things that make a good prompt!

Get this: Spataro also shared his #1 piece of advice for anyone looking to add ā€œAgent Bossā€ to their resume. Spoiler alert: start getting your hands dirty.

FROM OUR PARTNERS

Get from zero to SOC 2 in 15 hours with Delve AI agents

Tech founders, ever felt locked out of the $4 trillion healthcare market? As former healthtech builders themselves, the team at Delve is tearing down the HIPAA barrier.

If you need SOC 2, you can bundle HIPAA and get the fully compliant at no cost—instantly unlocking healthcare's golden gates for your enterprise product. Valid through April!

Beyond the free HIPAA upgrade, Delve's platform gets you fully SOC 2 compliant in just days, not months. The proof is in the results:

  • Lovable became fully SOC 2 compliant in just days

  • 11x saved 143 hours on SOC 2, and unlocked $1.2M ARR

  • Bland achieved SOC 2 compliance and unlocked $500k ARR within 7 days

Their secret weapon? CUA (Computer Use Agents) for compliance—AI-powered assistants that automatically collect screenshots and evidence, doing the grunt work while you build your empire. Book a demo with this link to get cracking.

Prompt Tip of the Day

Everyone’s concerned about AI taking jobs, so here’s a prompt to use AI to help you GET a job. There’s also this one, an alternate version shared in the comments.

The poster CLAIMS they used this (and had TikTok followers reply that it helped them get a job), but no guarantees. Either way, it’s probably a good starting point to help you update your resume if you don’t have any other resources!

Treats To Try.

  1. Use gpt-image-1 in the OpenAI Playground to test out prompts and images for OpenAI’s imagen (short for image generator, better than Sam’s ā€œImagegen", pass it on!).

  2. Peek coaches you through daily financial decisions by analyzing your spending patterns—free during beta, subscriptions later.

  3. BuildPad guides you step-by-step through validating and building your product idea, acting as an AI co-founder to ensure market demand—free for 2 phases.

  4. Suna is an open-source agent that executes real-world tasks for you through conversation by navigating websites, managing files, and running system commands (demo).

  5. Manna is like Duolingo, but for Bible study (genius) with beginner-friendly lessons and AI-powered answers to your questions—free to try.

  6. UI-TARS from ByteDance helps you control your computer using natural language, allowing you to perform tasks like changing settings or browsing websites through simple text commands.

  7. CUA helps your AI agents control computers within secure virtual containers, running macOS or Linux environments at near-native speed on Apple Silicon.

  8. Alter lets you talk to your Mac to complete tasks across all your apps directly from the notch—free 14-day trial, then connect your own AI or purchase.

  9. Perplexity’s voice assistant lets you set reminders, craft messages, and book restaurant reservations with your voice on both iOS and Android phones.

Around the Horn.

  • Sam Altman said on X that OpenAI doubled the rate limits for o3 and o4-mini-high for Plus subscribers, and that the OpenAI imagen (pass it on!) is now available via API with special controls for moderation sensitivity, quality, speed, background, output format, and more for $5/1M text or $10/1M image input and $40/1M image output, so ~2-19 cents per image (docs here).

  • OpenAI forecasted $125B in revenue by 2029 and $174B by 2030, but get this: it expects the majority of this revenue to come from agents and new products, not ChatGPT alone.

  • The US government will now integrate AI into K-12 education, establish AI courses for students, and create a White House Task Force on AI Education.

  • There’s a ton of juicy data coming out of Google’s antitrust court case right now:

    • Like how Google’s contract w/ Motorola apparently blocked the company from making Perplexity the default AI assistant on its devices…

    • Or how Gemini’s chatbot has about 35M daily active users and 350 monthly active users last month (vs 160M daily and 600M monthly for GPT)…

    • …And how ChatGPT has scooped up some ā€œhomework and mathā€ search queries from Google, but not the profitable ones (like searches where someone wants to buy something)… yet.

  • We don’t cover chip stuff here, but SemiAnalysis does (and is arguably the best in the business at it), and it just released a new banger on ā€œAMD 2.0ā€

FROM OUR PARTNERS

Want conversations, not frustrations? Meet Flow. šŸ—£ļø

Speechmatics has built a Voice AI (available via API for devs) that understands you the first time—even with thick accents, background noise, or complex language.

With best-in-class accuracy across 55+ languages, Flow delivers conversations your users will actually enjoy.

Thursday Trivia

One is real, and one is AI. Which is which? (vote below!)

A.

B.

Here’s the results from last week’s poll:

Robots never had a chance

Here’s what you said:

  • S.B chose B: ā€œThe spilled coffee cup looks tacked on. Def AI.ā€

    • We also like J.K’s response: ā€œJust seems like the girls reaction was too normal for a cup to have just dropped on the floor.ā€

  • M.W. chose A: ā€œthe people in the backgrounds limbs seem offā€

    • Pretty much everyone else who chose A was sussed out by the blur, like M.L.B who said: ā€œThe motion blur effect just doesn't look like a camera did it, something is off.ā€

  • Eagle eye mode: C.G. chose B: ā€œhair and jacket both show evidence of backlighting in B - but she is standing entirely in shadow.ā€ This is us right now.

A Cat's Commentary.

That’s all for today, for more AI treats, check out our website.

NEW: Our podcast is making a comeback! Check it out on Spotify.

The best way to support us is by checking out our sponsors—today’s are Delve and Speechmatics.

What'd you think of today's email?

Login or Subscribe to participate in polls.