OpenAI’s $20k AI & Meta’s ad takeover

It’s Thursday, and you know what that means—a fresh edition of Insane AI is here.

1: OpenAI plots charging $20,000 a month for PhD-level agents

OpenAI isn’t just betting on ChatGPT’s chat-subscription success—it’s looking to rake in big bucks from “agents” that can handle complex tasks on users’ behalf. According to investors, OpenAI plans to charge:

• $2,000/month for “low-end” knowledge-worker agents (think advanced personal assistants for busy professionals).

• $10,000/month for mid-tier software development agents.

• $20,000/month for “PhD-level” research agents—ideal for tackling high-stakes scientific problems.

OpenAI projects that these agent products will eventually drive 20%–25% of its total revenue. Though it already sells ChatGPT Pro at $200/month, that offering doesn’t have the same action-taking “agent” capabilities. Meanwhile, competitor tools are testing everything from pay-per-task fees to bundling AI features inside existing software suites. But given OpenAI’s leading role in setting AI prices, many will watch how CEO Sam Altman prices these specialized “PhD-level” agents—some businesses might pay top dollar for an AI that replaces a high-salaried engineer or accelerates critical medical research.

Seamlessly integrate your tools and services on HubSpot

The HubSpot Developer Platform is a trusted way for thousands of organizations to connect new and existing software. With flexible UI and extensibility tools, you can build apps that help teams unify their tech stack across our platforms. From listing your app in the HubSpot Marketplace to building custom solutions, there’s something for everyone to grow better.

Learn More

2: How Meta and Google benefit from AI-created ads

AI is taking ad creation to new levels, and tech giants like Meta and Google are cashing in. By using tools like HeyGen, Midjourney, and OpenAI’s Dall‑E, advertisers can cheaply spin out hundreds of ad variations—everything from different languages and visuals to new audiences and angles. That means they can pinpoint the message that resonates with each slice of their consumer base, increasing downloads, signups, and sales.

One example: Sociaaal, a studio behind celebrity look-alike and dating apps, raised its monthly ad output from 40 to 400. The result? A 65% revenue jump in four months. With that extra income, they poured more money back into ads—driving even more growth. It’s a virtuous cycle that benefits both advertisers and the platforms hosting the ads.

Google’s gone so far as to reimburse advertisers willing to try out third-party AI tools; the company just wants them spending more on Google Ads, period. Meanwhile, Meta’s sophisticated targeting systems help tailor all those AI-generated ad variations so they land in front of the right eyes. The takeaway? As AI churns out mountains of “hyper-personalized” campaigns, the winners are the advertisers who see better returns—and, of course, the platforms that sell them the ad space.

3: Make way for the humatons

Key idea: As AI tools get better, companies are blending human workers with automation—creating “humatons”—to handle client-facing tasks more efficiently. Humans remain essential for delivering “real” connections, even if much of the communication is AI-assisted behind the scenes.

What is a humaton?

A worker whose digital interactions—emails, chats, videos—are partly or fully generated and managed by AI. They still provide a human “face,” but the content and timing of their messages might be algorithmically optimized.

Why not go full AI?

• Customer trust: People usually prefer human interaction.

• Legal concerns: Some states ban misleading bots for sales.

• AI mistakes: Even advanced AI can hallucinate or glitch, creating liability risks.

Practical dilemmas and principles

• Establish humanness: Show proof there’s a real person behind the screen.

• Transparency: Tell customers when AI is doing the talking.

• Human touch: AI is great at spotting trends, but real humans bring empathy, authenticity, and personality.

Outlook: “Humatons” may be the best way for people to stay employed in an AI-driven future, especially in jobs involving trust, empathy, or deeply personalized service. The sooner we recognize and protect this human edge—requiring authenticity checks, for instance—the more we can preserve truly human work in an AI-saturated world.

4: “Humans in the loop” make AI work, for now

In today’s AI-driven world, companies promise there will always be people guiding and checking AI systems—just in case the bots make mistakes or go off the rails. Whether it’s prompting a chatbot, manually entering payment details in a conversation flow, or double-checking a final result, humans often step in at crucial points.

Why it matters: AI can do impressive work, but it’s not flawless. By having “humans in the loop,” organizations aim to catch errors, handle sensitive data, and maintain user trust. It’s also a way to keep AI from appearing impersonal or risky—nobody wants a bot making all the calls unilaterally, especially for tasks like launching missiles or offering critical healthcare advice.

The big picture

1. AI assists humans: Tools like Slack’s AI and ChatGPT still need people’s creativity and context.

2. AI hands over the wheel at key moments: Even advanced agents often require human sign-off on sensitive steps (like handling money or logins).

3. Humans review AI’s final work: Someone still has to fact-check a model’s “hallucinations” or missteps—especially in high-stakes scenarios.

As AI gets smarter, figuring out when humans should step in (and how often) becomes tricky. Workers skilled at deciding what to trust AI with—and what to reserve for actual people—are in higher demand.

For now, “humans in the loop” is an industry promise that keeps errors in check and makes AI systems feel safer. But as AI improves and companies chase greater automation, there’s no guarantee that humans will always be part of the final decision.

Editors picks ☕️

The Information: Trump calls for end of $52 billion chip subsidy program.

Financial Times: Companies are failing to convince staff of AI benefits.

Forbes: How to make yourself AI-literate.