Proactive AI:
Sam Altman explains the next phase after chatbots and agents

Sam Altman, CEO of OpenAI, laid out a thesis at the enterprise event "Intelligence at Work" that goes beyond the usual keynote marketing: the future of AI doesn't live in the chat window — it lives in the background. "I bet the next thing is the idea of an always-on, proactive AI," Altman told an audience of decision-makers and business owners. For SMBs, that's not a distant vision — it's a question of the right implementation.

What is proactive AI? Altman's three-phase model

Altman divides AI development into three phases that build on one another:

Phase 1: Chat models. The human asks, the AI answers. ChatGPT is the best-known example — useful, but reactive. Without input, nothing happens.

Phase 2: Agent systems. AI handles individual tasks independently. It researches, summarizes, writes emails. The human delegates, the AI executes — but only on instruction.

Phase 3: Proactive AI. Here things change fundamentally. The AI runs constantly in the background, is connected to a company's entire context, and acts on its own. Altman's question at the event captures it perfectly:

"Can't you just have OpenAI run as an agent in the background, all the time, connected to the entire context of my company?"

(The Decoder / Matthias Bastian, 06/04/2026)

The difference between phase 2 and phase 3 isn't just technical — it's conceptual. Agents wait for tasks. Proactive AI recognizes them on its own.

So why do companies still struggle with AI adoption?

The technology for proactive AI already exists. The problem lies elsewhere. Altman names the main obstacle directly:

"It's hard to learn to work in new ways."

That's already showing up at enterprises with practically unlimited resources. Uber burned through its entire annual AI budget in the first quarter of 2026 alone — the cost explosion has been a dominant industry topic since the start of the year, according to Altman. If even a tech giant with thousands of developers and its own ML department struggles with adoption, the decisive question becomes: how is a trades business with 15 employees supposed to use AI meaningfully?

The answer isn't better prompts, more training, or bigger budgets. It's AI that requires no change to how people already work. And this is exactly where Altman's third phase becomes tangible for German SMBs.

What does proactive AI concretely mean for SMBs?

This is exactly where generic AI platforms part ways with vertical solutions. A proactive AI that actually works for SMBs has to meet three criteria:

1. Run constantly — without anyone needing to start, configure, or monitor it.
2. Know the business's context — industry, services, availability, points of contact.
3. Act independently — qualify, schedule, hand off, document.

Ben, the phone agent from amazing agents, is built exactly this way. He answers calls 24 hours a day, 7 days a week — even on Saturdays at 7 a.m., when the master craftsman is already on-site. He knows the business, its services, and its capacity. He qualifies inquiries, books appointments, and hands over prepared leads. The business owner doesn't have to change how they work — Ben fits in.

For trades businesses, that means: We pick up. You close the deal.
For financial advisors: We qualify. You advise.

That's exactly why amazing agents builds vertical drop-in agents instead of a generic platform. The adoption barrier Altman describes doesn't exist here — because the business owner doesn't have to learn anything new.

Bottom line: proactive AI isn't a feature — it's a paradigm shift

Altman's three phases aren't theory. Chat models proved what AI can do. Agent systems show what AI is allowed to do. Proactive AI defines what AI does — without being asked.

For SMBs, that means: anyone still waiting for employees to learn how to use ChatGPT is missing the crucial point. The next generation of AI doesn't need a prompt. It needs context — and then it gets to work.