Mark Zuckerberg Wants Meta's AI Agents to Work for Everyone, Not Just Coders
On Meta's Q2 2026 earnings call, Zuckerberg sketched out a vision of personal AI assistants that handle your health, finances and relationships. The product doesn't exist yet, but the infrastructure bill is already enormous.
Key points
- Meta CEO Mark Zuckerberg announced on the Q2 2026 earnings call that personal AI agents are the foundation of the company's next wave of products and revenue.
- More than 1 million businesses use Meta's AI agents on WhatsApp and Messenger every week as of Q2 2026.
- Meta expects to spend between $130 billion and $145 billion on capital infrastructure in 2026, including a 1-gigawatt data centre campus built with BlackRock.
- Instagram reached more than 2 billion daily active users, Zuckerberg confirmed on the same call.
- Meta moved roughly 7,000 staff onto AI projects and cut about 8,000 jobs in May 2026.
Mark Zuckerberg has a simple pitch: an AI assistant that your mum could actually use.
On Wednesday's Q2 2026 earnings call, the Meta CEO described a future where personal AI agents, programs that handle multi-step tasks around the clock, go far beyond writing code. Think managing your health appointments, nudging your savings, even tending your friendships. "Soon we will have agents that can work 24/7 on your behalf to help you achieve your goals and improve your life, your health, your relationships, your finances, whatever you want," Zuckerberg said.
Why does this matter to ordinary people?
Right now, AI agents are mostly used by developers comfortable spending hours fiddling with settings. Zuckerberg's explicit target is "billions of people" who need something that works straight out of the box. That ambition puts Meta up against Google's personal agent Gemini Spark, plus Anthropic and OpenAI, labs that have focused heavily on business and coding tools rather than everyday consumers.
Meta's got real obstacles, though. Its AI models still trail the leading labs on most benchmarks. Unlike Google or Microsoft, it has no built-in access to your emails or calendar, the data that makes a personal assistant genuinely useful. There's also the trust problem: Meta's history with user data means plenty of people will hesitate before handing it control of their daily lives. As we reported on 24 July, Meta's own chatbot only recently gained calendar-reading and daily briefing features, which shows how much ground the company still needs to cover before a personal agent is a real product.
What is Meta actually building right now?
Business-facing agents are live. Over a million companies use Meta's AI agents on WhatsApp and Messenger every week to handle customer queries, and Zuckerberg confirmed the rollout is expanding to Instagram.
On infrastructure, the numbers are staggering. Meta plans to spend $130 billion to $145 billion on capital expenditure, covering servers, chips (the specialised hardware that runs AI calculations), and buildings, in 2026 alone. It announced a 1-gigawatt data centre campus, roughly the power draw of a mid-sized city, being built with investment firm BlackRock.
The company also refreshed its Muse Spark AI model, releasing a version 1.1 update with stronger coding capabilities last quarter.
| Item | Figure | When |
|---|---|---|
| Businesses using Meta AI agents weekly | 1 million+ | Q2 2026 |
| Instagram daily active users | 2 billion+ | Q2 2026 |
| Capex forecast for 2026 | $130bn to $145bn | Full year 2026 |
| Staff moved to AI projects | ~7,000 | 2026 |
| Jobs cut | ~8,000 | May 2026 |
Zuckerberg promised more concrete details on personal agents "soon." The vision is bold. The product isn't in anyone's hands yet. For anyone watching this space, the number to watch isn't the capex forecast. It's whether Meta can close the trust gap before Google's Gemini Spark does it for them.
