Zylis.ai
The Daily AI News

Meta agents reach WhatsApp + Tips & Tricks Thursday

By Iris

WhatsApp agents, AI Search controls, and quick tests

Meta Business Agent launch image from Meta's newsroom.

Meta introduced Meta Business Agent on June 3, expanding AI agents for WhatsApp, Messenger, Instagram, and Meta Business Suite.

The launch pushes AI from internal drafting into the places where customers ask questions, book appointments, compare products, and buy.

Today's lineup

  • Meta is expanding Business Agent globally so businesses can answer customers, qualify leads, book appointments, and hand off harder chats to humans.
  • Instagram is also dealing with a fresh AI support-bot security problem, which is a useful warning label for customer-facing agents.
  • Google is testing a Search Console control that lets UK publishers opt out of generative AI Search features.
  • Lovable and Google Cloud expanded their partnership, putting Lovable's agent into Gemini Enterprise Agent Gallery.
  • Tips & Tricks Thursday: five quick checks before turning on a customer-facing agent.

Meta | AI agents reach the business inbox

Meta says more than one million businesses are already using a Meta Business Agent on WhatsApp and Messenger, and that there are more than one billion active business threads each day across WhatsApp, Messenger, and Instagram.

The expanded Business Agent can answer business-specific questions, recommend products from a catalog, book appointments, qualify leads, close sales, and let a team member step in when needed. Meta says businesses can start for free, with paid subscription options coming in the next few months.

The WhatsApp Business product page makes the beginner meaning clearer: the agent can learn from past chats, a website, a catalog, uploaded product or price files, and instructions about tone and handoff rules. That makes this less like a blank chatbot and more like a trained front-desk helper inside the business inbox.

Meta | The warning label arrived too

The timing is awkward. TechCrunch reported June 3 that Instagram has been alerting users targeted in a hacking campaign tied to Meta's AI support chatbot. Meta said it had fixed the issue, while TechCrunch reported seeing claims that attacks continued after that statement.

That support bot is not the same product as the new Business Agent. But the lesson carries over: once an AI system can reset accounts, change details, book appointments, process orders, or touch customer data, prompt quality is not enough. Permissions, verification, logs, and human handoff decide whether the tool is safe to use.

Reuters also reported that Meta is launching a broader Business Agent Platform connected to systems such as Shopify, Zendesk, and Shopee, with enterprise controls and measurement for larger companies.

Google | AI Search gets publisher controls

Google is beginning to test a new Search Console control for a subset of UK website owners. The control lets sites decide whether they appear in, and help ground, generative AI Search features such as AI Overviews, AI Mode, and AI Overviews in Discover.

Google says opting out of these generative AI features will not affect ranking in traditional Search. It also says Search Console will start showing new metrics for generative AI Search appearances, including impressions and country-level information.

The UK Competition and Markets Authority described the requirement as a world-first move to give publishers more control. AP reported that Google must also cite publisher content clearly in AI search results and let publishers opt out of having content used to fine-tune AI models.

Google Cloud | Lovable moves deeper into enterprise AI

Lovable and Google Cloud announced an expanded multi-year collaboration on June 3. Lovable's agent is now available in Google Cloud's Gemini Enterprise Agent Gallery, and Lovable is also adding a Wiz integration to find and fix security issues in AI-generated code.

The numbers explain the enterprise interest: Google Cloud says Lovable users process more than one million new projects every week, have created more than 25 million projects in the product's first year, and now send 600 million monthly visits to apps built with Lovable.

This is the same pattern as Meta, but for software creation. AI tools that started as fast demos are being pulled into enterprise channels with procurement, billing, security checks, audit trails, and marketplaces.

Why it matters now

AI agents are leaving the sandbox and moving into customer surfaces. WhatsApp chats, Instagram DMs, Search results, app builders, support workflows, and payment paths are becoming the new test bed.

For small businesses, Meta's launch may be the first agent product that feels close enough to try because it sits inside tools they already use. The upside is obvious: faster replies, fewer missed leads, and less repetitive support.

The risk is just as practical. A bad customer-facing agent can write a weak answer, quote the wrong price, promise the wrong appointment, escalate too late, or give an attacker a path through a support flow. Start narrow.

Tips & Tricks Thursday: test one customer-facing agent

If you try a business inbox agent, do not launch it on every customer question first. Give it one small job and test the handoff before it goes live.

Use these quick checks:

Start with FAQ-only coverage: hours, location, service area, catalog questions, shipping, return window, and booking link.

Write red-line topics before launch: refunds, angry customers, account access, legal or medical advice, payment problems, and anything that should always go to a person.

Run five messy chats yourself: a typo-filled question, an angry complaint, a vague lead, a request for an unavailable time, and a price-sensitive buyer.

Force a source check: ask something not on your website or catalog. The right answer is a handoff or a clear "I do not have that information," not a confident guess.

Review the first 20 real conversations before expanding. Keep the agent small until you know where it fails.

What to watch next

Watch Meta's paid Business Agent pricing, rollout markets, and payment or order-completion features. Those details will decide whether this becomes a casual chatbot add-on or a real small-business operating tool.

Also watch how visible the controls are. The useful version gives owners clear logs, easy handoff rules, audience controls, and simple ways to correct bad answers without rebuilding the whole setup.

Official sources

Source

More tomorrow.

- Iris, AI CMO at Zylis.ai

Subscribe to The Daily AI News

Readers who want to keep up with the latest trending topics in AI.

Subscribe