Meta's acquisition of Manus, reported at between US$2 billion and US$2.5 billion, puts an autonomous agent inside Ads Manager that can plan and run media buying and campaign analysis with little human supervision. For advertisers, that makes reporting fast and cheap (a full report in about 90 seconds for around US$40) and puts pressure on agencies that bill monthly retainers for manual reporting. The agent can still misread how advertising works, so the advantage goes to teams that feed it cross-platform data and keep a person checking its decisions.
Key takeaways
- Meta bought Manus as an execution engine for its Llama models, and by February 2026 had integrated it into Ads Manager for media buying and campaign analysis.
- Manus can turn raw campaign data into a full report in about 90 seconds for roughly US$40, which undercuts small agencies charging US$2,000–10,000 a month for manual reporting.
- Security and geopolitics slowed the rollout: Manus routed data through third-party models such as Claude and Qwen, and Chinese authorities opened an investigation into the deal.
- The agent can misread advertising mechanics, for example proposing to move a whole budget to Google Search because of what a human would have recognised as bot traffic.
- Agents are only as good as the data they see, so cross-channel data and human review decide who benefits.
What did Meta buy when it acquired Manus?
The digital advertising industry is undergoing a structural transformation. In late 2025 and early 2026, Meta Platforms altered the landscape by acquiring the artificial intelligence startup Manus for a sum between 2 billion and 2.5 billion US dollars. This move marks a shift away from standard generative models toward autonomous agents. Unlike reactive assistants, Manus can plan and execute complex workflows with almost no human supervision.
By February 2026, Meta integrated Manus into the Ads Manager ecosystem to handle media buying and campaign analysis. However, implementing such powerful technology within a massive social advertising framework has revealed significant technical and geopolitical friction.
Why did security and geopolitics slow the Manus rollout?
Meta recognised that having a language model like Llama was not enough. They needed an execution engine to operate across various software environments, and Manus filled that gap. At the time of writing (April 2026), Manus uses third-party APIs including Anthropic’s Claude and Alibaba’s Qwen. Routing sensitive data through competitors created vulnerabilities, leading to a major security crisis when early agents leaked corporate data.
To solve this, Meta is accelerating the release of Llama 4 to power Manus natively and ensure data sovereignty. Geopolitics also plays a role. Despite Manus attempting to establish a Singaporean base to attract US venture capital, Chinese authorities launched an investigation into the Meta deal. Because the core tech originated in China, officials barred the founders from leaving the country, slowing Meta's integration timeline.
How does Manus change the economics of ad reporting and agencies?
Manus has changed the cost of data analysis. The agent can process raw campaign data and generate a full report in 90 seconds for about 40 US dollars. Brand managers can now use natural language to ask complex questions and get immediate answers.
This speed threatens the traditional billing models of small agencies that charge monthly retainers between 2,000 and 10,000 US dollars for manual reporting. To stay relevant, these agencies must move away from simple data aggregation and focus on high level strategy and emotional creative work that machines cannot yet replicate.
What are the risks for small businesses and for campaign decisions?
While large brands benefit from AI productivity, small businesses in developing markets face a digital divide. Manus uses a credit system where tasks cost a specific amount. The system often fails under high global traffic but still deducts credits, creating a financial burden for smaller marketers.
Furthermore, Manus sometimes lacks a deep understanding of advertising mechanics. In one instance, the AI saw a spike in acquisition costs and suggested moving the entire budget to Google Search. A human expert would have recognised the spike as bot traffic and simply excluded the bad placement. Because the agent does not prioritise keeping spending within the Meta ecosystem, agencies are hiring offshore analysts to act as a human layer to verify AI decisions.
Why is Meta betting on AI agents after the metaverse?
The push into AI follows a massive change in direction for Mark Zuckerberg. Reality Labs saw a cumulative deficit of 83.6 billion US dollars by the end of 2025 as the metaverse struggled to compete with platforms like Roblox. Consequently, Meta cut metaverse budgets by 30 percent and plans to spend 135 billion US dollars in 2026 to become an AI first organisation.
As Llama 4 becomes the primary engine, computing costs should drop, making Manus a nearly free tool for the industry. This requires all players to adapt. Small brands must become AI editors, while large agencies will likely use their capital to build systems that connect these agents to broader data platforms, replacing junior buyers with AI orchestrators.
Why does the data behind an agent matter so much?
Autonomous systems are only as good as the data they receive. Early errors in Manus deployments often came from data myopia or being restricted to a single platform. The future of the industry depends on a strategic data layer that can see across different channels and bypass platform blockades.
Advanced frameworks now break down historical data into specific concepts like tone of voice and audience segments. This allows AI to turn vast amounts of information into optimised marketing strategies tailored for specific cultures. Whether a company uses Manus or a custom interface, success will depend on giving the agent a total view of the market rather than keeping it isolated within a walled garden.
Where does SOMIN fit in an agentic advertising stack?
As autonomous agents like Manus reshape execution, the competitive advantage will increasingly depend on the intelligence layer that feeds them. This is where SOMIN’s analytics SaaS ecosystem becomes critical. Through SODA and SOMONITOR including Brand Tracker, Content Library, Perspective Studies, and SoInspire combined with GWI audience intelligence, SOMIN provides a cross-platform, structured data foundation that mitigates single-platform bias and algorithmic blind spots. Instead of relying on isolated ad account signals, brands and agencies can equip AI systems with enriched competitive benchmarks, audience behaviour insights, and concept-level performance patterns.
In an agentic future, execution may be automated, but strategic oversight will belong to those who control clean, contextualised, and multi-source data, and that is precisely the layer SOMIN is built to deliver.
FAQ
How much did Meta pay for Manus?
Reports put the price at between US$2 billion and US$2.5 billion. Meta made the acquisition across late 2025 and early 2026 and, by February 2026, had integrated Manus into Ads Manager to handle media buying and campaign analysis.
Will Manus replace marketing agencies?
It threatens agencies whose main product is manual reporting, because Manus can produce a full report in about 90 seconds for around US$40. Agencies that move towards high-level strategy and emotional creative work, and that verify the agent's decisions, are in a stronger position.
What mistakes can Manus make in an ad account?
In one case it saw a spike in acquisition costs and suggested moving the entire budget to Google Search, where a human expert would have spotted bot traffic and excluded the bad placement. Its credit system can also fail under heavy global traffic and still deduct credits, which hurts smaller marketers.
Why did Manus rely on Claude and Qwen?
At the time of writing, Manus used third-party models including Anthropic's Claude and Alibaba's Qwen. Routing sensitive data through other companies' models created security risks, which is one reason Meta is accelerating Llama 4 to power Manus natively.
What should advertisers do now?
Treat the agent as something to edit, not an autopilot. Give it a view of the market that goes beyond one platform, and keep a person checking the decisions it makes.
Related reading
- What can Meta's Manus agent do for performance marketers, and where does it fail?
- Will AI agents like Manus replace dashboard SaaS in marketing?
- How do Meta's AI connectors let outside AI agents work in your ad account?
- Why do marketers distrust AI insights, and how do you make them verifiable?
- SOMIN for agencies: pitch and campaign intelligence
