xAI developing new Automations feature for Grok

๐กxAI is evolving Grok from a chatbot into an agentic automation platform for scheduled, multi-step workflows.
โก 30-Second TL;DR
What Changed
Grok's Tasks feature is being integrated into a broader automations system.
Why It Matters
This shift suggests xAI is moving Grok toward an agentic framework, enabling it to perform multi-step, autonomous workflows rather than just conversational queries. It positions Grok as a more competitive tool for productivity-focused power users.
What To Do Next
Monitor the xAI developer documentation or Grok interface for the release of the Automations API to begin building custom agentic workflows.
Key Points
- โขGrok's Tasks feature is being integrated into a broader automations system.
- โขThe new system supports scheduled routines for recurring AI tasks.
- โขUsers will gain access to expanded skill sets and customizable model selection.
๐ง Deep Insight
Background and context from public sources โ not the original article. 9 sources cited.
๐ Enhanced Key Takeaways
- โขThe 'Skills' feature, launched in May 2026, provides reusable workflow packages, allowing Grok's agent to invoke saved capabilities on demand for automated tasks.
- โขThe existing 'Tasks' feature, which the new 'Automations' system will integrate, was launched in June 2025 and already enabled scheduling AI-driven prompts and analyses using live data from X (formerly Twitter).
- โขThe upcoming 'Automations' system will offer users explicit control over which 'Skills' an automation can utilize and provide flexible model selection, moving beyond the current 'Expert mode' toggle.
- โขGrok's automations are designed to perform recurring operations such as monitoring X trends, generating reports, and sending external notifications via email, with individual tasks capable of handling up to 131,000 tokens per run.
- โขThe underlying Grok models, including Grok 3.5 and Grok 4.x, utilize a Mixture-of-Experts (MoE) architecture, with Grok 4.x featuring a multi-agent system to reduce hallucination rates and a reported context window of up to 2 million tokens.
๐ Competitor Analysisโธ Show
| Feature/Aspect | xAI Grok (Automations) | OpenAI (e.g., Codex/ChatGPT) |
|---|---|---|
| Core Function | Scheduled, AI-driven routines with expanded skills and model selection | Programmable AI agents, reusable skills, model selection within automations (Codex); general conversational AI with plugins/APIs (ChatGPT) |
| Real-time Data | Native integration with X (formerly Twitter) for live data analysis | Access to real-time web data (ChatGPT with browsing) |
| Customization | Expanded skill sets, flexible model selection, custom scheduling | Reusable skills, model setting within automations (Codex); custom instructions, plugins (ChatGPT) |
| Output/Notifications | Results via notifications or email | In-app results, API integration for external actions |
| Pricing | Included with X Premium/SuperGrok subscriptions | Various tiers, including free, Plus, Team, Enterprise; API usage based on tokens |
๐ ๏ธ Technical Deep Dive
- Grok's core architecture is based on a Mixture-of-Experts (MoE) model, which divides the model into specialized subnetworks, allowing for more modular and efficient information processing by selectively activating parameters.
- Grok 2.5, a 270-billion-parameter model, utilizes Sparse Mixture of Experts (SMoE), activating approximately 23% of its parameters per token for efficiency.
- Key architectural components also include Grouped-Query Attention (GQA) to reduce KV cache size for faster inference, Rotary Positional Embeddings (RoPE) for encoding word order, RMS Norm for network stabilization, and the SwiGLU activation function.
- Grok 4.x models feature a multi-agent architecture where several specialized AI agents work in parallel, debate, and cross-check queries, reportedly reducing hallucination rates from around 12% to 4.2%.
- Grok 4.20 has a reported context window of up to 2 million tokens, which is shared collectively across its active agents.
- The 'Tasks' feature, and by extension the new 'Automations,' can handle a generous context window of up to 131,000 tokens per run.
๐ฎ Future ImplicationsAI analysis grounded in cited sources
โณ Timeline
๐ Sources (9)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
Weekly AI Recap
Read this week's curated digest of top AI events โ
๐Related Updates
AI-curated news aggregator. All content rights belong to original publishers.
Original source: TestingCatalog โ
This is a summary, not the original. Read the source, or get the weekly briefing.
Weekly AI briefing
One email a week. Unsubscribe anytime.

