xAI has taken the concept of an AI colleague to the next level. On September 28, it officially launched Team Bots—intelligent agents called Grok that can work alongside an entire team and learn with them. Grant it access to the necessary files, applications, and expertise, then share it, and everyone can work together based on the same context, while each person's private conversation with it remains locked in their own drawer.
Each Team Bot grows around a shared role or workflow within the team, built from four core capabilities. Context is fed into it through relevant files, instructions, and skills, including brand guidelines and internal documents; plugins allow it to dive into apps like Salesforce, Notion, and GitHub, either connected individually by each person or configured for the entire team; credentials enable secure access to third-party APIs without existing plugins; and memory helps it retain what it learns, gradually becoming more efficient over time in its role. The bot is shared, but each person's conversation remains private—it keeps independent contexts and memories for each user while reusing shared skills. Teams can also collaborate directly with it in Slack. Each Team Bot has its own handle, which can be added to a channel so everyone can ask questions, add context, and view its responses.
Within SpaceXAI, Team Bots have already been integrated into some of the most context-heavy workflows. In sales and customer success, each major client has a dedicated Team Bot, shared by the account manager, success manager, architect, and sales lead. It reviews company news, recent Gong calls, Notion documents, and related Slack discussions every night and sends a briefing each morning in the client channel, explaining what has changed and what each person should do next, even providing tailored draft materials. During the day, the team uses it to evaluate strategies, test ideas against customer data, and plan next steps. It remembers every decision the team makes, becoming a record system when people change roles, quickly helping new members catch up. Harper Insurance has taken it even further: instead of manually checking each client across three platforms, pulling balances, and sending personalized emails, they built a Team Bot in 24 hours that automated the process, recovering over $120,000 in policies for clients, and now it runs continuously as an engine.
In product and engineering, the Engineering Team Bot resides in project Slack channels, connected to Notion, Linear, Hex, Datadog, and Cursor, tracking product decisions, sorting bugs, creating tickets, and even initiating Cloud Agents to tackle well-defined fixes. The team taught it the release process, so it knows how to review PRs, when to ask the team for help, and what evidence is needed before a change is completed. After coordinating across tools and agents, it reports progress and obstacles in Slack. Interestingly, Team Bots themselves were built using this setup—a bot guided a Cursor Project, orchestrated hundreds of Cloud Agents, and helped a five-person team submit over 100 PRs daily, getting Team Bots launched within weeks.
In marketing, the Marketing Bot consumes brand guidelines, blogs, and social media copy. When someone in Slack drops off a draft for review, it checks it line by line according to tone and the latest information. Regional teams get feedback in their own time zones and can continue pushing without waiting for headquarters approval. For website and SEO changes, it takes the work from review all the way to publishing. Once content is approved, it makes the changes and shares preview links in Slack for final sign-off. Anyone on the team can submit a website update without raising a ticket or waiting for engineering. The data analytics team receives dozens of one-time analysis requests daily, so they built a Data Bot, allowing anyone in the company to get answers directly without waiting for analysts or requesting warehouse access. It uses shared read-only credentials to check approved tables in Databricks and connects to Datadog, Hex, and Statsig to find issues and generate results. This team spent two years building a skillset library covering 45,000+ tables, all given to the Data Bot, including instructions on finding the right data, analyzing fraud feature usage, following chart standards, and updating security dashboards across the company. The bot also remembers corrections made to its queries, improving the answers it gives to everyone based on knowledge learned from one person.
Team Bots are now available in public beta in Teams and Enterprise plans. Users can directly share their favorite Grok Bots with their team or start with the pre-built sales, product management, marketing, and data analytics Team Bots. When xAI combined the four essentials—context, plugins, credentials, and memory—into a shared intelligent agent, and let it work alongside the team in Slack during the day while digesting company-wide updates at night, it has already taken a real step forward, moving AI from a personal assistant to a true team colleague.
