AI agents move from sidecar tools to embedded coworkers
AI agents embedded in business platforms are autonomous software systems that live inside tools like chat, payments, and dashboards, where they observe activity, act on behalf of users, and coordinate workflows across data sources without forcing workers to switch contexts or copy information between separate applications.
The most important shift in enterprise AI workflow is this: agents are no longer destinations, they are coworkers sitting inside the tools teams already use. Anthropic’s Claude Tag moves an advanced model out of a private chat box and into shared Slack channels, summoned with a simple @Claude mention so teams can collaborate with it in the open. In parallel, Square is wiring ChatGPT and Claude directly into its commerce rails so AI can help customers discover merchants and complete orders inside AI conversations, rather than redirecting them elsewhere. These AI agents inside business platforms promise faster work and smoother buying journeys—but they also raise hard questions about control, security, and compliance that enterprises can no longer postpone.
Claude in Slack: multiplayer AI and the new enterprise workflow
Claude Slack integration is a deliberate attack on one of knowledge work’s biggest time sinks: bouncing between chat, browser, docs, and internal tools. Standard generative software forces employees to copy data out of team chats into separate browser windows, then paste results back. Claude Tag reverses that. Teams pull the model directly into shared channels with @Claude, delegate tasks in public, and let everyone watch status updates inside the same thread. This is what AI agents business platforms should look like: multiplayer by default, context-aware, and visible to the whole team.
The agent runs on Anthropic’s Opus 4.8 engine and can break work into steps, connect to internal databases and code repositories, and operate asynchronously without constant human nudging. In “ambient” mode, it quietly monitors conversations, surfaces priority alerts, and keeps track of unresolved tasks over days. That is a profound workflow change: instead of project managers chasing updates, the agent keeps the to-do list alive. Anthropic reports its own product group now generates 65% of its code with an internal Claude Tag variant—a quotable sign that this is not a toy, but a rewire of how engineering work gets done.
Agentic commerce: Square turns AI chats into checkout lanes
If Claude in Slack shows how AI agents reshape internal workflows, Square’s agentic commerce automation shows how they will reshape revenue. Square has rolled out integrations with ChatGPT and Claude to “help enable agentic commerce” by letting buyers discover and transact at the moment of decision inside AI-powered conversations. Instead of searching, clicking through ads, and navigating clunky menus, a customer can ask an AI what to eat, see a restaurant using Square, and place an order within that same interface.
The first sellers live on this system are Food & Beverage merchants with active Square Online Ordering profiles in the U.S., and Square is not charging additional marketplace commissions on those AI-initiated orders. For sellers, the pitch is blunt: “accessing customers without new complexity”. There is no new API, no extra tool to configure, and no added fee; eligibility is opt-in by default, controlled from the existing Square Dashboard while Square syncs menus, hours, and ordering data in real time. This is agentic commerce automation wrapped in an interface sellers already trust, and it sets a precedent: the smartest way to adopt AI agents is to hide their complexity under familiar dashboards.
Why embedded beats standalone—and why governance will decide the winners
AI agents embedded in existing platforms reduce friction because they respect how people already work and buy. Tagging Claude like a coworker inside Slack channels is powerful precisely because it removes any extra cognitive or technical step; as Anthropic’s Cat Wu points out, the form factor mirrors how teams already mention colleagues in threads. Likewise, Square’s sellers stay inside the same dashboard while Square handles the messy integration work in the background. This is how AI agents business platforms should evolve: invisible wiring, familiar surfaces.
But the convenience comes with a governance bill. Claude Tag’s ambient monitoring and cross-app access demand a serious security model. As Anthropic notes, background agents that can read chat histories, connect to email, and tap corporate tools “require a distinct security infrastructure to protect proprietary information”. Scoped identities, channel-level memory, logging of user queries, spending caps, and careful access boundaries are not optional extras; they are the price of letting an always-on agent roam Slack. Enterprises will have to decide whether the gain—lower task friction, richer shared context, and less manual tracking—is worth the overhead of audits, compliance reviews, and configuration work. Those that treat governance as a first-class product requirement, not an afterthought, will turn these agents into durable advantages instead of quiet risks.
The next phase: standards, competition, and the risk of invisible sprawl
These launches are not isolated experiments; they are early moves in a platform land grab. Anthropic’s Claude Tag beta comes on the heels of a USD 65 billion (approx. RM299.0 billion) Series H that pushed its valuation to USD 965 billion (approx. RM4,442.9 billion), ahead of OpenAI’s USD 852 billion (approx. RM3,920.0 billion) mark. According to Ramp’s May 2026 AI Index, Anthropic has already edged past OpenAI in enterprise adoption, 34.4% versus 32.3%. At the same time, Square says ChatGPT and Claude are only its first live AI integrations, with more coming and active work in agentic commerce protocol groups to define how AI agents and commerce platforms should interact.
The direction of travel is clear: more agents, more channels, more invisible automation. The danger is quiet AI sprawl—agents wired into every chat, payment flow, and dashboard without a unified view of what they can do or see. The conclusion for leaders is straightforward. AI agents inside Slack, Square, and other business platforms are where real productivity and revenue gains will come from. But the winning enterprises will be those that move fast on embedded use cases while moving even faster on governance: standardising permissions, demanding open protocols, and treating every new agent not as a cool add-on, but as a powerful user whose access must be earned and monitored.
