AI Agents Stopped Being an App. They Became a Coworker.

Four products shipped between Sunday and Monday that make the same bet: AI agents should live inside the apps where work already happens, not in a separate tab you have to remember to open.

Claude Tag from Anthropic, Agent Architect from Zoom, Spark from Productboard, and Connect AI Developer Edition from CData all landed June 22-23. Each takes a different angle on the same problem, but they agree on the solution. Stop sending people to a chatbot. Bring the chatbot to where the people are.

Claude Tag: Claude joins your Slack channels

Anthropic launched Claude Tag on June 23. Claude joins a Slack workspace as a team member. Admins grant it access to specific channels and connect it to tools, data sources, and codebases. Then anyone in the channel types @Claude with a request.

The internal number is worth pausing on: 65% of Anthropic's product team code is now created by Claude Tag. They use it to chase down product metrics, work through support tickets, and debug tricky issues. Engineering first, but spreading into operations.

Three things separate it from a Slack bot.

It is multiplayer. One Claude per channel, visible to everyone. Anyone can see what it is doing, pick up where someone else left off. That is the shift from personal assistant to team member.

It builds persistent context. Claude remembers what happens in its channels. You do not re-explain the project every time. It can pull from other channels too, if you grant permission.

It takes initiative. Turn on ambient mode and Claude proactively flags things it notices, follows up on stale threads, surfaces information from connected tools.

Pricing: available in beta for Claude Enterprise and Team customers. Admins set monthly token spend limits per channel and per org. Anthropic is issuing introductory launch credits so entire companies can try it. Claude Tag works with Opus 4.8.

This is the product that makes "tag Claude in" a normal workplace phrase. Whether that excites or unsettles you depends on your team.

Zoom Agent Architect: Build AI agents from a prompt, pay when they resolve something

Zoom announced Agent Architect and Agent Performance Suite on June 22 at CCW Las Vegas. Agent Architect generates production-ready voice and digital customer service agents from a text prompt. You describe what you want the agent to handle. It reads intent, fills in context, connects to data sources, and assembles the workflow.

The Performance Suite adds the testing layer most companies skip. You simulate customer interactions before launch, then track resolution rates, containment, CSAT, and cost per resolution after launch. KB Suggestions watches successful human-assisted resolutions and drafts knowledge base articles from them for review.

Zoom also introduced outcome-based pricing for Virtual Agent. You pay when an interaction is resolved or successfully routed. Not per seat. Not per message.

ZoomMate, at $20 per user per month, now includes embedded AI agents that handle follow-ups, execute workflows, and coordinate action across Salesforce, ServiceNow, Google Drive, Confluence, and OneDrive. You create them from a short prompt inside Zoom Chat.

Futurum Group analyst Keith Kirkpatrick called the lifecycle integration a direct hit on "the persistent gap between AI promise and actual business value." That is analyst-speak for: most AI deployments stall because nobody built the measurement layer. Zoom built the measurement layer.

Pricing: Agent Architect and Performance Suite available on Zoom Virtual Agent plans. Outcome-based pricing is optional. ZoomMate is $20 per user per month with embedded AI agents included.

Productboard Spark: Product management rebuilt as AI-native

Productboard announced Spark on June 22. The claim: "the first agentic product system." They rebuilt the platform from the ground up as AI-native.

Spark runs specialized agents for different product jobs: voice-of-customer analysis, spec writing, roadmap prioritization, stakeholder updates. It surfaces customer opportunities from your real feedback data, writes delivery-ready specs, and scales your team's existing thinking instead of starting from scratch.

The gap they are filling is real. Product managers spend enormous time on manual research, spec-writing, and coordination. Those tasks live in a different tool from where the product data lives. Spark collapses the gap.

Hubert Palan, Productboard's CEO, put it plainly: "Your job is to ship great products to your customers, not become an AI infrastructure team."

Pricing: Spark is available now in existing Productboard workspaces. You enable it from the dashboard. Specific pricing tiers not yet disclosed.

CData Connect AI Developer Edition: Free governed data access for AI agents

CData launched Connect AI Developer Edition on June 23. It is free.

The problem it solves: most enterprise AI projects die at the data layer because getting governed access to production systems (Salesforce, Snowflake, NetSuite, Workday, Microsoft 365) requires IT involvement at every step.

Connect AI exposes enterprise APIs as a queryable data layer with standardized schema, read/write support, and automatic auth handling. Developers write queries. IT gets visibility and control.

The free Developer Edition includes MCP server support, per-user authentication passthrough, query logging with user attribution, and works out of the box with Claude Code, Codex, Cursor, and LangChain. It also ships with Toolkits, which let you package governed data access into a single MCP Server URL scoped to specific use cases. Agents get exactly what they need and nothing more.

The open-source Python SDK provides DB-API-compliant access for existing Python workflows. The CData CLI gives coding assistants direct connector access.

Raviv Levi, CData's Chief Product and Technology Officer: "Developers have been forced to choose between moving fast and meeting the governance their company requires. That tradeoff doesn't hold up anymore."

Pricing: Connect AI Developer Edition is free. Python SDK is open source. CLI is free. All available now at cdata.com/developers.

CData's customer list tells you why this matters: Anthropic, Databricks, Microsoft, Google, Palantir. They are powering AI workloads for the companies building the AI tools everyone else uses.

What ties them together

All four products bet on the same thing: AI works better when it meets people where they already are. Claude Tag goes to Slack. Zoom goes to the contact center. Productboard goes to the product roadmap. CData goes to the data layer that feeds all of them.

None of these products ask you to learn a new interface. They ask you to do what you already do, with an AI teammate added to the workflow.

The pricing models tell you something too. Zoom's outcome-based pricing and CData's free Developer Edition both bet that lowering the barrier to entry matters more than extracting per-seat revenue. Productboard folds Spark into existing plans. Claude Tag runs on token credits with admin-set limits.

If you are evaluating AI tools this month, the question is no longer "which AI app should I buy?" It is "which of the tools I already pay for just became AI-native?"

---

Sources: Anthropic, Zoom, Productboard, CData (PR Newswire), CMSWire, Futurum Group, CXM World