Hops Opens Its Collaborative AI Workspace After 2,500-Team Early Access

Hops opened its Collaborative AI workspace to everyone on November 10, 2026, after more than 2,500 teams used it during early access. The product, built by For Good AI and led by Wrike founder Andrew Filev, puts AI agents into shared team documents, reusable work methods and custom apps rather than confining them to private chats.

The launch targets a measurable gap between individual and organizational productivity gains. Filev said his conversations with founders and executives of established companies usually produce estimates that their organizations are moving 10% to 20% faster than a year earlier — short of a 2x improvement and short of what individual AI gains would suggest. His framing reaches back to earlier workplace computing: "The PC era didn't pay off when PCs got faster. It paid off when they were networked. I believe AI will follow the same path."

Who Is Behind Hops?

For Good AI develops Hops, and the same team builds Zencoder's AI coding agents, with Filev serving as founder and CEO. According to the company's about page, Filev previously started Wrike, the work-management platform that was later acquired for $2.25 billion. That track record gives the launch weight in a crowded AI workspace market, where vendors are competing to move AI from individual assistants to team-level systems of record.

What Does the Shared Workspace Actually Change?

In Hops, people and the AI agent work together in shared chats, documents, tables and decks. When someone asks the agent to draft or revise material, it writes directly into the shared file, so all participants work from the same version instead of passing AI output between private chats and team documents. The product page describes live editing, with AI document edits visible in version history — a feature aimed at making agent contributions auditable rather than invisible.

The workspace rests on three structural elements:

• Shared files. The agent writes into documents, tables and decks that the whole team uses, eliminating the copy-and-paste loop between personal AI chats and team tools.
• Reusable skills. Teams can save a working method as a reusable skill containing instructions, context and format. The person who supplied the knowledge controls access and can update it, so colleagues apply the same method without reconstructing the process each time.
• Custom apps. If a team lacks a tool, it can describe an application and have Hops build and run it inside the workspace using the same data and permissions.

Hops illustrated the skills concept with its own marketing lead, who taught the system the company's voice so a salesperson could request a customer email written directly into a shared document. The company says the workspace connects to more than 2,000 apps, including CRM and cloud-storage tools, allowing the agent to work from connected team data rather than isolated context.

How Do Permissions and Pricing Work?

The permission model governs what the agent can see and send. Hops will not use a file that the person making a request cannot open. If an answer draws on information others in the thread cannot see, the reply is held until the requester approves it. External sends and sensitive changes also wait for human approval, and administrators can control which teams use connected tools or make those tools read-only. The design addresses a persistent enterprise objection to AI agents: the risk that automation quietly widens access to restricted information.

Pricing is usage-based rather than per-seat:

• The pay-as-you-go plan includes 5,000 free credits at $0 per user per month.
• Top-ups cost $20 for 20,000 credits.
• Teams draw from a shared credit balance.
• Administrators can set spending limits per teammate.
• Purchased credits remain valid for 12 months.
• Chat, documents and collaboration do not consume credits.

Consumption for an AI task depends on the model used, the amount of information read, the tools invoked, the number of steps and the output length. The Enterprise tier adds the full integration library, SSO (single sign-on), SCIM (automated user provisioning), audit controls, a security review, custom terms and the option to bring an API key. Hops states that it does not use customer data to train models — a position consistent with growing regulatory and contractual pressure on AI vendors handling corporate information.

Has Hops Proven the Model Internally?

Hops reports that 30 engineers and a fleet of AI agents rebuilt the company's collaboration stack in a year. The company now runs its own CRM, support desk, IT requests and project tracking inside the product, and employees can ask about leave policy and receive an answer linked to the handbook. The internal deployment illustrates the product's ambition: coordinating routine operations alongside drafting work, not merely generating text on request.

One early customer supplied a named endorsement. Discern Security CEO Sai Venkataraman said in the launch release that his sales and marketing teams use Hops for pipeline reviews, event and website-prospect follow-ups, and campaigns. He also said the product replaced tools including Slack and Jira at his company — a concrete consolidation claim that, if replicated across customers, would position Hops as a substitute for established collaboration infrastructure rather than an add-on to it.

What Comes Next?

The public opening turns Hops's early-access period into a broader adoption test. Shared skills and permission controls are designed to make AI output reusable without widening access to underlying information, while credit pricing removes per-seat charges and ties heavier automation to actual consumption.

The central question is whether teams will move enough recurring work into the workspace for those shared methods to close the company-level gains Filev described as lagging — the 10% to 20% organizational improvement that falls far short of what individual AI users experience. His networking analogy sets the benchmark: if shared context does for AI what networks did for PCs, the workspace model could reset expectations for team productivity software. If it does not, Hops becomes another AI layer atop the tools it aims to replace.