News
LLMs changed what workplace software is for
The last two years did not add a summarizer to the sidebar. They changed the job of chat, work management, and CRM, from a place people talk to a place software can act.
What actually shipped
Salesforce put Agentforce inside Slack: agents that search the workspace, act in a channel, and sit on CRM data. Microsoft put Copilot collaborative agents inside Teams, grounded in Graph, with identity and compliance attached. Microsoft’s own Work Trend Index says active agents in that ecosystem grew about 15× year over year — 18× in large enterprises. That is no longer a demo reel. It is a line item.
The same split showed up in the other two catalogs. Rovo in Jira, monday AI, and Asana all sell AI that drafts, assigns, or chases work inside the board. HubSpot Agent Hub sells follow-ups, drafts, and agents in the record. Salesforce Agentforce is the place you build. The products did not become “AI companies.” They became runtimes — or they did not, and that gap is now the grade.
Chat is now a runtime
A workplace chat used to be scored on channels, search, apps, and whether IT could turn someone off. Those still matter. The new axis is whether an agent can do work in the thread: a handoff, a task, a follow-up, a custom action a developer will actually ship.
That is why Slack and Teams both score as agent platforms in this grader, and why they assume a paid AI SKU. It is also why a company that only wants people messaging people should say so. Loading the agents axis for a café that will never buy Copilot is how you get a pretty ranking that nobody uses.
WhatsApp did not get this update
Frontline communication research still finds deskless teams on noticeboards and ownerless WhatsApp groups. Hospitality, retail, and site crews live there because it is already on the phone. LLMs did not change that. They made the cost of staying there clearer. A personal group chat is not a corpus an agent can read, not a place IT can revoke, and not a workflow with an owner and a due date.
WhatsApp can still win a grade. A two-person shop with no leavers and no appetite for agents should not pretend it needs Slack. The point is the trade: simplicity stays a 5, agents stay a 1. If the company now wants software that acts, the default messenger is the wrong runtime. That is an industry update, not a branding preference.
Boards and pipelines got the same split
Work management used to be “can people see the work.” The LLM year added “can the tool chase the work.” Summaries are cheap. Assignment and follow-up inside the record are the useful layer. Custom agents on Jira’s Rovo platform are a different purchase from a visual board with a chatbot bolted on.
CRM split the same way. Pipedrive is still the faster pipeline. HubSpot is still the inbound platform that earns a click on drafts and reminders. Salesforce is still the place you build. “We need AI in the CRM” is not an answer. Suggest, act, or build is the answer — and it moves the winner.
The buying question changed
Procurement used to ask which chat people would open. That question is still decisive for deskless teams. The new question sits next to it: how far is AI allowed to go — chat only, suggest drafts, act in the thread, or a platform developers will build on.
Two things did not change, and vendors hope you forget both. First, hourly and multi-site teams still will not open a desk tool because it grew a better model. Second, lock-in got worse. Agents grounded in Microsoft Graph or a Slack workspace are not “export a ZIP.” The corpus is the product. Switching cost was already the hidden TCO line. LLMs thickened it.
Company-owned history matters more for the same reason. An agent that cannot see last month’s shift chat, last quarter’s deal notes, or the task that slipped is a parlour trick. Personal messengers fail that test even when the model is excellent.
What the grade does with it
SaaS Grade already asks this as the last question in every catalog: Chat, Suggest, Act, or Build. “Nothing yet” does not punish anyone. Native and platform answers load the agents axis — and, for Build, the ecosystem axis — because that is what the last year of product shipping actually did.
Zenzap is scored on the same rules. It is a 4 on agents for in-thread tasks and handoffs, not a 5 on a developer platform, and it does not always win. The honest move is still the same: answer the company you have, including how far you want a model to go, then send the poster.
Grade the company you actually have
90 seconds. The last question is Chat, Suggest, Act, or Build. Then send the poster to the person still arguing in the abstract.
Sources
Vendor pages and research the claims above rest on. Open them.
- Salesforce — Agentforce in Slack: Agents that search Slack, act in-channel, and sit on CRM data.
- Microsoft — Copilot collaborative agents in Teams: Agents grounded in Graph, with enterprise identity and compliance.
- Microsoft Work Trend Index 2026: Active agents in the Microsoft 365 ecosystem grew ~15× year over year, 18× in large enterprises.
- Atlassian — Rovo in Jira: AI search, summaries, and agents inside Jira workflows.
- monday.com — AI Feature Catalog: Credit-metered AI blocks, updates, and agents on the work platform.
- HubSpot — Agent Hub: CRM-grounded agents for drafts, follow-up, prospecting, and custom builds.
- Salesforce — Agentforce: The CRM platform play for building and running agents on customer data.
- ChangeEngine — Frontline communication tools, 2026: Deskless teams still run on noticeboards and ownerless WhatsApp groups.