For most business users, the asset system is somewhere they have to remember to log into. The brand director, the QC lead, the compliance manager, the music supervisor, the delivery operator: they live in Slack or in Microsoft Teams all day, and every trip out to "the MAM" is a context switch. So we stopped asking them to make the trip. The Ceivo agent now lives in the channel they were already in.
The pitch is simple. Why pull a brand director out of their team's Slack channel to read a brand-compliance finding, when the finding can land in that channel as a card, with a thread already open, and the next question they have can be answered by the agent that produced the finding? Why force the QC team to scrub a marker timeline when they can ask, in plain language, "how many other QC failures have we had from this vendor this quarter, and what should I tell them?"
Subscribe Your Own Channel to the Topics You Care About
The model is built around the way real teams already organize themselves. Each business group has its own Slack or Teams channel. The brand-compliance team has one. The QC group has one. The delivery operations team has one. The music-supervision pod has one. Inside each channel, the team owner runs a single command to subscribe that channel to the topics that matter to them, at the severity threshold they want to be paged at.
A typical post-production channel looks like this after a couple of minutes of setup:
Subscriptions for this channel
• compliance ≥ warning
• qc ≥ warning
Topic catalog (topic → agents that emit it)
• brand → brand_standards_check, brand_standards_report
• captions → sdh_style_scanner
• compliance → compliance_scanner, compliance_marker_report
• delivery → mezzanine_verification, edl_export
• ingest → youtube_downloader
• music → music_recognition
• qc → compliance_scanner, brand_standards_check, baton_qc,
mezzanine_verification, rating_verification,
sdh_style_scanner, atrium_submissions_assistant
• rating → rating_verification
• reports → compliance_marker_report, brand_standards_report, edl_export
• rights → music_recognition
• submissions → atrium_submissions_assistant
A topic is a label, not a single agent. The qc topic catches everything an upstream QC pass might emit; the brand topic catches the brand checks; the reports topic catches the rolled-up PDFs that summarize a file. Subscribing the channel to a topic means every agent that emits to that topic, at or above the severity the channel chose, will land as a card in that channel as the run completes. A brand team that only wants to see WARNING and above on brand work, and HIGH and above on rights, just says so: subscribe brand ≥ warning, subscribe rights ≥ high. No engineering, no admin ticket, no sprint.
Cards Are the Hook. The Conversation Is the Product.
When an agent finishes, the bot posts a card. The card carries the file, the verdict, the highlights, links to the marker report, and a thumbnail or two so the team has context without opening anything. That part is just notification, and notification is the easy half.
The product is what happens in the thread under the card.
A QC failure lands. The QC lead opens the thread and asks the questions a senior reviewer would ask out loud:
"Why did this fail? How many other failures have we had from this vendor this quarter? Are they trending in the wrong direction on the same defect? What do other vendors do better here? Help me draft an email to them that explains the finding, points to the spec they missed, and offers a path to get back into good standing."
The bot answers each one in turn. It pulls the file's marker report. It cross-references the vendor's history against the same QC profile. It surfaces whether the same defect appears across other recent deliveries. It drafts the email in the team's tone, references the right clause in the house spec, and stages the message so the QC lead can read it, edit it, and send it without ever leaving the channel. The agent that flagged the failure is the same agent that helps decide what to do about it.
The same loop runs against any topic. A brand-compliance card lands; the brand director asks why a particular ΔE color check tripped on a partner-supplied still and what other partner deliveries were close to the threshold this month. A music-recognition card lands; the music supervisor asks which detected cues are absent from the cue sheet and which sheet entries have drifted in timing. A compliance card lands; the legal lead asks how many similar findings have surfaced this quarter and what the typical adjudication has been.
The bot is not pretending to be a human. It is doing the part of the human's job that is assembly: pulling files, joining history, drafting prose, surfacing precedent. The judgment stays with the person in the channel. They confirm, edit, and send.
Built on the Same Substrate, Surfaced Where the Work Lives
There is no second platform to learn. The bot is backed by the Ceivo MCP, the same ceivo-api skill our agents use, and the agent-run lookup tool that sits on top of the architecture we laid out in Beyond the API. Every action the bot takes is the same action a Ceivo user could take inside Ceivo. The difference is purely surface: the surface is now Slack, and Microsoft Teams is in scope alongside it, so the brand director who lives in Teams gets the same experience the QC team gets in Slack.
The bot also carries memory at the channel and the user level. It learns which projects a channel watches, which thresholds the team has tuned, which vendors map to which liaisons, which tone the team uses when emailing partners. By week two, the QC channel's email drafts read like the QC lead wrote them, because the lead has been editing the drafts and the bot has been listening.
What This Means for the Business Group That Owns the Work
This is the natural extension of the agent-native architecture we laid out in Beyond the API and the workflow we walked through in the agentic newsroom. If the people closest to the work are the ones who set the rules and tune the agents, the agents should land in the same room they already work in. A brand team's room is their Slack channel. A QC team's room is their Teams channel. A music-supervision pod's room might be a single thread in a shared channel. The agent meets them there.
For the business owner, it means the work surfaces where the work is being done. For the agent, it means every run gets a thread, every thread gets a human reader, and every reader's adjudication is feedback the rule can absorb on the next run. The catalog gets sharper because the people closest to the failures are the people closest to the bot.
What's Next
The Slack chatbot is live now in customer pilots, with the topic catalog above as the starting set and per-channel customization available to any team that asks. The Teams build is in scope and on the roadmap as a peer surface, not a port: same MCP, same skills, same memory, same conversational loop, rendered as Adaptive Cards instead of Block Kit. Customers running Ceivo today can subscribe a channel this week.
If your brand team, QC group, or compliance pod is tired of leaving the channel they live in to talk to the system that knows what just happened, reach out and we will get the bot into one of your channels and tuned to your topics on the first call.