--- title: "Our team updates client records by talking to a Slack bot, and every change keeps a name" description: "One\" lives in our company Slack. It answers questions, takes notes, changes facts and files requests in plain words, and every change it makes is recorded in the Hub under the name of the person who asked." section: Inside Sneeze It author: David Steel published: 2026-10-10T20:15:06.379Z url: https://one.sneeze.it/blog/one-in-slack-updates-the-record tags: ["inside-sneeze-it", "hub", "one-in-slack", "slack", "ai-guardrails"] --- # Our team updates client records by talking to a Slack bot, and every change keeps a name _One" lives in our company Slack. It answers questions, takes notes, changes facts and files requests in plain words, and every change it makes is recorded in the Hub under the name of the person who asked._ Most of what an agency knows about a client is said out loud before it is ever written down. A strategist hears on a call that a location changed its website. A coach learns the owner prefers texts to email. Someone notices an offer ended early. In most agencies that knowledge lands in a chat thread and stays there, because opening the client record, finding the right field and typing it in is a second job nobody has time for in the moment. We decided the record should come to where the team already talks. So the [Hub](https://hub.sneeze.it), the one record every Sneeze It tool reads from, has a teammate in our company Slack. It is called One. You mention @one in a channel or send it a message, in plain words, and it does the thing in the Hub. ## Seven things from one message One does seven things, and only seven. It can show help, list what is still open, refresh the numbers, open a ticket, set a fact, answer a question, and take a note. Anything said to it about a client that is not one of the other six becomes a note on that client in the Hub. It finds the client the way people actually refer to them: a full name, the name without the brand, a distinctive first word, initials, or an alias. Questions ("how did last month's offer do?") are answered from the Hub's numbers and live ad reads, never only from a stored snapshot. "Refresh" reads Meta, the join data and the call center now. :::stats 7 | things One does from a single message 0 | AI guesses in deciding what a message means 1 day | from "I want it to be @one" to live source: the Hub's Slack command router and build history, September 29, 2026 ::: ## Rules decide what you meant. AI only answers. This is the decision we care most about. When a message arrives, a fixed set of plain rules decides which of the seven actions it is. No AI model is asked to interpret the request. The same words always do the same thing, which means a teammate never has to wonder whether "set the website" will be read as a question this time. AI is used for exactly one of the seven: answering a question. Then it works with read-only tools over the Hub's data and live ad reads. It can look; it cannot change. Every change to the record comes from a rule the person triggered on purpose. :::pullquote The same words always do the same thing. AI answers questions; it never decides what you meant. ::: :::flow "What happens to a message" Slack | A teammate mentions @one or sends it a message Check | The Hub confirms it came from Slack and from a Sneeze It account Rules | Plain rules pick one of seven actions Hub | The note, fact or ticket lands on the client's record History | The change is recorded under the person's own name ::: One is internal only. It works for Sneeze It staff, recognized by their company account, and it refuses to work in any channel shared with another company. A client can never read an answer meant for our team. ## Every change still has a name and a time Changing a fact looks like this, using a sample client: > **@one** set Sample Fitness Riverside website to samplefitness.com > > **One:** Done. Website changed from (blank) to samplefitness.com. Recorded in the Hub. Before it saves, One checks that the value looks real for the field. A ZIP code has to look like a ZIP code. Then the change is made as the person who typed it, not as "the bot," and it goes into the same history every Hub page keeps: the old value, the new value, who, and when. Someone opening that client's page next week sees exactly what changed and who changed it, whether the change came from Slack or a click on the page. :::compare "Knowledge in a chat thread" "Knowledge on the record" The website changed, and it was mentioned in a channel. Nobody updated the record. The next person looks it up, finds the old one, and asks again. --- "@one set ... website to ..." The fact changes in the Hub, under the name of the person who said it, with the old value kept. Every tool that reads the Hub sees the new one. ::: :::figure wide ![An agency team room where a teammate types on a laptop while a small magenta card slides from the laptop into a wooden card catalogue drawer that already holds many grey cards](/blog/media/11b74161daadec7a86fee1cf.jpg) Said in Slack, kept on the record, with a name and a date. ::: ## A request, one question at a time "ticket" plus a client opens a request without leaving the conversation. One walks through the same questions the Hub's request form asks, one at a time, with skip, back, cancel and send. When it is sent, it is filed to the Hub and to the team's Jira desk, the same way a ticket opened on the location page would be. The forms themselves, and the 99 old requests that shaped them, are in [We read 99 old requests and rebuilt every form](/blog/99-old-requests-rebuilt-into-forms). One also works the other way around. Other Sneeze It tools can post a question to the right teammate through One, and any reply in that thread closes the question and becomes a note on the client. When a client's success coach changes, One announces it in the coaches' channel and messages each person it affects. ## Listening, carefully, in one channel The newest piece is the most limited on purpose. In one internal channel only, One listens. When a teammate says something about exactly one client, a small AI model restates it plainly and sends it to our shared Brain as a candidate lesson, so the knowledge reaches the other tools instead of scrolling away. It never replies in the channel, it skips anything about money, and it has a daily cap. It is switched on for that one channel and nowhere else for now. All of this rests on the same principle as the rest of the Hub: one record, read by every tool, with every change traceable. That is the subject of [One record, every tool](/blog/one-record-every-tool). :::takeaways "What this means if you run 20 locations" - What your agency hears about your locations should end up on the record, not in a chat thread. Ask where a mentioned change goes. - Every change to your locations' facts should carry a name and a date, whether it was made on a page or in a message. - Ask which parts of an agency's tools use AI to decide, and which use it only to answer. "Rules decide, AI answers" is a good answer. - Client-facing channels and internal tools should never mix. A bot that refuses shared channels is a simple, checkable rule. ::: :::cta Want to see how your locations' records stay current? Book a walkthrough and we will show you the Hub and a sample @one thread. :::