The follow-up is not a personality problem. On most ops teams it is a missing workflow. The same question leaves someone's desk, lands in email or Slack, and then depends on memory. Memory is a bad system of record.
Give the request an owner, a reminder, and a place to close. Keep a person in the loop only where the answer is actually unclear.
What the chase looks like on a normal Tuesday
A request goes out before lunch. "Can you confirm the handoff time?" The recipient is in meetings. By late afternoon the sender copies the note into Slack "just in case." Wednesday morning a third person asks the same thing in a different thread, because they cannot tell whether anyone answered.
Nothing is broken in the tools. Email delivered. Slack notified. The failure is that the request has no state. It is open, nudged, or done, but nobody wrote that down outside the messages.
Why another channel or chatbot does not fix it
A new channel feels like progress. It usually adds a third place to check. People still answer where they already live, which is their inbox or the team channel they had open.
A chatbot that says it can remind you has the same gap. A reminder with no record is another ping. If it cannot see the original request, the last reply, and whether the item is closed, it will nag the wrong person or repeat a question that was already answered.
The fix is a loop with a beginning and an end, not a louder notification.
Map the loop: ask, wait, nudge, escalate, close
Write the loop in five moves before you automate any of them.
Ask. One request, one owner, one question. "Need the handoff time by Thursday at 3." Vague asks create vague threads.
Wait. Set a real window. Four business hours, or the next morning, is a plan. "Soon" is not.
Nudge. One follow-up, in the same thread, that restates the question and the deadline. Do not start a new conversation.
Escalate. If the window passes, the workflow tells a named backup, not the whole channel. The backup needs the original ask and the fact that nobody answered.
Close. Someone, or the workflow after a clear reply, marks it done and writes the answer where the next person will look. A thread that fades out is not closed.
If you cannot name those five moves, automation will only send more messages.
What the workflow should remember between messages
The messages are not the record. The workflow should store a small set of fields.
Who asked. Who owns the answer. The exact question. When it was asked. When the nudge is due. The last reply, even if it was "not yet." Whether the item is waiting, escalated, or closed. Where the final answer should be written, such as a ticket, a sheet, or a project task.
That memory lets a later step see that Wednesday's Slack note is the same request as Tuesday's email. Without it, every ping looks new.
Where an agent can draft or route, and where a person must answer
An agent is useful on the messy edges. It can read a reply and tell "Thursday at 2" from "we're still checking." It can draft the nudge in the team's tone. It can route a clear answer to the record and leave an unclear one for a person.
A person still has to answer when the reply changes the plan. A conflict between two dates, a no that needs a new owner, or anything that commits the team to a customer. The agent can prepare a short brief. It should not send the commitment on its own.
Draw that pause on purpose. If the only signal is a message someone might miss, you still have a chase. You just added a model to it.
A starter workflow you can run with tools you already use
Start with one request type that happens every week. Vendor confirmations, internal handoffs, or a draft that needs review. Do not boil the whole company into one scenario.
Trigger it from the email or Slack message you already send. Create one tracked item with the fields above. Schedule a single nudge. If there is no clear answer, assign a backup and stop. When the answer is clear, write it back to the ticket or doc and close the item.
You do not need a new inbox. Keep the conversation where it already happens, and give it a state. A visual workflow can sit on top of that email and Slack without replacing them.
How to tell it is working without fake time-saved numbers
Skip invented hours. Watch a few plain signals for two weeks.
Count open items that have an owner. Count nudges that went out once, not five times. Count items that closed with an answer written down, versus threads that simply went quiet. Ask the coordinator whether they still re-ask the same question from memory.
If those counts move the right way, the loop is doing its job.
Innflow is a general agent and workflow builder for this kind of loop, on top of email, Slack, calendars, and the SaaS you already run. The app is at https://app.innflow.ai.












