Two kinds of follow-up: the chase and the commitment
Two jobs share the name follow-up. One is chasing a reply that never came: you sent the quote, a week went by, silence. The other is delivering what you said you would deliver, which usually lives in a sentence like “I’ll get you the revised numbers by Thursday.”
Software is good at the first one. The second leaves a different trace, but most products in the follow-up category are looking for the first.
The chase runs on thread state
A tool can find an unanswered message without understanding a word of it. It needs the date something went out and the absence of anything coming back. That’s a query rather than a judgment, which is why so many products do it well.
Google ships a version in the base product. Gmail’s nudges, announced in May 2018, push a thread back to the top of the inbox both to reply to messages you may have missed and to follow up on messages that got no response.
Sales tools built a category on the same mechanic. Outreach, Salesloft, HubSpot sequences and lemlist enroll a contact in a cadence, send on a timetable, and stop the moment a reply lands. Inbox assistants sell the one-off version of it: Boomerang tracks a sent message and resurfaces it, SaneBox’s SaneNoReplies collects the mail that went nowhere, and Serif pre-drafts the nudge when a thread goes quiet.
If your team keeps losing the second touch on a quiet prospect, a sequencer runs timed cadences against a prospect list better than we do, and reports reply rates while it does it. We do ship the chase, though: the Follow-up Writer in our catalog drafts a follow-up for threads that have gone quiet, and a Follow-up tracker keeps the list of what is owed one.
The commitment leaves no trace a query can find
You answer a customer at 6pm and, inside the reply, agree to send a revised SOW on Thursday. They write back “perfect, thanks.” The thread now holds a message from you, a message from them, and no open loop a timer can detect. Every signal says handled.
The commitment lives in one place, a clause inside a paragraph inside a message. It has no due date and no owner. A timer has nothing to fire on and a task list has no entry for it.
That’s what people are asking about when they search for a way to track commitments made over email and chat. The information was written down. It never got transcribed into anything that tracks work, which is the same gap we describe on unrecorded knowledge.
Transcribing it yourself is a small task that repeats once per message, so how much time it takes depends on how fast the stream moves. Microsoft’s telemetry study Breaking down the infinite workday puts the average worker at 117 emails and 153 Teams messages a day, interrupted every two minutes, 275 times in a working day. A capture step is quick. Against that volume, it’s the step you skip in a busy week.
Panopto measured waiting on a colleague. Panopto’s 2018 survey with YouGov, 1,001 US adults at organizations of 200 people or more, found the average employee spends 5.3 hours a week waiting for assistance or insights from colleagues, with 66% of those delays lasting up to a week and 12% lasting a month or more.
Three levels of visibility
| What happened | What software sees without reading the words | What covers it today |
|---|---|---|
| You sent, nothing came back | Thread state | Gmail nudges, sales sequencers, response trackers |
| They sent, you never replied | Thread state | Gmail nudges, triage tools, an “awaiting reply” view |
| You replied, and promised something inside the reply | Nothing. The thread reads as handled | Software that reads the sentence |
The top two rows are why follow-up looks like a solved category. The third row is where deals go quiet without anyone noticing, because the last visible event was a friendly exchange.
”No reply in five days” is the wrong test for the third row
A timer measures silence. A commitment is a statement about the future made in words, and silence tells you very little about one.
It misses in both directions. When the other person answers “great, thanks,” the thread stops being quiet, so nothing counts down to Thursday. What can fire, days later, is a nudge that you never replied to “great, thanks”. Wrong clock, wrong thing, and the revised SOW is still unwritten. When a thread goes quiet because the work actually finished, the timer fires anyway and somebody gets chased over nothing. Teams find this out the slow way: they switch on a no-reply rule, get nudged about threads that ended fine, and switch it back off.
What catching the third row takes
Four requirements, and which vendor you pick matters less than whether the product does all four.
Read every place your team talks. A promise made in a chat channel binds you the same way one made by email does, so anything reading only a mailbox starts out missing a large share of them. We read Gmail, Outlook, Slack, Microsoft Teams, and Google Chat.
Read the sentence, not the header. “I’ll send”, “we’ll get you” and “by end of week” live in the prose. Metadata says nothing about any of them.
Produce a record with an owner and a date. A chaser needs something to chase. Extraction that ends in a highlight or a summary leaves the same gap it just found.
Let a person correct it. Extraction gets things wrong, and a list people distrust gets ignored. The correction has to change what happens next time, rather than dismissing one card and waiting for the next.
How we handle it
DoBox reads conversations across Gmail, Outlook, Slack, Microsoft Teams, and Google Chat and pulls the work out of them: requests, deadlines, follow-ups and the commitments you made, each arriving as a task with the thread attached before you open it. Tasks can have an owner, a due date, priorities, labels and comments, and you can reassign whenever.
We built correction in. Rejecting a suggested task and saying why creates a standing feedback rule that shapes later extraction, so “a quote confirmation isn’t a task for us” is a sentence you write once. Those rules sit under DoBox settings where you can read and edit them.
Delivering the commitment is a workflow. You pick what starts it. A DoBox event can, whether that’s a task created, a label added, a comment posted or an approval granted. So can a message arriving on any connected channel, a new contact, a Brain page, a cron schedule, or you pressing run. The schedule is how a workflow checks a board each morning for what’s due. From there it can branch on a condition, ask Claude for a judgment call mid-run, draft the reply, update a connected tool over MCP, and park anything that shouldn’t send itself in an Approvals section. It can read your Brain before it writes, so a follow-up that owes somebody a number goes out with the number you actually quoted.
A promise made out loud on a call reaches us too, through the notes email your notetaker sends. We don’t transcribe meetings ourselves. We read that email like any other message and pull the commitments out of it, so the recap you were going to skim becomes tracked work. Meeting follow-ups covers that path.
Pointing software at a mailbox is a decision your security team will want to review, and the security page has the detail. We hold SOC 2 Type I as of July 31, 2026, with Type II underway, and CASA Tier 2.
Which one your team drops
- Pull the last five items that got dropped. Was each one a reply you never got, or a promise you never kept? The answer usually skews hard one way, and it tells you which kind of product to shop for.
- Check whether those threads were quiet. A dropped item sitting under a cheerful “thanks!” would never have tripped a no-reply timer.
- Count where the promises were made. If many of them happened in chat, an email-only tool leaves those untouched however good it is.
- Ask what happens the second time. A nudge reminds a person to do the work again. A workflow does the routine part of it and asks for your input only where a person has to decide.
Frequently asked questions
What tool automatically follows up on emails I haven’t replied to? Gmail’s nudges do a basic version at no cost, and most email assistants sell a better one. For messages where the ask is buried in a long thread, the useful version is extraction: read the message, turn the request into an assigned task with a date, and let a workflow handle the routine reply. That’s what DoBox does across Gmail, Outlook, Slack, Microsoft Teams, and Google Chat, so a request that arrives in chat gets the same treatment as one that arrives by email.
How can a team track commitments made over email and chat automatically? Something has to lift the commitment out of the prose and turn it into a record before anything can track it. That means software reading the messages themselves, creating a task with an owner and a due date, and giving people a way to reject a bad extraction so the next one improves.
Can this+that replace our sales sequencer? Closer than you would expect on the mechanics: a scheduled workflow can loop over a list, work out who is due a touch and send it. What a sequencer adds is enrollment management, reply-rate reporting and deliverability work, and rebuilding those is a project rather than a configuration. If you sell by cadence, run both. The commitments your reps make inside those threads are the part a sequencer doesn’t see, and catching those is what we built.
Our problem is lead follow-up. Which row is that? Usually both, in sequence. Speed to lead and qualification are the first row, and a dialer or a sequencer is the right purchase for them. The revised quote you promised that lead by Friday is the third row. Lead routing covers the inbound side.
Does the whole team have to change how they work? No, and that’s the point of reading the messages rather than asking people to file them. Tasks appear from the conversations people are already having. You still open the board. What you don’t have to do is type the work into it.
If what your team loses is the promise made inside a thread rather than the reply that never came, that’s what DoBox was built to catch. Get started with this+that and point it at the mailbox and the channels where the promises get made.