
ArticlesWriting tools
The best AI reply app depends on whether your problem is volume or stakes
Four kinds of tool answer to the same name and barely compete. Email triage, sales sequencers, support bots, and the one for messages that matter.
By Samet Durgun · Co-founder of Subtext · 10 min read
· Updated November 14, 2026
In short
It depends on whether your problem is too many messages or one message that matters too much. I sort reply apps into four groups that barely compete: email triage (Superhuman, Shortwave, Agentys), outbound sequencing (Reply.io, GoExtrovert, Skywork AI), support and DM automation (Instant Reply, Vellum), and single messages where how they land matters, which is where I put Subtext, the app I co-founded. Most of what I say about the other tools comes from vendor material, not head to head tests.
Before you pick one
- Decide first whether your problem is too many messages or one message that matters too much.
- Pick only from that group: a tone checker will not clear your inbox, and a sequencer has no opinion on whether your message sounds cold.
I co-founded Subtext, one of the tools named here, and say so at the start of the post.
One message that matters too much?
Subtext asks what you want from it first, then offers up to three versions.
“What’s the best AI reply app” is a question with four different right answers, and the person asking usually only wants one of them.
I went through the category properly, partly because I build in it. Subtext is mine, an app that reads how a single message is likely to land before you send it, which is one of the four answers and not the other three. It shows up in the fourth group below, and you should weigh everything I say about it accordingly. The other three groups I have no stake in at all, which is most of this article.
The useful thing I found is that these tools barely compete with each other. People file a sales sequencer and a tone checker under “AI reply app” and they are not alternatives in any meaningful sense. I compared the AI tone checkers that actually check your tone separately. One is trying to move a pipeline. The other is trying to stop you sending something at 11pm that you cannot take back.
The four categories
Email triage is for volume. Superhuman, Shortwave, Agentys. The unit of work is the inbox, and success is getting to zero faster.
Outbound sequencing is for pipeline. Reply.io, GoExtrovert, Skywork AI. The unit of work is the campaign, and success is a booked meeting.
Support and DM automation is for coverage. Instant Reply, Vellum. The unit of work is the queue, and success is that nobody waits.
Relational calibration is for stakes. Subtext is the one I work on. The unit of work is a single message, and success is that it lands the way you meant it. One real example of it is just below the table, and another further down.
Only the last one starts from how a single message will land. The first three are logistics.
The table below sets out four kinds of AI reply app, with Subtext in the last row.
| Category | If your problem is | Tools the article names | What success looks like |
|---|---|---|---|
| Email triage | Four hundred emails | Superhuman, Shortwave, Agentys | Getting the inbox to zero faster |
| Outbound sequencing | Pipeline | Reply.io, GoExtrovert, Skywork AI | A booked meeting from the campaign |
| Support and DM automation | A support queue | Instant Reply, Vellum | Nobody in the queue waits |
| Relational calibration | One message that matters too much | Subtext | A single message lands the way you meant it |
Subtext is the author’s own product. The author has not run these tools head to head. Almost everything on the other tools comes from vendor material and comparison sites that run affiliate programmes, checked August 2026.
A table row says little about how a tool behaves, so here is one example of the fourth, on a request made up for this example: “I’m so angry at my sister, she bailed on mum’s birthday again and I had to do everything. Help me say something to her”. Subtext did not draft anything at first. It put the word Anger on the feeling, flagged “again” as a Generalization that could make the message a broad accusation if it went into the draft, and asked four questions. The first, “What do you most want from her with this message?”, came with three answers to tap, “An honest apology”, “Help going forward” and “Just to be heard”. An answer tapped there is context a general chatbot would need you to type out, for reasons I come to below. That is the app’s real output, unedited.
I'm so angry at my sister, she bailed on mum's birthday again and I had to do everything. Help me say something to her
That sounds exhausting, and it makes sense you're angry after doing it all yourself again. Since this is a heavy one, I have a few quick questions so the message says what you really mean.
- What do you most want from her with this message? An honest apology
- Did she give a reason for not coming this time? Yes, a weak excuse
- How do you want to come across to her? Calm but firm
- Is there anything else you want to say or ask her that we haven't covered?
Thanks, that helps. Here are a few ways to say it calmly and firmly, while making clear you want a real apology.
- Short and direct (Firm, heads-up): I'm really upset that you weren't there for mum's birthday and I ended up doing everything. The reason you gave doesn't really cover it for me. I'd like an honest apology from you. It's clear and firm and asks for an apology outright. Calling her reason weak could make her defensive.
- Hurt but open (Vulnerable, safe to send): I've been wanting to talk about mum's birthday. I ended up doing everything on my own, and it left me feeling pretty hurt and tired. I didn't really feel like your reason was enough, and I think I need you to acknowledge that. An honest sorry would mean a lot to me. It speaks from your own feelings rather than blame, so she's more likely to listen. The apology request is clear but gentle.
- Calm conversation opener (Constructive, safe to send): Hey, can we talk about mum's birthday? I was on my own with all of it, and that was really hard. I know you had your reasons, but it didn't sit right with me, and I'd feel a lot better if you could own that and say sorry. I'd rather sort this out between us than let it sit. It opens a conversation instead of an accusation and states your need clearly. The wish to sort it out together makes it easy for her to respond well.
I left out "again" so it doesn't turn into a list of old grievances. Want it shorter, warmer, or more confident?
Help going forward
Thanks, that helps. Here are a few calm but firm ways to tell her how it landed and ask for help next time.
- Short and direct (Firm, safe to send): Hey, I wanted to talk about mum's birthday. I ended up doing everything on my own and it was a lot. I'd really like us to share the planning next time. Can we sort that out together? States the facts plainly and moves straight to a request for shared planning, with no blame. The brevity could read as a bit curt.
- Warmer, honest (Honest, safe to send): Hey, I've been thinking about mum's birthday. I was disappointed you couldn't make it, and I ended up carrying all of it myself, which left me pretty drained. I know things come up, but I'd love for us to do the next one together. Could we plan it as a team? Shares your feelings with "I" statements and gives her an easy way to say yes, so it should land without putting her on the defensive.
- Clear ask, own words (Assertive, heads-up): Hey, can we talk about mum's birthday? I didn't feel the reason you gave for missing it was enough, and I was left to do everything. I'm not trying to start a fight. I just need things to be more even from here on. Before the next family event, can we agree who's taking care of what? I'd rather sort it out early than end up resentful. Says plainly that her excuse didn't sit right with you, which is honest but could make her defensive. The clear plan for next time balances that.
I didn't repeat what her excuse was, since I don't know the details. Want it warmer, shorter, or more confident?
Just to be heard
Thanks, that helps. Here are a few ways to be heard without turning it into a fight.
- Short and direct (Calm, safe to send): Hey, I need to tell you that I was really disappointed you weren't at mum's birthday. I ended up doing everything on my own, and it was a lot. I'd just like you to know how that felt. States the disappointment and the workload plainly without blaming. The ask is only to be heard, so it should be received calmly.
- Warmer, hurt but honest (Vulnerable, heads-up): Hey, I wanted to talk about mum's birthday. I missed you there, and the excuse you gave didn't really sit right with me. I ended up organising everything myself and I felt pretty alone with it. I'm not looking for a fight, I just want you to know how it was for me. Names the weak excuse and your feelings gently. The line about the excuse could make her defensive, so it's slightly riskier.
- Fuller, calm but firm (Grounded, safe to send): Hi, can we talk about mum's birthday? I've been thinking about it and I'd rather say it calmly than let it sit. I really missed you there, and I ended up carrying all of it by myself, which was tiring and a bit lonely. I'm not trying to start an argument. I just need you to hear that it mattered to me, and that it matters to me that we show up for mum together. Opens with an invitation to talk, owns your feelings and says plainly that you're not after a fight. It's firm yet warm, so it should land well.
I left out any request for her to do something differently, since you only want to be heard. Want it shorter, warmer, or a bit more confident?
Email triage with Superhuman, Shortwave and Agentys
If your problem is four hundred emails and a Tuesday, this is your row.
Superhuman centres on a latency promise that no interaction should take longer than about 100 milliseconds. Keyboard shortcuts, split inboxes, fast triage. Its own AI agents draft replies, and since Grammarly bought it in 2025 it sits in the same Superhuman suite as Grammarly. It optimises how fast you can process mail yourself. Somewhere around $30 to $40 a month depending on plan.
Shortwave goes the other way. It is Gmail only, and it leans on an assistant that can search years of archive in plain language, summarise long threads, and run saved workflows. Where Superhuman makes you faster, Shortwave tries to read the thread for you. Roughly $14 to $24 a month.
Agentys solves the part that stops most people switching, which is that you have to switch. It works inside Gmail, Outlook and Microsoft 365 as they already are, generating draft replies in your own trained voice inside the inbox you already use. About $17 a month.
My split is Superhuman if you like keyboards, Shortwave if you want the machine to do the comprehension, and Agentys if the migration is the blocker.
Outbound with Reply.io, GoExtrovert and Skywork
If your problem is pipeline, none of the above helps and neither does anything in the last two sections.
Reply.io runs autonomous outbound agents that research a prospect, sequence across email, LinkedIn, SMS and calls, sort the replies that come back, and put meetings on a calendar.
GoExtrovert replaces the fixed sequence with something that watches for signals, tracking prospect posts and generating suggested comments and messages in your voice, with a human approving before anything sends. That review step is what keeps a LinkedIn account from looking automated.
Skywork AI does research before drafting, reading documentation, filings and market material, then producing formatted cited documents. Skywork targets formal proposals more than cold email.
Support and DMs with Instant Reply and Vellum
If your problem is a queue that never empties, this row.
Instant Reply serves social commerce, so Instagram, Messenger and WhatsApp Business. Two details stood out to me. It transcribes inbound voice notes, which the company says are around 30% of WhatsApp inbound, and it runs several AI providers behind a failover so an outage somewhere upstream does not take the queue down. That second one is unglamorous and probably the reason people keep it.
Vellum is an open source personal workspace that spans desktop, mobile, voice, Slack, Telegram and email, keeping memory across all of them so a follow-up three weeks later still knows what was said.
Why general chatbots struggle with the fourth category
General models serve the first three categories well, because the job is throughput and a competent generic sentence is fine. The fourth, the one Subtext addresses, is where it gets awkward, and I’d rather be specific about why than hand-wave at it.
A general model writes the most probable next thing. For a status update that is exactly right. For a message where you are trying to hold a boundary without sounding cold, the most probable next thing is a wordy, over-apologetic, slightly performative paragraph, because that is what the internet is full of.
Three failure modes show up repeatedly. Models make implicit things explicit, so they say the thing you deliberately left unsaid. They smooth the rhythm out, so the result is coherent and characterless. And when the topic is conflict, they over-explain and over-apologise by default.
There is a second problem, which is effort. Getting a general chatbot to write a difficult personal message well means describing the relationship, the history, the tone you want and the thing you are afraid it will sound like. I looked at ChatGPT alternatives for writing, and what the big three do to your words. That is a lot of typing at the exact moment you are least able to think clearly. I wrote about the research on that at more length in what AI actually knows about the tone of your messages and in whether AI message rewriters actually work.
The fourth category, where I am not neutral
Subtext is my product, so read the next few paragraphs as an argument rather than a review.
Subtext starts from your draft, not from a prompt. You paste what you already wrote, or a screenshot of the whole thread, or a voice note where you talk it out. It reads the exchange around your line when you include it, then tells you how the draft is likely to land and which words are doing it, before it offers to change anything. It marks each version it offers “safe to send” or “heads-up”, and a tap shows a short line on what earned it, including what could be misread. Behind the mark is Subtext’s estimate, from 5 to 100, of how likely the message is to land the way you meant, not a measurement.
For messages that matter it stops and asks what you want to get across, as it did with the sister above. Holding a boundary and smoothing things over produce different messages, so it does not guess which one you meant. Subtext adjusts a draft with one tap instead of a prompt: Warmer, Shorter, More casual, Add humor, Add confidence. And you get up to three directions to switch between instead of one answer, which is a faster way to find out what you meant than describing it in advance. Since October 2026 Subtext also has Memory. It keeps short notes on how you write to one person, and writing styles such as one for work and one for friends, and every note can be read, edited or deleted.
A good reply app also tells you when no reply is needed. One example shows it, on a thread made up for the purpose: a LinkedIn connection, Daniel, praises a design studio’s portfolio, then nudges twice, once with “just checking you saw my message”, and promises to “check back in a few weeks”. Asked “is this even worth replying to?”, Subtext answered that it “looks like an automated sales sequence”, that nothing in it is personal or asks anything real, “so you don’t owe him a reply”, and that “a short, kind no thanks stops the follow-ups.” It offered three answers to tap, “Write a polite no thanks”, “I’m a bit curious” and “I’ll just ignore it”.

Asked to decline politely, it wrote “Thanks for reaching out, Daniel. It’s not something I’m looking into right now, so no need to check back. I wish you all the best with FocusFriday.” and said it had added “no need to check back” so the follow-ups stop. The last line was not clear to the writer, so that line went back with the question “I don’t get that part”. Subtext explained that FocusFriday is “the product Daniel created, as his profile headline says”, and offered the message with only that line changed, labelled “Closing line swapped”: “Thanks for reaching out, Daniel. It’s not something I’m looking into right now, so no need to check back. I wish you all the best.” Being able to ask about a line, and get one line changed and nothing else, is most of what makes a drafted reply your own.

Subtext works in 20+ languages, natively rather than translated. It never sends anything for you, which is deliberate. And the privacy posture matches the content. Nothing trains a model, conversations delete themselves five days after you last use them, or 90 days if you pin them, and deleting something removes it immediately. Notes saved to Memory stay until you delete them. Try Subtext in your browserDownload on the App StoreGet it on Google PlayScan it with your phone camera to install.
When not to use it. If your problem is four hundred emails, go back to the first section. If your problem is pipeline, second section. If your problem is a support queue, third. Subtext works on one message at a time, which makes it the correct tool for exactly one of the four problems on this page and the wrong tool for the other three.
A note on where these numbers came from
It changes how much weight you put on the rest, so I’ll say it plainly.
Almost everything in the first three sections comes from vendor documentation, vendor blogs and comparison sites that run affiliate programmes. That is not the same as testing. Prices move, and several of the figures above appear only in the vendor’s own material. The 30% voice note figure is Instant Reply’s own. The latency promise is Superhuman’s own. I have flagged them as such, so they do not pass for independently checked facts.
I have not run these tools head to head, and I would not trust anyone who claims they have across all four categories, because the categories do not share a success metric. There is no benchmark on which a sales sequencer and a tone checker can both score.
Which one
If you are drowning in email, Superhuman or Shortwave, or Agentys if you refuse to leave your current inbox.
If you are running outbound, Reply.io for multi-channel sequences, GoExtrovert if LinkedIn is the channel and you want a human approving each one.
If you are running a DM queue, Instant Reply for social commerce, Vellum if you need memory across channels.
If the thing keeping you up is a single message to a person who matters, that is the category Subtext serves, and I would rather say that plainly than pretend I arrived at it neutrally.
The mistake I would most want to save you from is picking from the wrong row. A tone checker will not clear your inbox, and a sequencer has no opinion about whether your message sounds cold. They answer to the same name and solve unrelated problems.
Think I have miscategorised a tool, or know one that belongs in a row it is missing from? Tell me on LinkedIn.
Samet Durgun is the co-founder of Subtext, an app that catches the emotional tone of your messages and rewrites them in your own voice. He’s based in Berlin.
Sources
Prices and features checked August 2026. Where a figure comes from a vendor rather than an independent test, the text says so.
- Superhuman, product and pricing pages.
- Shortwave, product and pricing pages.
- Agentys, product pages and the Superhuman comparison written by Agentys.
- Grammarly, “Grammarly Rebrands Company as Superhuman.”
- Reply.io, Jason AI product pages.
- Skywork AI, message generator guide, published by Skywork.
- Instant Reply, AI customer support comparison, published by Instant Reply.
- Vellum, best AI tools for lead capture, published by Vellum.
- Subtext, official site.
- Subtext, terms, privacy and your account.