The B2B Podcast Index
Index
All categories
MarketingSalesSaaSFinanceHROpsLeadershipCustomer SuccessAI & DataProductStartups & FoundersRevOpsEngineering & DevTools
MethodologySubmit
Best of:MarketingSalesSaaSFinanceHROpsLeadershipCustomer SuccessAI & DataProductStartups & FoundersRevOpsEngineering & DevTools
An independent project byFame
SearchBest episodesGuestsInsightsMethodologySubmit a podcast
Index/Marketing/Email After Hours: The Podcast for Email Senders
Email After Hours: The Podcast for Email Senders artwork

No-Reply Email Addresses Are Costing You Customers

Email After Hours: The Podcast for Email Senders · 2026-07-02 · 36 min

0:00--:--

Key moments - from our scoring

Substance score

61 / 100

Five dimensions, 20 points each

Insight Density12 / 20
Originality11 / 20
Guest Caliber14 / 20
Specificity & Evidence11 / 20
Conversational Craft13 / 20

Liam Aranda, head of on-demand generation and lifecycle marketing at Zapier, discusses why brands should retire no-reply email addresses and embrace genuine two-way communication. Reply inboxes contain a goldmine of actionable data: customer service problems, sales opportunities, complaints that signal engagement, and customer advocacy moments. The challenge isn't identifying value - it's managing volume at scale. For high-volume senders (millions of emails), Aranda shares how AI agents solve this. She describes building deterministic and agentic systems at Zapier that automatically triage replies into categories (support, sales, praise, complaints) while maintaining human oversight at critical decision points. Rather than one person drowning in an inbox, rules-based and AI-assisted workflows route complaints to support teams, flag sales-qualified inbound to sales with context (pulled from account activity), route praise to marketing channels for amplification, and flag edge cases to humans. The key insight: trust AI to help, not to act alone. Hallucinations and legal liability are real risks. For B2B operators managing email programs, this episode clarifies the business case for reply addresses and offers a practical starting point - audit your current reply inbox, identify your highest-volume reply type, and build a single workflow with AI triage and human review before scaling.

Key takeaways

  • →Unmanaged reply inboxes represent millions in missed pipeline, including outright sales requests that go unanswered because no-reply addresses prevent responses.
  • →AI agents can now automatically categorize incoming replies (support, sales, praise, complaints, OOO) and route them to the right teams, solving the scale problem that made reply-to addresses impractical before.
  • →Complaints and negative feedback in reply inboxes are engagement signals - responding with empathy can convert angry customers into advocates and surface product feedback.
  • →Every AI automation handling reply inbox triage must maintain a human-in-the-loop checkpoint, especially for sales and outbound responses, to avoid AI hallucinations that create legal or brand liability.
  • →Starting a reply management system requires auditing your current inbox to identify the highest-volume reply type, then building a single workflow with AI classification and human review before expanding to other categories.

Guests

Liam Aranda

Topics in this episode

ZapierAgentic workflowsEmail authenticationCustomer service automationNo reply email addressreply to inbox managementemail deliverability best practicesmailbox provider requirementsNo-reply email addressesAI agents for inbox managementReply-to inbox triageSales lead routingCampaign MonitorCustomer.io UnpackedDeterministic AI classification

Questions this episode answers

What kind of legitimate customer messages are going unanswered in no-reply email inboxes?

Sales-ready customers trying to buy but hitting error screens, customer support issues unrelated to the marketing email (like account lockouts), questions about new features, trial requests, and direct praise or feedback - all of which get lost because no-reply addresses prevent any response.

How does Liam Aranda use AI agents to manage replies at scale without overwhelming her team?

She built both deterministic and agentic systems that automatically classify incoming replies by type (customer support, sales, praise, complaints, out-of-office). Praise routes automatically to Slack channels; sales and support route to teams with human review gates; unknown replies flag to her team in Slack for decision-making.

Why is a human-in-the-loop critical when building AI systems to handle reply inbox automation?

AI can hallucinate - generating false promises like 'this plan is free this month' - which creates legal and brand liability, especially on branded email accounts. Humans must review at least the first instances of each automated response type, and all sales and outbound automations, to catch errors before they reach customers.

How should teams respond to complaints in their reply inboxes to turn them into opportunities?

Respond with empathy and curiosity rather than defensiveness; ask what specifically disappointed them and show you're reading their feedback. Aranda shares an example of responding to a harsh redesign complaint that led the customer to clarify what feature they wanted back - which the team then restored and recognized publicly.

What's the first step for an email team that has never managed their reply inbox intentionally?

Audit your current reply inbox to see what types of replies you're receiving most frequently, then build a single workflow starting with that highest-volume category, add AI classification, and maintain human review before scaling to other reply types.

What our scoring noted

Our reviewer’s read on each dimension, with quotes from the episode.

Insight Density

12 / 20

The episode delivers practical insights about reply-to inbox management and AI triage workflows, but a substantial portion consists of conversational filler, anecdotes (Chris the newsletter critic, lobster event story), and repeated framings of the same core idea. The actionable substance - segmenting reply types, building AI agents incrementally, human-in-loop validation - is concentrated in the latter half; earlier segments dwell on terminology preferences and general principles without novel specifics.

You're losing customers and sales if you don't.
I would say that's a great question. I would, uh, take a look at your inbox today... Start with that one. More than likely, it'll probably be either on the, uh, like customer service, um, support side, or potentially the sales side.

Originality

11 / 20

The core argument - that no-reply addresses waste customer feedback and pipeline - is well-established industry wisdom, not contrarian. The AI-triage workflow Liam describes is practical but represents standard contemporary thinking: classify replies, route them, add human checkpoints. The framing around complaints as engagement signals and the mention of Microsoft's 2023 bulk sender requirements are solid but not novel. No first-principles rethinking or genuinely unexpected counterargument emerges.

Today's best practices will be required in the future. Uh, it might be a year from now, it might be three years from now.
Trust AI to help you do your work. Don't trust AI to do all of your work.

Guest Caliber

14 / 20

Liam Aranda is a legitimate practitioner - head of on-demand gen and lifecycle marketing at Zapier, with prior roles at Campaign Monitor and Customer IO. She has shipped real systems, managed reply inboxes at scale, and speaks from operational experience rather than theory. Her perspective on AI triage, dealing with complaint management, and sales-lead routing is grounded in hands-on work. However, she is not a founder or C-level operator at a major brand, which limits the caliber slightly.

I've led Lifecycle and email programs across zapier campaign monitor
we have an agent that goes through and looks at our inboxes because we've got a couple for like our newsletters events

Specificity & Evidence

11 / 20

The episode lacks hard numbers, metrics, and concrete examples. No data on reply rates, conversion lift, or pipeline value from reply-to inboxes is cited. The 'Chris' newsletter complaint story and lobster event anecdotes are vivid but not generalizable evidence. The AI workflow is described in process terms rather than results: no mention of accuracy rates, volume handled, or quantified time savings. The reference to 'millions of pipeline' is asserted without specifics. Microsoft's bulk sender requirements are mentioned but not detailed.

I'm going to guarantee that there's millions of pipeline. No, um, I don't know if there's millions, but there's definitely pipeline in there.
we're sending out millions of emails if you will

Conversational Craft

13 / 20

The hosts ask solid foundational questions and allow Liam to develop her thinking, particularly on AI implementation steps and human-in-loop governance. However, follow-ups are often softball or accept claims without pressure. When Liam says 'millions of pipeline' are being lost, the hosts don't press for evidence or specifics. The 'email morality alignment' closing feels like mutual affirmation rather than productive tension. There's minimal disagreement or challenge; the interview feels like a friendly conversation with a guest who shares the hosts' worldview rather than a rigorous interrogation.

So what specific use cases have you been tackling and then what have you learned that maybe surprised you about that?
I was just about to say, I think that, like, what you were describing in terms of like, oh, certain reply types are being tackled already, some aren't. Is that for senders who want to start?

Conversation analysis

Computed from the transcript - who did the talking, and the words that came up most.

Share of words spoken

  • Speaker A74%
  • Speaker B15%
  • Speaker C11%

Most-used words

email34reply31customer23today19support17build15sales14start13team13inbox12human12send11sure11hours10best10back10

Episode notes

There’s probably a treasure-trove of pipeline and praise getting dusty in your no-reply inbox. In this episode of Email After Hours , Guy Hanson and Danielle Gallant sit down with Leah Miranda, Head of Demand Gen and Lifecycle Marketing at Zapier, to talk about killing one of the most destructive “best practices” in email marketing: the no-reply inbox. For years, marketers have relied on no-reply addresses to avoid managing inbound email at scale. But as mailbox providers place greater emphasis on engagement signals and customer experience, that approach is quickly becoming outdated. Leah explains why reply-to inboxes are no longer optional and how ignoring replies could be hurting both deliverability and customer relationships. She also walks through how she builds AI agents to create the best customer experience while keeping the human touch. If you're still sending from a no-reply address, this episode may completely change the way you think about email. About Leah Miranda Leah Miranda is Head of Demand Gen and Lifecycle Marketing at Zapier, where she leads customer journey, lifecycle, and demand generation strategies that drive engagement, conversion, and long-term growth.

Full transcript

36 min

Transcribed and scored by The B2B Podcast Index.

Speaker A: Foreign.

Speaker B: Welcome to Email After Hours by SenderScore. Powered by validity.

Speaker C: We're your hosts. My name is Guy Hanson.

Speaker B: And I'm Danielle Gallant and this is Email After Hours. Today we're tackling a topic that feels long overdue and that is the potential death of the no reply email address. So in the past it was recommended as a best practice that senders use valid reachable reply addresses, but it was perhaps one of the least adopted best practices in email. Kind of got forgotten about. But now, thanks to some, um, updated mailbox provider recommendations, the tide might be turning there. So we're going to get into why on today's episode.

Speaker C: That's right. And joining us, we've got Liam Aranda, head on demand gen and Lifecycle marketing at Zapier. Uh, and a self described email geek who builds high performing teams and scalable growth strategies. Leah has led Lifecycle and email programs across zapier campaign monitor, uh, and has recently spoken at Customer IO Unpacked about using AI to triage reply to inboxes. So absolutely the right person to have with us today. Leah, welcome to Email After Hours. It's really great to have you with us today.

Speaker A: I am super excited. I'm going to wave. I like to wave all these things. M. Uh, yeah, I'm super excited. I've been a longtime listener, fan, uh, of the show. So you're making dreams come true here on a Friday.

Speaker B: We love that. And really, before we get into today's topic, a little starter for us, a little icebreaker. Your LinkedIn headline says, I'm just a girl standing in front of a CEO asking them not to call it an E Blast, which is maybe my favorite headline so far on LinkedIn. And so question for you. What is the worst email terminology that you still hear in the wild?

Speaker A: I. I don't know if it's worse and maybe it's a phrase. Um, I bought a list. Um, or can we buy a list? I think this is always just sends me into a panic. Um, I do get a little persnickety about there being a dash and in, in email. Um, I just, you know, it's a camp that I'm, I'm, I'm living in. In a hill I'll die on. It's. I think nothing is more jarring though than an E Blast. And hopefully you're a fan of Notting Hill and understand the reference. But uh, yeah, anytime someone says blast E Blast, it just, it is so descriptive of what they're trying to do and accomplish that I'm like, no one has ever wanted that for a customer. No one has woken up and feel like, you know, we should just blast some people.

Speaker B: Just blast it out today, right? No, you should just let it rip. No.

Speaker C: Okay. I think that sets the tone straight away. So, yeah, let's start with the big picture. I feel like we spent a lot of time over the last couple of years especially talking about why the use of no reply at really needs to be retired. So from your perspective, why should brands actually want their subscribers to reply to their emails?

Speaker A: It just feels common sense. Like you would never send somebody and be like, I mean, a lot of emails are like, we launched this thing or we did something cool and it almost invites them to ask a question or some sort of follow up. Um, or if it's like an automated message, uh, they did a form or, you know, they realize, oh, I actually forgot to add something into just feels like the start of a conversation. And it's hard to like start a conversation and immediately turn around your back to someone. Like, that would be an odd thing to do. And that's, that's really what that no reply does. And so, you know, for me, I think it's, it's a place where you can, it's a treasure trove, if you will, of just information, um, and people reaching out. You know, you have the errant person of like, I don't like this redesign and that's fair. Um, but then you have that customer who's like, oh, I'm interested in this product. And it's like, how many times have people, like, without even knowing about it, responded to a no reply email of like, I would like to give you my money. Um, and then it just, it goes nowhere. And that, that makes me sad for the business and the customer because they

Speaker B: were trying, we, when we were having our kind of internal call before to talk about maybe a great topic for today and getting onto this one, we discussed our own reply, like no reply or like mailboxes that we've seen these, um, responses come into. And a treasure trove is a perfect way to put that. I think now on paper, switching from like no reply at to reply at addresses sounds simple. But for higher volume email programs that are sending tens of millions of emails and potentially generating thousands of replies, the inbox can be overwhelmed very quickly and there can be a lot kind of backed up in there. So why is the switch easier said than done?

Speaker A: I think it comes down to like the management of it. I think back in the day, like when we think about if you were uh, being like a good little email marketer and you had a reply. You, unfortunately, and I think we maybe even talked about this before. Like we had to manage those replies and those replies, like it's a treasure trove. Sure. But there's some, there's some weird things in that, that inbox. Um, I remember, uh, being a Campaign Monitor. I think my biggest fear was like seeing some sort of like, image included. Um, and it was just like I was like, I don't know what I'm about to click into. I don't really. Or you have, you know, like you've got singles in your area and you're like, I didn't really need to know that. And those are just like, you know, the mild versions. And so, you know, it was tough, you know, like you switch and you really wouldn't feel that, like, was this worth it? Because I've just spent three hours going through. When you're such a high value sender, I've spent three hours going through really uncomfortable replies and emails. Um, but I think if I could, like, why, I think it makes it better today. And I'm going to say the fun word, AI agents, agentic. Um, and that's what makes I think the switch a lot easier now for companies versus having someone or that poor lonely. Because it usually happened after hours. Like it wasn't like a number one thing. You'd be like, oh my gosh, it's been three months since I checked that inbox. I should probably go check it. At least that's what I used to do. Sorry to my coworkers at Campaign Monitor. You now know my horrible secret.

Speaker C: In your customer IO talk. You describe the reply to inbox as something of a goldmine. But a lot of teams, a lot of email teams are still ignoring it. I guess the next question is what kind of messages might be sitting in there. We're probably talking about complaints and unsubscribe, but sales opportunities, even praise. And I guess we start with complaints. I think they're particularly interesting because marketers often perceive them as a negative. But at the end of the day it's, it's an engagement signal. It's like the other side of the engagement coin where you got clicks on the one side and complaints on the other. But I did say to Danielle, I introduced her to a new expression, but back in the day we used to refer to reply to complaints as flames because these are people who aren't just hitting a button, they're really mad at you, they're mad at you to the point that they're actually going to write a little letter and tell you why they're mad. So is that still a good thing? Can that still be useful customer experience feedback?

Speaker A: I mean, I think so. Um, I would love to do a study on people who do this. Um, I don't think I've ever once gotten an email. I may have been miffed about it, but I've never replied to it. So, so I would love to know just like the whole, the ethos, psyche of that reply, however, it hit something. It hit a nerve. It hit. And I think sometimes, like, especially, um, I'm m going to age myself a bit. You know, we kind of grew up with the Internet and then, you know, we realized, oh, we could say things almost into the void and we see that come to life in those sort of. I do like that flame. Um, you know, we get flamed, if you will. Uh, uh. But I think what's great about it is I always like to respond to them. You know, we, we, we respond to, to a certain degree. Um, um, I, I, you know, don't flame back, I guess is what I would caution everybody.

Speaker B: Like, how dare you? Absolutely dare.

Speaker A: You maybe write it in like, you know, your, your AI agent and you, you yell at your agent and then say, can you make this now a professional, um, response. But I always like to, you know, we, a famous one. We redesigned our newsletter and someone was not happy about it. They were just angry and they wrote a very long email. I think it was longer than the actual newsletter. And so what we did is I responded, it was like, hey, we're going to call this person Chris just for fun. It's like, hey, Chris, like, thank you so much for your feedback. I hear that you don't like the redesign. I'd be curious to know what about it. Didn't you, like, is there a section that we removed that you weren't, that you were expected to see? Um, and then what was great is Chris respond. He was like, oh, I'm sorry. I guess I didn't, I didn't realize someone was going to read this. And I was like. And so I took someone who is like, really sort of like, miffed at us. And you can take that relationship even if it's a complaint, you know, like, treat it with empathy, go at it with curiosity. Like something, something happened that they wanted to engage and if you can, like, he was like, oh, well, really what I wanted to see was like this section. I was upset that like this section removed. I was like, Great. We can look to put it back in, in the next one. And we even did like a little shout out to Chris. Um, and it was fun and he was excited about it. He responded. And so, you know, we took a kind of a negative situation, turned it positive and that could be said for anything. You know, if someone's having, um, I think a big one that we see often is the complaint is, has nothing to do with maybe the message you sent. It's a customer support question. They're like, I've been logged out of my account for months and I can't get anybody on the phone. And it's like, that's a way for you to step in and be like, hey, well gee golly, let me, let me help you out. You can service it up to your support, uh, team, get them back on the right track and you take something that could have been a potential turn for an account and made it into a positive relationship. So that's why I think like it's worth going through and reading the flames and, and responding to them with empathy.

Speaker B: Yeah, it's funny, right? Because just that is like the Internet in a nutshell, right? If people think that they are screaming into a void, they will scream or say something like they will express themselves. And a relationship can so easily be built with, with the, with just the option that somebody's reading it. But what about praise? I'm sure they're, I'm sure we're not getting praise. Well, maybe I'm wrong. My assumption is that we're not getting praise in there in those reply addresses as much as maybe the flames or the feedback or the customer service emails. But they do happen for sure. And we all understand the great power of customer advocacy. So. So how should teams identify and then use those positive replies?

Speaker A: Yeah, I think, um, you hit on like the first one is the most ideal, is that customer advocacy. And you say, oh my gosh, I've been a long time fan. I don't get that much fan mail. I'm going to be really honest. But we do like every. How many people, like, especially when you work at some really fun brands. Um, like I've been using Zapier for like years and I love this. And you did this new feature and it's like, okay, cool, um, let's hand that off to our customer advocacy. Like could we either, you know, do a story either for, you know, our socials. Could it be a larger brand story if it's a bigger account? Um, or can we just send that person some swag you know, um, it's, you know, it's taking that and connecting with the customers I think is really great. The other side of the praise is when people, you know, comment on a marketing campaign that you did, like, that's exciting. And it's, you know. No, for most people, most teams, marketing campaigns aren't built in like a single person. It's. You've got your ads team, your brand team, your product marketing. Um, they might not be like a 50 person team on each, but, you know, it's a group of people that are bringing this stuff to life and it's always great to throw those in public channels. Um, so that's something I love to do with the team is if we launched something or did an event and we got some really great positive feedback, I throw that out and you know, that I'm definitely like tagging or, you know, CMO and our CEO and be like, hey, like, we did this really cool thing. Um, and that's just, you know, I think those positive vibes kind of help with the flames, if you will, out. And uh, yeah, I think sometimes they just don't get surfaced or sometimes we, we read them in the inbox and we're like, oh, that was nice. And we kind of move on. And it's like, it's, it's a nice pause to say, hey, we're doing something. We're connecting with people. It's not just us, like behind screens pressing buttons.

Speaker C: That's right. And especially when it's like a little unexpected, it's even nice sales interest. Feels like it could be the big missed opportunity here. Do you think there's potentially a load of pipeline, you know, unexpected pipeline that's quietly rotting away in unattended inboxes?

Speaker A: Yes. I'm going to guarantee that there's millions of pipeline. No, um, I don't know if there's millions, but there's definitely pipeline in there. Or, um, I would say maybe even like pipeline adjacent. People are just, uh, you know, we see that sometimes with, with Zapier where people are reaching out and they're like, hey, I'm just kind of curious. I want to do some sort of trial. I'm not sure which trial to use. Um, and you know, if that goes unanswered, it's tough. You know, they may have trial at the end, they may have figured it out. But it's also a part for us to be like, hey, uh, I know you want a trial, but we'd actually love to build like a custom trial for you. And that's, that's where it's like pipeline adjacent. It's like hey just a little nurturing and we get to a uh, potential large deal size and then you will have those errant people of hey I'm trying to buy this and I can't. And the amount of those type of messages especially if you're in um A, D2C um even B2B uh especially like in the SaaS, um like industry you'll have a lot of people do that. They're like hey I'm running up to this error screen. I, I'm trying to buy this, I'm trying to self serve by this and if you get enough of those like that's a significant amount of sales revenue um and self serve revenue that you're just missing because no one's responding. Now you could argue most people will find a way and go to like your help, maybe your chat on your website but I mean at a certain point I don't know about you but if it's difficult to like purchase something I'm done because there's probably another product that was on my list that I'm like oh like if they're going to respond to me in a quicker way like that's what I'd prefer. And so I mean yeah, I would highly encourage you if you have an inbox that you haven't checked in like the last like three months, go in there, go check, go see what's in there. Um, and then uh, I think from that is you can start to build those automations. I think it's tough to have this um, just like as a single person's job because you'll forget about it or if a team is doing it they'll forget about it too. We're human. But I think um, the great thing like I said is that you can build agents around this. You can build rules even if you're say not agentic but you can build rules within your inbox to say is it an out of office reply? Yeah, you probably don't want to deal with those. Um, if it is something that you can um, have an agent look at and say hey this was a complaint, is it support related? Let's send it over to support. And that is, I don't know, I think one of the coolest things about AI aside like the coolest things about how we're working um, in today's age and the cool things that we can do now and I don't know, I get a little excited and nerdy about it.

Speaker B: Well, speaking of which, that's a perfect segue because in your role now you have been building response handling systems using AI. So what specific use cases have you been tackling and then what have you learned that maybe surprised you about that?

Speaker A: Yeah. So we have an agent that goes through and looks at our inboxes because we've got a couple for like our newsletters events, uh, um, and uh, for any sort of like high touch events for instance, uh, we do this really great shout out to my events team. Um, it's like signature outpost events where we'll fly people out. We recently took a whole bunch of people to the Cape, um, and they had lobster. It was great. We talked about automation AI. Ah, automation AI, lobster. Those should all always go together. So you know we have, you know we can set up an agent where someone can reply and say hey, uh, you know I have a question about an agenda. And we, those are pretty easy questions that we can answer. And we say hey, great, here's the information you have or here's the agenda that's going to be going out. We don't really need someone to really manage that. So that's a really nice thing to do. Or the agent can say hey, I don't actually know what this person's asking. And then it'll flag us in Slack saying hey, you had this email. Can someone on the email team reach out to them? And I like this setup because it keeps a human in the loop. It gives them guardrails. And so um, when building, if you're going to build this, I think there's two ways that you can do it. One is truly just agentic. You have an agent, it can make decisions on its own. You can add in the, if you're unsure, send me like a message or send me an email, ping me somehow. Or you can have something that's maybe a little bit more deterministic. And so that's what I use for our large send. So like our newsletters, you have to imagine we're sending out millions of emails if you will. So how do I go through and make sure that it is filing them the correct way? And so that's where that deterministic and so I can have AI basically label it, say hey, I'm pretty sure this is going to be customer support. It'll go down the customer support realm. Um, if it is praise, it'll go down the praise way. If it's an unknown again, I always have that human in the loop. So we can go in and Review it. Um, if it sales, that is the one that I actually always have a human in the loop. Because the last thing we want to do as marketers and be like, I sent you a sales lead. And they're like, leah, this was not this, please don't ever send me these type of leads again. Uh, we don't want that tension, you know, that's never great. So, so I always go through and say, okay, this is being marked. AI thinks this is a sales lead. Is it truly a sales lead? Are they just asking like, hey, I need to, I, um, want to buy this account. Is that something that we could do if it's a self serve model? Or is it something that, what I like to do is say, oh yeah, this is a sales lead. And then I have another agent that sits on top of that. Once I say yes, this is a sales lead that will go through. It'll look at that person's account if there's any activity either from that person or the like the top level account, and then put together a nice little package for the sales team so we can say, hey, they reached out from this, here's their question, here's a little bit about their account. Please reach out. And I think that that's a, a, uh, really nice way to support your team. Make sure that they have context without working, I think 80 hours a week.

Speaker C: Gotta tell you, I, I, I was counting though. You mentioned three times the importance of a human being in the loop. And that's important, isn't it? Because it feels like for it to work, there's got to be a balance between letting AI do the heavy lifting on the one side and um, the fact that human judgment is still essential on the other. So what does that look like in practice? What have you learned?

Speaker A: Yeah, I think eventually, um, and maybe I've just had too much AI Kool Aid lately. I think eventually I'll get to the point that I'm like, yeah, AI can definitely handle this. I don't think we're anywhere near that right now. Um, I am optimistically cautious is probably the best way I can say that. So I think having that human in the loop and that judgment also just helps feed the AI as well. And so how I've got it set up is that it will basically take a look and say, hey, no, we actually got this wrong and here's why. And I can put in that why. And so it'll then begin to feed that model and so it'll become smarter and smarter. So if you build something out like this. Have a human in the loop for every single step. Build out that model for you, and then you can start taking it back a little bit. Um, so for instance, ah, praise is the one flow that I don't have, um, any human in the loop. It'll just go into, uh, we've got a praise like Slack channel. But for everything else, I think if you're not sure about the model or if you have any sort of responses that you are like, going to be automating from that. Have a human in the loop. Because the last thing you want is for AI to hallucinate to say, yeah, you can have this plan and it's free this month. And so like, and it's like, oh, my gosh, wow. It's like on a zapier branded, you know, email account. Um, and legally binding. Like, we don't want to do that. So trust AI to help you do your work. Don't trust AI to do all of your work is at the end of the day where I tell my team to focus on because, yeah, it's one of those things that if you get it right, it's wonderful. If you get it wrong, you get it really, really wrong. And that is what should keep us up at night. Sorry, the tone got really dark there for a second.

Speaker B: No, it didn't. I was just about to say, I think that, like, what you were describing in terms of like, oh, certain reply types are being tackled already, some aren't. Is that for senders who want to start? Whenever we talk about AI, which is always, uh, we're all on the Kool Aid, um, I think that it becomes really overwhelming to be like, I want to do this, but what actually do I do? What are the steps first? Where do I begin? So for people who want to start managing replies more intense, intentionally, what should they be doing first? Is it starting with a certain type? Like, what's step one?

Speaker A: I would say that's a great question. I would, uh, take a look at your inbox today. So imagine we don't have any AI set up. Um, take a look at your inbox today. What type of responses are you getting the most of? Start with that one. More than likely, it'll probably be either on the, uh, like customer service, um, support side, or potentially the sales side. So let's say the customer support side, you could, um, build up just a single flow, get an email in. There's an AI step in there to say, can you qualify? Is this person looking for customer support? Then you can put like, Common phrases that you see from the kind of the research you just did and say, they're looking for, I can't access my account, I'm having trouble logging in, my payment failed. Those things that might come up, add those in and then, um, add in the human in the loop for that and say, hey, we've identified this is a customer support and say if you wanted to have it draft a message and say, hey, I was going to send a message, here's what the message would be. And then have that human in the loop approval step. So what it might look like for you is you get like a teams or a Slack notification saying, hey, here's their message, here's how we've defined it, and then here's my reply. And at that point you can say, yes, this is correct. Um, or you could potentially edit that message. And that's how I would start. It's just a single use case, a single kind of reply. And even with the, uh, reply, I mean, give your AI an example of what a good reply is because usually if it's a payment, here's the link that they need to go to. If it's a more serious complaint, like here's our support hotline that you can call 247 or I'd encourage you to chat. I've also forwarded this onto our support team and you can have those kind of like base templates in there that it'll start to pull from. And then once you get confident in there, go to that next most common one, which, like I said, might be sales, might be praise. And then you can start to like build branches if you will, if you have like a deterministic workflow or just add on new agents. Um, and that would be my recommendation. Don't try to, yeah, don't try to build the example that I gave first because that's going to be overwhelming. The example I gave is months and months of iterations. We've added things on. Um, and also I would say expect things to be wrong at the beginning. I think it's like if you've ever had any sort of like ChatGPT scale or Claude skill at the beginning, it's just a lot of like, you've got to put in a lot of input into it before you start getting some semi good, uh, responses. And then suddenly around after a couple of months you're like, oh yeah, this is really getting it. It just needs that long. So think of it as like you're onboarding a new employee and you wouldn't expect A new employee just to know everything off the gate. Like, you're gonna have to do some coaching, really.

Speaker C: So Claude's gonna have to watch about a week's worth of videos.

Speaker A: Exactly. And some sort of harassment training, I assume.

Speaker C: Let's talk about deliverability for a moment, because we love deliverability almost as much as we love AI. And what I'm thinking of specifically here, Leah, is round about this time last year, Microsoft belatedly launched their bulk sender requirements. And part of their guidance was an explicit statement that they specifically recommend that senders should use a replyable address, which was really interesting. Um, especially because we all know that today's best practices have a bit of a habit of becoming tomorrow's new bulk sender requirements. So is this something that we should be paying attention to? And, you know, does operating a real reply to inbox align with what mailbox providers increasingly are wanting to see from legitimate senders?

Speaker A: I mean, 100%, I, uh, think, uh, especially, like, if we could, like, go back in time. Do you remember gdpr? That was a thing. Um, yeah, you remember, it was. It was always interesting to see, like, who was freaking out about it. And like, without passing any judgment, we won't say any names here, but like, the people freaking out about it, it was kind of like. But like, you do need to opt in. Like, are you emailing people who didn't want to be emailed? Like, like that? Sort of. And so, yes, I 100% agree. I think this. Today's best practices will be required in the future. Uh, it might be a year from now, it might be three years from now. Um, my guess is the EU would do it first, which is great. But build those best practices in now. Um, and like, this is one of those, uh, I think you don't want to be caught in a GDPR situation of like, well, we have just been emailing and buying lists and that's not cool. Um, if you're doing that now, stop it. Um, it's 20, 26, y'.

Speaker C: All.

Speaker A: That's not cool. Uh, and then for this, yeah, I mean, it should. I would be really surprised to see, like, really reputable, great email programs not have this. And if you are sitting there thinking, uh, I'm doing it all, but this is like, that's okay too. Like, these best practices come in. I would recommend taking and building out that reply too, and then also manage it. I think the other problem with this is it will become a, a standard eventually. Um, but if it's a standard and you have a Reply to email address and you're back at the beginning of this conversation where just no one's checking it. That's also not cool. And you know the, the email industry can't standardize whether or not you respond to replies. It's just whether or not you have that reply to address. So like bake those into your, your team, how you're structured. Like I said, if it's someone just handling replies one on one, great. If it's an agent and you have and tool stack, like great, build that out. But yeah, it always be. I mean I think email marketers worth their sort of weight if you will, are always forward looking, are always thinking because at the end of the day this comes down to the principle of uh, what's best for my customers, what's best for my prospects. Am I just, you know, if you're just blasting into the void, like gross. Sorry, there's so much judgment on this. Uh, um, I have like strong feelings. Like we're with you. Yeah, you can email like a lot of people and I'm saying like every email needs to be like, to like bespoke to five people. Like you can still send a mass newsletter but have that newsletter segmented, have it be dynamic content. I think if you are a marketer that you're like, yeah, this makes sense, you're going to be fine. Um, you're going to have those forward looking practices and this is just one of those. And I kind of am excited about making this. I would like to see the industry make this a standard. Um, I wish if we had the ability to make people actually manage the inboxes too because it becomes a void, if you will. So yeah, build those practices.

Speaker B: Today we had a uh, 2026, our litmus state of email report and it's really one of the things that we found or there were themes right. Among senders that are those top performers that are achieving the highest ROI from the email channel. And one of those things is that they are, well two of them, many of them is that they're focused on building relationships through much more targeted um, communications that catch the subscriber where they are in the journey. Like that is the personalization element. And so I wonder with this too, like does this uh, functioning monitored reply address kind of fit into that like moving kind of a little bit away from broadcast messages and more to this like conversational relationship building customer journey? I think it sounds like a pretty like an integral part of that.

Speaker A: I would think so. Uh, I think where we're going in The. We're in this, like, weird space right now where everyone can write emails and send emails. You can build them with AI and blast them all over the place. And that's cool. Not really, but, you know, it's what people are doing. I think what people are going to be expecting in the future is exactly what you just said. And I think the amount of tools that we have, uh, to. You know, when I think about my email campaigns that I'm sending out that are kind of those relationship buildings I'm looking at, okay, what product data do they have? What have they been doing? Um, you know, from a brand standpoint, what have they been doing on social media? Not in, like a creepy way, but, like, also, like, what. What site pages on our site have they been visiting? Like, how can that all influence that conversation? And we'd want that same love and respect for this inbox, this reply inbox. And so you can have. I mean, I'm, um, talking about, like, really advanced building, but you can then have somebody, like, say, it's a customer support thing. All of those things I just described that you would do on the front end, you could potentially do on the back end if they're reaching out. And then you can say kind of like the, the sales version of that, but for customer support. You could have a customer support agent take a look at their account, um, see some information, and then you are starting from, hey, I'm Leah from our customer support. How can I help you today? You're jumping in and already addressing their problems. And so, uh, yeah, I think that is where we should be going with the technology. And I think that's us using the technology for good, rather than I can now email, you know, millions of people a minute with AI. Like, that's. That's not fun.

Speaker B: Good for you guys. Should we be doing it, though?

Speaker A: I say this in, like, I say this, like, not saying that we have all the answers or that we're doing everything right here. Like, we screw up, we grab the wrong data points. Um, like. Like, that is natural, but I think it is the spirit of which you're trying to do it. You know, we're trying to have a conversation with somebody, not just be like, I see leaders like you doing. It's like, ugh, please don't ever send me an email that says, that guy.

Speaker B: And I have such good examples that we've talked about on the podcast before where people have reached out to us. Like, we loved your appearance at this conference that we didn't attend. They were supposed to attend, but we didn't and it just is just like little tiny more looking would solve it. But this has been such a great conversation. I am with you 100%. It sounds like our um, our email morality is kind of aligned in what we want to see and what we don't want to see. So kind of just to close it out today. If a brand is still using a, uh, no reply today, what's like a one liner that you would deliver to convince them to change to make the switch?

Speaker A: You're losing customers and sales if you don't.

Speaker B: Bam.

Speaker A: True managen and lifecycle person. So Pipeline is a little, you know, I'm thinking about Pipeline but I'm also, you know, like you're losing customers. At the end of the day you will lose customers if you don't have this and if customers are important to your brand, you should do this.

Speaker C: I think we will wrap up on that note because we can't beat it. Leah, thank you so much for joining us today and sharing your perspective on um, reply to inboxes. And man, we covered a lot of stuff today. AI triage, lifecycle marketing, why customer replies deserve a lot more attention than they normally get. If our listeners want to connect with you and learn a bit more about what you're doing, what's the best way for them to reach out?

Speaker A: Uh, LinkedIn. Yeah, connect with me on LinkedIn.

Speaker C: Absolutely. To our listeners. Be sure to tune in next time. In two weeks time, hit subscribe so you don't miss any future episodes. And um, absolutely don't Forget to visit senderscore.org for more great resources to help you become even stronger senders.

Speaker B: And Guy and I are always super proud of the guests that join us on email after hours and we're always looking for new guests to feature on future episodes too. So if you'd like to join us for an upcoming episode, drop a comment, message us too on LinkedIn or fill out the guest form that's linked in the show notes. Thank you for listening and we'll see you in a couple weeks. Bye guys.

Speaker C: Be sure to tune in next time and hit subscribe so you don't miss any future episodes. And don't Forget to visit senderscore.com for loads more great resources to help you become a stronger sender.

Speaker B: Uh, to all you sleepless senders out there, thank you for joining us after hours and see you next time

Speaker C: email after hours.

Related episodes across the Index

Other episodes covering the same guests and topics, from across The B2B Podcast Index.

  • More Agents Than Employees: How Zapier Disrupted Itself Before AI CouldTalking AI · on Zapier90 / 100
  • Building A Company of AI Super Workers - Tracy St. DicFNDN Series · on Zapier87 / 100
  • Gong's Amit Bendov on Powering Your Company BrainAgents of Scale · on Zapier83 / 100
  • The API of the Physical World: Sy Bohy on the Future of Connected HardwareHow I Grew This: Real Stories of Digital Growth · on Agentic workflows81 / 100
  • Ep. 29- How Integrations Impact Your Pre-Sales and GTM Team with Cristina Flaschen, CEO of PandiumPre-Sales Unplugged: Leadership Playbook · on Zapier80 / 100
  • The Artisan Approach to AI Governance with Karl Herbert Grabbi from Credo AIParadigm Shock · on Agentic workflows80 / 100

More from Email After Hours: The Podcast for Email Senders

All episodes →
  • Life After Opens: New Consent Rules and the Metrics Worth Trusting84 / 100
  • Stop Blaming Marketing: The Real Reasons Your Emails Land in Spam85 / 100
  • Gmail Changed the Inbox…Again: What Marketers Need to Know64 / 100
  • DMARC Updates Every Email Sender Needs to Know64 / 100
  • The Hidden World of ISP Relations: Reputation, Removals, and Inbox Trust66 / 100
Explore the best B2B Marketing podcasts →
All Email After Hours: The Podcast for Email Senders episodes →