Latest / After the Demo / Read the First Thousand Emails

Transcript

Read the First Thousand Emails · After the Demo

0:00
John Arndt: Somebody posted a rule this month. Before your AI sales rep sends its first thousand emails, a person reads every one of them.
0:07
Kate: A thousand.
0:08
John Arndt: Not spot checks. Every one.
0:11
Kate: Wait. Your AI sales rep. The thing you bought so you'd have less email to deal with.
0:17
John Arndt: That one.
0:17
Kate: So you hired a robot and got homework.
0:20
John Arndt: Hired a robot, got a very long first week.
0:23
Kate: A week? Who reads a thousand emails in a week?
0:27
John Arndt: Okay, a long first few weeks. I was rounding in the robot's favor.
0:31
Kate: And if I don't read them?
0:33
John Arndt: Then it's still sending them. To real people, every day, with your name at the bottom.
0:37
Kate: Great.
0:50
John Arndt: This is After the Demo. I'm John Arndt. I run about fifteen small software products out of Sacramento, most of them built and run with AI agents, plus an AI consulting practice.
1:01
Kate: And I'm Kate. So say your sales team, or a vendor you already pay, wants to switch on one of these AI sales reps this quarter. It finds the leads and it writes the emails.
1:12
John Arndt: That's the pitch, pretty much everywhere. Salesforce, Apollo and Gong all announced AI agents for sales teams this fall.
1:19
Kate: So if that's my company, first thing I want to know. How much do I have to babysit it?
1:24
John Arndt: Some.
1:25
Kate: Some. And does the babysitting last, or does it quietly go away next year?
1:30
John Arndt: And one you didn't ask. What happened when one company sent a lot more.
1:34
Kate: Did it work?
1:35
John Arndt: Hold that. We'll get there.
1:36
Kate: He does this. He holds things.
1:39
John Arndt: I do hold things.
1:40
Kate: And when he says it depends, my job is to make him say on what.
1:45
John Arndt: That's fair. Start with the guy who says read all of them, because he's not wrong.
1:53
John Arndt: So the rule's from Jason Lemkin. He posted it earlier this month. Never let an AI sales rep send its first thousand emails unless a person has read every single one.
2:03
Kate: You said that already. I was hoping it'd shrink.
2:06
John Arndt: It doesn't shrink. And after the first thousand, he says twenty to thirty minutes of spot checks a day is enough.
2:12
Kate: Does anybody actually do that?
2:14
John Arndt: I'd guess not many. That's probably why he posted it.
2:17
Kate: I'd quit at email forty.
2:20
John Arndt: Forty.
2:21
Kate: Forty. I'm out at forty. By forty they all sound the same and I'm just clicking next.
2:26
John Arndt: Then let's do some of your forty together, and see what you'd miss. Say you sell scheduling software to dental offices. Your new AI sales rep wrote its first batch overnight. Open email one.
2:36
Kate: Email one. Hi, I noticed your practice is growing. Fine. A little greasy, but fine.
2:43
John Arndt: Read to the bottom.
2:44
Kate: And if you book a call this week, I can get you a special discount on your first year. Wait. Do we do that?
2:51
John Arndt: You tell me. Did anybody approve a discount?
2:53
Kate: No. I'd remember approving that.
2:56
John Arndt: That's one Lemkin says you'll catch. Pricing that gets aggressive. It offered a deal you never signed off on.
3:01
Kate: And it sounded so sure about it.
3:04
John Arndt: They always sound sure. Keep going.
3:06
Kate: Email six. Dear Doctor, I hope this message finds you well. Very polite. Email seven. Doc! Quick one. You're leaving money on the table and I can fix that by Friday.
3:18
John Arndt: He's got that one too. The tone that swings around.
3:21
Kate: Email six is your uncle at a wedding. Email seven is a car lot.
3:25
John Arndt: Same tool. Same afternoon.
3:27
Kate: Okay, but that one I'd catch. That's funny. I'd be wide awake for that.
3:31
John Arndt: You'd catch it at email seven. Would you catch it at email thirty-eight?
3:35
Kate: At thirty-eight I'm two emails from quitting.
3:38
John Arndt: Right. Next one. Email twelve.
3:40
Kate: Our software syncs directly with your insurance billing, so your front desk never has to touch a claim again. Ooh. Does it?
3:47
John Arndt: Does it?
3:48
Kate: I don't know, it's your pretend software.
3:50
John Arndt: It doesn't. And that's on his list. A feature it made up. It wrote the exact sentence a dental office wants to hear, and it isn't true.
3:59
Kate: Hang on. How would I even catch that? I'm not the product person.
4:02
John Arndt: You'd ask. That's part of the reading. Whoever reads them has to know what the product actually does, or go find out.
4:09
Kate: So it's homework with a pop quiz.
4:11
John Arndt: Kind of, yeah. And the last one on his list is my favorite. The email that almost went to the wrong group.
4:17
Kate: Almost?
4:18
John Arndt: Almost. Because somebody read it first. That's the whole point.
4:22
Kate: Like what, the pitch for brand new practices going to dentists who already pay you?
4:27
John Arndt: That kind of thing. They get an email saying, have you considered trying us. They have. They're paying you. Now they're wondering who's running the place.
4:35
Kate: And they forward it to somebody.
4:36
John Arndt: To everybody. With a question mark.
4:39
Kate: Why can't I just spot check from day one? Read every tenth one. Same idea, way less time.
4:45
John Arndt: Because you don't know what to spot check for yet.
4:47
Kate: I know what a bad email looks like.
4:49
John Arndt: You know what a bad email from a person looks like. This thing goes wrong in its own ways. The discount nobody approved. The tone that swings around after
4:57
Kate: After lunch. Yeah. Okay.
5:00
John Arndt: And you didn't know either of those was on the list until you read the batch.
5:04
Kate: So the first thousand is how I find out what the list even is.
5:08
John Arndt: The first thousand is how you learn the tool. Lemkin says it pretty plainly. After a thousand, his words, you'll truly understand your AI agent. Then the twenty to thirty minutes a day is maintenance. You know where it breaks, so you know where to look.
5:22
Kate: Okay, but who has that kind of time? Honestly. Who's sitting there reading a thousand of anything?
5:28
John Arndt: Who has the time to call a dental office and apologize for a discount you don't offer?
5:32
Kate: That's a cheap shot.
5:34
John Arndt: It's a little cheap.
5:35
Kate: It worked, though. That's the annoying part.
5:38
John Arndt: So where are we on forty?
5:40
Kate: Fine. Sixty. If they're short.
5:42
John Arndt: We're making progress.
5:43
Kate: Don't push it. I said if they're short.
5:46
John Arndt: And he ends the post on this, which I like. Quote. But you have to do the first thousand yourself. All of them.
5:52
Kate: He really wanted that one to land.
5:54
John Arndt: He's not leaving you a loophole.
5:57
Kate: Which, fine, I get why. If you skip the reading, you never find out it's offering discounts.
6:02
John Arndt: You find out when a customer asks for one.
6:05
Kate: Okay, but the vendors have an approve button now. Isn't that the reading?
6:14
John Arndt: Sort of. And the optimistic case is real. Apollo, the sales software company, put out three AI agents this fall that live inside its new Slack app. One finds leads that match the customer you're after
6:26
Kate: In Slack. Where I already ignore everything.
6:29
John Arndt: In Slack. One finds leads, one proposes whole email sequences, and one watches the ones already running and suggests fixes. And every one of them needs a person to approve it. That's the part I like.
6:42
Kate: Okay. That's better than I expected, honestly.
6:45
John Arndt: And it's not only Apollo. In September, Anthropic put out a beta built with Salesforce. It's called Salesforce in Claude. And Anthropic says by default, Claude asks the seller to approve each proposed change before it's written.
6:58
Kate: By default.
6:59
John Arndt: Yeah. And Salesforce described what it calls long-horizon agents. You give one a goal. Their example is, re-engage my at-risk deals. And it works on that for days, weeks, even months.
7:12
Kate: That's a long time to leave something alone.
7:15
John Arndt: It checks in for your approval, based on guardrails you define. That's in pilot, and Salesforce says it's generally available in November.
7:22
Kate: A robot that won't let your dead deals die. For a whole quarter.
7:26
John Arndt: With check-ins. So my point is, the big vendors put a human in the loop. I think that's good news.
7:33
Kate: Can I read you Apollo's own sentence?
7:35
John Arndt: Sure.
7:36
Kate: Quote. By design, every agent requires that approval today as we want to earn your trust before we take the human out of the process entirely in future iterations.
7:46
John Arndt: Yeah.
7:49
Kate: Today. Earn your trust. Take the human out entirely. So your seatbelt comes with a countdown.
7:56
John Arndt: They didn't say when.
7:57
Kate: They didn't have to say when. They told you which direction.
8:01
John Arndt: Okay, but come on. Approval on every single email doesn't scale either. Nobody's approving a thousand a day forever. Lemkin himself says after the first thousand, you move to spot checks.
8:12
Kate: Spot checks by a person. That's not the same as nobody.
8:15
John Arndt: Okay. Fewer checks is fine. Nobody checking at all is what you watch for.
8:21
Kate: And if the approve button turns into a sample?
8:24
John Arndt: If a named person still reads the sample, fine. That's what Lemkin's describing.
8:28
Kate: Fine. We split that one. What's the other thing? Because you've got a face.
8:33
John Arndt: Permissions. What the agent can actually reach. There was a post this month I loved. Danielle Morrill posted that her Claude Code setup has reached the point where she doesn't do the data entry around passwords and logins anymore. Her words, it is glorious.
8:48
Kate: I believe her. Nobody likes typing passwords.
8:52
John Arndt: And she clearly thought hard about it. She had her agents write up how it works. Each of her computers gets its own read-only, limited account in her password manager, and she can shut them off one machine at a time. And her rules don't let the agents print secrets.
9:06
Kate: Let me be the IT person at a credit union for a second.
9:10
John Arndt: Oh, no.
9:11
Kate: Hi. Who holds the keys?
9:13
John Arndt: Each machine has its own limited access, and a person can shut it off.
9:17
Kate: What can the agent touch?
9:18
John Arndt: Whatever that machine's access reaches. It's read-only, and it's limited.
9:22
Kate: And when that laptop walks out of the building?
9:25
John Arndt: You shut that one machine off.
9:27
Kate: After you notice it walked out.
9:28
John Arndt: After you notice.
9:30
Kate: Out of character now. Her own write-up says there's no approval prompt each time a password gets used. Anything running as her on that machine can read them until she revokes the access.
9:40
John Arndt: She does say that. Plainly.
9:42
Kate: And she says that's the right trade for her, because the machines are hers and the access is limited.
9:47
John Arndt: Which is fair.
9:48
Kate: It is. But if you need a check every time a password gets used, like a fingerprint prompt, she says this isn't that. So even she says it depends whose machine it is.
9:58
John Arndt: Yeah. You got me there. Convenience isn't an access policy.
10:02
Kate: Can I get that in writing?
10:03
John Arndt: For her own machines, it's her call, and she made it carefully. For a company, somebody signs off on what the agent can reach.
10:10
Kate: Somebody with a name.
10:12
John Arndt: Right. Not the team.
10:13
Kate: Because the team is how nobody does it.
10:16
John Arndt: So, more homework, since you've got some anyway. Find the approve setting in whatever tool you run. Write down the one person who owns it. And while you're in there, write down what the agent's allowed to reach.
10:27
Kate: And for the record, the approve button doesn't change my forty.
10:30
John Arndt: Sixty.
10:31
Kate: Sixty. Short ones.
10:34
John Arndt: What bugs me is where Apollo says it's headed. Eventually, the human comes out. And if nobody can read all this, that sounds like the answer. There's another one.
10:43
Kate: Send less.
10:44
John Arndt: Hold that.
10:45
Kate: He's holding things again.
10:47
John Arndt: Let me show you somebody who sent more.
10:54
John Arndt: Okay, I'll be the brochure for a minute. The point is scale. A small team can't hand-write every email. One rep with an agent reaches a lot more people.
11:03
Kate: And honestly? I'd buy that brochure.
11:06
John Arndt: So here's a company that scaled up and wrote it down. Voiso. Andreas Gregoras, who runs brand and PR there, wrote in October about their own sales development team. These are their own numbers, self-reported, so take them that way.
11:19
Kate: Noted.
11:20
John Arndt: Through twenty twenty-five, that team put out two to four campaigns a month. Since February of this year, the same team launches eight to sixteen. No new hires.
11:29
Kate: That's about four times as many. And nobody quit?
11:33
John Arndt: Not that he says. Same people.
11:35
Kate: Huh. So the brochure was right.
11:37
John Arndt: Separately, he looked at the replies. Over the past twelve months they sorted three thousand seven hundred four replies. Direct meeting requests were roughly one percent of those.
11:47
Kate: Wait. Of the replies. Not the emails.
11:50
John Arndt: Right. People who actually wrote back. Roughly one in a hundred of them asked for a meeting.
11:55
Kate: I was rooting for them.
11:57
John Arndt: So was I. And he tested the other thing everybody wonders about. What happens when you just send more?
12:03
Kate: Nothing good, I bet.
12:04
John Arndt: He says when everybody's inbox started filling up with competent AI email, they tested what he calls the obvious response, and raised their sending volume a lot. And in his words, the number of replies did not scale with it.
12:17
Kate: They turned the hose way up and the bucket didn't fill.
12:20
John Arndt: That's how he tells it.
12:21
Kate: One company, though.
12:23
John Arndt: One company, their own numbers. I'm not calling it a law of nature.
12:27
Kate: Then what were all the other replies? If only one percent was somebody saying let's meet.
12:31
John Arndt: He says the rest came back as a question or an objection.
12:34
Kate: That's it? Question or objection?
12:37
John Arndt: Or somebody asking for more detail, or pointing you at somebody else. And his line on that is, each of those has to be answered by a person who knows what is actually true.
12:47
Kate: Great. The sending scales, and a person still has to answer.
12:50
John Arndt: Every reply that isn't a meeting, yeah.
12:53
Kate: Which is ninety-nine percent of them.
12:55
John Arndt: Roughly. But I'm not giving up scale that easily. A small team can't compete without it. I'd say reviewing is a phase. You read a lot early, and then you grow
13:04
Kate: No. Wait. Flip it. If no person can read what it sends, it's sending too much.
13:10
John Arndt: That's a strong rule.
13:11
Kate: It's a simple rule. Cut the volume down to what somebody can actually read, and Lemkin's thousand fits in your mornings.
13:18
John Arndt: A thousand emails fits in my mornings?
13:21
Kate: Fifty does. Fifty a day is a thousand in four working weeks. I can do fifty. I'd make it.
13:27
John Arndt: So it's a very long first month.
13:29
Kate: It's a normal first month. With coffee. The homework gets smaller when the pile gets smaller.
13:35
John Arndt: Okay, that's annoyingly good.
13:37
Kate: And the guy who wrote the article agrees with me. Doesn't he?
13:40
John Arndt: Partly, yeah. He wrote that, quote, a precise offer for a narrow segment will outperform a general offer sent to a larger audience.
13:49
Kate: Narrow means fewer people. I'll take partly.
13:52
John Arndt: And he wrote that writing that's recognizably human became, his words, unusual again. So when everybody's inbox fills up with competent machine email, the one that stands out is the one that sounds like a person wrote it for you.
14:05
Kate: Which is hard to do at sixteen campaigns a month.
14:07
John Arndt: It's hard to do and also read every one, yeah.
14:11
Kate: And let me guess, he also says a great email to the wrong person still goes nowhere.
14:16
John Arndt: Almost word for word. A good sentence sent to the wrong person still fails.
14:20
Kate: So you could write the perfect email to the wrong thousand people.
14:24
John Arndt: And nobody books anything. Very politely.
14:27
Kate: So.
14:28
John Arndt: So you win that one. I was treating review as a phase you grow out of. The rule only works if you cut the volume first.
14:35
Kate: Say it slower.
14:36
John Arndt: You. Win. That. One.
14:40
Kate: That's two. I'm keeping a list.
14:42
John Arndt: And to be fair to him, he's not saying the job goes away. He expects fewer sales development roles, filled by people who are paid more. His words, the role has not disappeared.
14:53
Kate: Fewer of them, paid more. I can live with that. Mostly the paid more part.
14:58
John Arndt: Of course it is.
14:59
Kate: So if your sending went up this year and your meetings didn't?
15:03
John Arndt: Stop adding volume. Pick the slice of accounts a person can actually read for, and send to those.
15:08
Kate: Okay. I'm reading fifty a day. What am I reading them as?
15:12
John Arndt: As the person who gets them. I learned that one the expensive way. Almost.
15:22
Kate: Almost expensive. Go on.
15:24
John Arndt: So, late August. I decided to buy every plan on seven of my own products, the way a customer would.
15:30
Kate: Your own products?
15:32
John Arndt: My own.
15:33
Kate: Why would you need to buy your own stuff? You built it. You know where everything is.
15:37
John Arndt: You'd think. That's the trap. Knowing where everything is isn't the same as buying it like somebody who doesn't.
15:44
Kate: In through the front door, then.
15:45
John Arndt: Yep. Pick a plan, pay for it, see what shows up after. Every plan.
15:50
Kate: And?
15:50
John Arndt: And it caught real bugs. Before any customer did. Including a checkout that would have let someone subscribe for zero dollars.
15:57
Kate: Zero.
15:59
John Arndt: Zero.
16:00
Kate: So for a while, it was free?
16:02
John Arndt: It would have been. I caught it first. But it was my mistake, it was sitting right there, and the only reason I found it is I tried to buy my own thing.
16:10
Kate: Which product?
16:11
John Arndt: I'm not saying.
16:12
Kate: It's the good one, isn't it.
16:14
John Arndt: Moving on.
16:16
Kate: No, wait. You must've been furious.
16:19
John Arndt: Mostly embarrassed. Nobody had hit it, so nothing broke. But I'd have sworn that checkout was fine.
16:24
Kate: Everybody would swear that about their own checkout.
16:27
John Arndt: That's the problem. You don't look at your own checkout as the person paying.
16:31
Kate: And now there's a rule. I can see it on your face.
16:34
John Arndt: Test the path the money takes, as a customer, on a schedule.
16:37
Kate: Meaning somebody who's never seen the inside.
16:40
John Arndt: Right. From the inside, everything looks fine. You know what the button's supposed to do, so you see it doing that.
16:46
Kate: What does that have to do with emails? That's a checkout.
16:49
John Arndt: The AI sales rep is the front of that same path. It's the first thing a lot of prospects ever see from you. Reading its emails as the person getting them is the same test.
16:59
Kate: Oh. The made-up feature.
17:01
John Arndt: Yes. The insurance billing thing. That's the email version of the zero-dollar checkout. To whoever wrote it, it looks fine. It's a nice sentence. It only breaks when the dentist reads it and believes it.
17:12
Kate: And then calls you asking for it.
17:14
John Arndt: And then you get to explain why your email promised something you don't sell.
17:17
Kate: When I read them, I'm the dentist.
17:19
John Arndt: Exactly. You don't know what the product does. You only know what the email told you.
17:24
Kate: Okay, but I know too much. How do I read it like a stranger?
17:28
John Arndt: Every promise in it, you ask one question. If I paid today, could I actually get this?
17:34
Kate: And if the answer's no?
17:35
John Arndt: Then that sentence doesn't go out. Same as the checkout. You don't ship it and hope somebody notices.
17:40
Kate: Say I test my checkout once. It works. Done. Why a schedule?
17:45
John Arndt: Because the thing changes underneath you. Every release is a new checkout.
17:49
Kate: Like Apollo's next version.
17:51
John Arndt: Like any of them. Remember the approval sentence. Approval today. The tool you checked in October might not be the tool you're running in March.
17:59
Kate: Great. March me has homework too.
18:02
John Arndt: March you has a calendar reminder. That's lighter.
18:05
Kate: Then I reread every time they ship something.
18:08
John Arndt: At least then. And on a calendar, so it doesn't depend on you noticing a release note.
18:13
Kate: How often, though? Give me a number.
18:15
John Arndt: Pick one you'll actually keep. The point is it lives on the calendar, so you don't have to remember.
18:20
Kate: That's a dodge.
18:21
John Arndt: It's a little bit of a dodge. A schedule you skip is worse than a slower one you keep.
18:26
Kate: Fine. Monthly. Because I don't read release notes either.
18:30
John Arndt: Nobody reads release notes. Which is kind of this whole episode.
18:34
Kate: Then explain the guy whose agent built a whole business while he was on a date.
18:38
John Arndt: Alex Finn. He posted that his AI agent built and launched a whole business for him. The website, a newsletter, and reaching out to sponsors.
18:47
Kate: Was the date good?
18:48
John Arndt: He didn't say. His post says that in the fifteen hours after launch, the site had twenty thousand visitors and fifteen hundred email signups. No sponsor yet. He says one will be found shortly.
18:59
Kate: By his count.
19:00
John Arndt: Right. And readers added a note under the post. It says he announced the site himself the day before, already live, and later set up a bot to maintain it.
19:09
Kate: So the timeline's a little fuzzier.
19:11
John Arndt: Maybe. Either way, I don't think the launch is the interesting part. Launching got cheap.
19:16
Kate: Great. Who reads it on day two?
19:18
John Arndt: That's the whole job now. Somebody reading what goes out under your name, every day.
19:23
Kate: Fifteen hundred signups, and somebody has to read what goes to them.
19:27
John Arndt: Fifteen hundred people who each think a person wrote to them.
19:31
Kate: And if nobody reads it?
19:32
John Arndt: Then you find out the way I almost did. From a customer.
19:40
Kate: Okay. One of the big vendors says AI wins deals for you now. I have questions, and you're answering them fast.
19:48
John Arndt: I can do fast.
19:49
Kate: Rapid fire. Short answers. You're bad at short answers.
19:53
John Arndt: I'm working on it.
19:54
Kate: Five questions. I'm timing you. Question one. Does AI win deals now?
19:59
John Arndt: Gong's CEO, Amit Bendov, said this at their September event. Quote. For years, AI told you how to win. Now, AI wins deals for you.
20:09
Kate: Or it loses one, a made-up feature at a time.
20:12
John Arndt: Fair. They also announced Gong Activate, which they say will use always-on agents that keep evaluating every account and deal. And an Agent Builder, for custom agents that run on triggers you define.
20:24
Kate: Triggers you define. So somebody still has to define them.
20:28
John Arndt: Somebody with a name.
20:30
Kate: I knew that was coming. Also, I stopped the timer. That was not short.
20:35
John Arndt: Then here's the short half. Separately, Gartner predicts that by twenty twenty-eight
20:39
Kate: Is this the short half?
20:40
John Arndt: This is the short half. AI agents will outnumber sellers ten to one, and fewer than forty percent of sellers will say the agents improved their productivity.
20:48
Kate: Ten agents for every seller, and most of the sellers shrug.
20:52
John Arndt: Yeah. And Gartner's analyst, Dan Gottlieb, had a line I liked. If those systems are fragmented, the agents will scale the fragmentation.
21:00
Kate: In English.
21:01
John Arndt: If your stuff's a mess, the agents make more mess, faster.
21:05
Kate: Now that was short.
21:06
John Arndt: I've been practicing.
21:07
Kate: Question two. Do I still need salespeople?
21:10
John Arndt: Someone posted a clip quoting Elon Musk on why his company went after Cursor, the coding tool. He gives a few reasons. Coding data. Talented people. And then, from the post's transcript, that good enterprise sales team with relationships with thousands of companies.
21:25
Kate: Hang on. Where's the clip from?
21:27
John Arndt: It's somebody's post. I can't check the clip myself, so treat it that way.
21:31
Kate: If that quote's right, even the AI company wanted the salespeople.
21:35
John Arndt: If it's right. And the relationship is exactly what a bad email spends. One weird discount email to a customer you've had for years, and you're spending something you can't buy back quickly.
21:45
Kate: Not even with a discount.
21:47
John Arndt: Especially not with the discount nobody approved.
21:50
Kate: Question three, and this one's mine. Is there a version where the human never leaves the loop? No countdown. No homework.
21:57
John Arndt: One of mine, Intakra, works that way. It watches public sources, things like job posts and filings, for the companies a team already sells to, and when one of them starts something the team could help with, it emails the team a heads-up with the proof.
22:11
Kate: And then it writes to them?
22:13
John Arndt: It doesn't write to those companies or send them anything. A person on the team decides whether to reach out, and writes it.
22:18
Kate: So it's homework too.
22:20
John Arndt: Short homework. It's a few a week, on purpose.
22:23
Kate: Fine. I'll allow it. Question four. Say I want an agent anyway. Where do I start?
22:29
John Arndt: Inbound. Where somebody already raised their hand.
22:32
Kate: Why there?
22:33
John Arndt: Because they came to you. They asked a question. Answering a question is a smaller job than starting a conversation with a stranger.
22:40
Kate: So the robot gets the easy ones.
22:42
John Arndt: The ones with a hand already up. Salesforce describes one called Piper that works around the clock on inbound prospects. It answers questions over text, voice or video, qualifies the leads, and books meetings. They describe it as available now. They list another one, Hunter, for November, and trade coverage calls that the outbound one.
23:01
Kate: And you'd wait for Hunter?
23:03
John Arndt: My opinion, not a fact. I'd start with the people who came to you before the people who didn't.
23:08
Kate: And somebody still reads what it says.
23:11
John Arndt: Yes.
23:11
Kate: Even inbound? They already like you.
23:14
John Arndt: That's why. They already like you, and one bad answer is how they stop.
23:18
Kate: Huh. I had that backwards.
23:21
John Arndt: Inbound feels safe because it's friendly. That's my opinion, anyway.
23:25
Kate: Okay. That's actually worse.
23:27
John Arndt: Right. Somebody asked a real question.
23:29
Kate: Same rule, then.
23:30
John Arndt: The answer had better be true. That's the made-up feature problem again, just faster.
23:35
Kate: That one was actually short. Timer's back on.
23:37
John Arndt: For one question?
23:39
Kate: Last one. What do I do this week? And I want two things. Not five.
23:43
John Arndt: Two. Pull the last fifty emails your tool sent, and read every one as the person getting them. Like testing your checkout.
23:49
Kate: That's one.
23:50
John Arndt: And name the one person who reads what it proposes. A name, not the team.
23:55
Kate: Two. Done. Don't add a third.
23:57
John Arndt: I wasn't going to.
23:58
Kate: You were. You took a breath.
24:00
John Arndt: I was going to.
24:02
Kate: See? You can stop at two.
24:04
John Arndt: It was physically difficult.
24:05
Kate: I could tell. Now squeeze it. Give me the one thing. The one I'll actually do.
24:12
John Arndt: The one thing. Pull the last fifty emails your AI sales tool sent. Read every one as if you were the person getting it, and circle anything you wouldn't have signed yourself.
24:22
Kate: And if it hasn't sent anything yet?
24:24
John Arndt: Then read its first fifty before it does.
24:26
Kate: And put one name, not a team, next to the approve button.
24:30
John Arndt: That's two things.
24:31
Kate: It's one thing with a name on it.
24:34
John Arndt: I'll allow it.
24:35
Kate: Read it like you'd test your checkout.
24:37
John Arndt: As a customer. On a schedule.
24:39
Kate: And since you've been counting all episode, final number. I'd make the thousand. At fifty a day.
24:45
John Arndt: From forty to a thousand in one episode.
24:48
Kate: Don't make it a thing.
24:49
John Arndt: It's a little bit of a thing.
24:51
Kate: Send this to whoever just switched on the AI sales rep. Before it sends anything else in your name.
24:56
John Arndt: Links are in the show notes. Jason Lemkin's post, the Voiso article, and the vendor announcements we talked about.
25:02
Kate: So you hired a robot and got homework.
25:04
John Arndt: Fifty emails of it. One morning. That's After the Demo.
25:08
Kate: Go read your fifty.