The Rude Director of Sales Was Fake. The Anxiety Was Real.

I've been building a cold call simulator inside Galadon, and I haven't released the video yet. But the concept behind it is something I need to talk about now, because I think most people are thinking about AI sales training completely backwards.

Here's what I built: five different user personas, all powered by GPT-4. You call one of the personas - let's say the Director of Sales at a mid-size agency - and you try to pitch your startup and book a meeting. The whole thing is audio. Text-to-speech in the background. But when you're actually doing it, it doesn't feel like you're talking to software. It feels like you're talking to a person.

And that person is rude.

Not rude as in incompetent. Rude the way a real Director of Sales is rude when you catch them cold on a Tuesday afternoon and they've already taken six unsolicited calls that day. They push back. They go cold. They make you feel like you're wasting their time. And that feeling - that specific social discomfort - is the entire point.

What Actually Breaks Salespeople

I've worked with over 14,000 agencies and entrepreneurs. I've helped them generate more than 500,000 sales meetings. And in all that time, I've watched a version of the same failure repeat itself constantly: someone reads every objection-handling guide ever written, internalizes the frameworks, knows all the right things to say - and then completely freezes the moment a real prospect's voice goes cold on them.

Information isn't the problem. Consequence is.

When you're reading a script, there are no stakes. When you're role-playing with a colleague who's trying to be helpful, the stakes are fake and you both know it. Your colleague isn't actually going to hang up on you. They're not going to make you feel stupid. They don't want to. So you practice in a consequence-free environment and you wonder why the skill doesn't transfer when it counts.

Real sales calls have stakes. Somebody might actually say no. Somebody might get annoyed. Somebody might make you feel small. That's the emotional context where your training either holds or it doesn't. And you can't replicate that with a PDF or a classroom exercise or a partner who's trying to be kind to you.

What you need is a situation where the consequence feels real - even if technically it isn't. And that's exactly what a threatening AI persona manufactures.

The Simulator I Built (And Why It Works)

The way the Galadon simulator works: you dial in, you get the Director of Sales persona, and you try to get through the call. The AI is programmed to act the way a cold-call recipient actually acts. It's skeptical. It's impatient. It's not trying to help you practice - it's trying to get off the phone. You have to push through that resistance and actually book the meeting.

What I noticed when testing it is that people get genuinely anxious. Not pretend-anxious. Actually uncomfortable. Their palms sweat. They stumble over words they've said a hundred times in a mirror. They lose the thread of their pitch when the persona interrupts them, because that's what happens when the emotional stakes kick in - your brain switches modes and suddenly the script you memorized isn't in your fingers anymore.

That's a feature, not a bug.

Because the moment you can stay composed through that - when a rude, dismissive AI persona is cutting you off and you still find your footing and pivot and get the call back on track - you've actually wired something. You've practiced under the emotional conditions that matter. Not the calm conditions. The hard ones.

The tech behind it is GPT-4 with text-to-speech. It's audio. It's not complicated from an engineering standpoint. What makes it work isn't the sophistication of the model. It's the design of the persona - threatening enough to trigger real social anxiety, but not so cartoonishly hostile that it breaks immersion.

Consequence Is the Variable

Think about how most sales training actually works. You get a script. Maybe you watch some calls. Maybe you shadow someone on a few live conversations. Then you're thrown into the deep end with real prospects and real pipelines.

The problem is the gap between the training environment and the performance environment. In training, there's no consequence for failure. In performance, there is. And that gap is where skills go to die.

Athletes figured this out a long time ago. You don't get good at performing under pressure by only practicing when there's no pressure. You simulate game conditions in practice - noise, fatigue, hostile crowd, time running out. The point isn't to make practice miserable. The point is to narrow the gap between what practice feels like and what the game feels like so your body already knows what to do when it gets hard.

Sales is the same. If you want to get good at handling the cold call that goes wrong - the prospect who interrupts you, the gatekeeper who's openly condescending, the decision-maker who gives you exactly twelve seconds before they're done - you have to practice those conditions specifically. Not conditions-adjacent. Those exact conditions.

Until now, the only way to do that was to make real cold calls, which means burning real leads while you're still learning. Or to hire a professional actor to run you through adversarial role plays, which most teams can't afford. Or to get a manager to do it, but managers don't want to spend three hours a day being artificially rude to their own reps.

The AI persona solves all three of those problems at once. It's available whenever you need it. It costs a fraction of what a sales coach does. It doesn't get tired of being the villain. And it doesn't pull its punches to spare your feelings.

There's a Viral Parallel That's Worth Noting

Around the time I was building this, a Tesla AI sales bot made the rounds on Twitter. The clip went viral - an AI making outbound phone calls on behalf of Tesla, handling objections, setting test drive appointments, talking like a real person. People lost their minds over it. The comments were split between amazement and unease.

What most people missed is that the AI wasn't impressive because it was technically perfect. It wasn't. There were tells. But it was impressive because it was close enough - close enough that the person on the other end wasn't entirely sure, close enough that the social dynamic felt real. That's the threshold that matters. Not perfection. Believability.

Same principle in training. You don't need the AI persona to be indistinguishable from a real human. You just need it to be threatening enough that your body treats it like a real call. That's a much lower bar to clear, and we're well past it with current models.

What's interesting about that Tesla demo is also something that caught my attention: the company behind the bot was apparently so bad at selling it that you literally couldn't buy it. You'd try to get access, the calendar would be full, and you'd never hear back. A reminder that building something great and being able to sell it are two completely separate skills - and the second one is what this simulator is designed to develop.

Where This Goes Next

The immediate application is obvious: train your cold callers before they touch a real prospect. Run them through the rude Director of Sales ten times. Make them get hung up on. Make them fumble and recover. Make them earn the ability to stay calm when it goes sideways - because it will go sideways, and the reps who've already lived through a version of that are the ones who don't fold.

But the broader opportunity is bigger than that. Think about what a sales emulation layer could do inside a platform like Galadon. Your reps can practice their pitch, handle objections against a realistic persona, get consistent conditions every time - no manager required, no live leads burned, no schedule to coordinate. That's sales enablement that actually scales.

And then there's the calendar integration angle. What I was exploring with the team is connecting the simulator to something like the Google Calendar API - so the AI persona can find real availability and recommend actual time slots when it's time to book the meeting. At that point the simulation isn't just training. It's infrastructure. You're not practicing how to book a meeting. You're booking one, inside a framework that tracks every rep's progress and flags where they're losing the call.

We're also at the point where you can send 133,000 cold emails a month - which is roughly what we're doing right now - and route the responses into a system that processes and qualifies them automatically. Combine that with a trained rep who's gone through fifty rounds in the simulator before touching a live conversation, and you've got a real outbound machine. The volume is there. The training infrastructure is now possible. The bottleneck used to be getting reps ready fast enough. That problem is being solved.

If you're building your own outbound system and need to get your prospect list dialed in first, ScraperCity's B2B database is what I use for lead sourcing - unlimited records, no per-lead fees, built for the kind of volume where this training approach actually makes sense. Tools like Smartlead and Instantly handle the sending infrastructure. And if you want enrichment and automation layered on top of your outreach, Clay is worth looking at.

The Insight Nobody Talks About

Everyone in this space talks about objection handling frameworks. They talk about scripts, sequences, subject lines. Those things matter - I wrote an entire book about it, the Cold Email Manifesto, and the fundamentals in it still apply. You need a sharp offer, a real case study, a clear call to action. That part hasn't changed.

But frameworks are inert until they're tested under pressure. And pressure has always been the hard part to manufacture in a training context. You can't just tell someone how to handle a hostile prospect. They have to feel it - the coldness in the voice, the shortened responses, the sense that they have about eight seconds to say something worth hearing or the call is done. That feeling has to become familiar before it stops being destabilizing.

That's what the simulator does. It makes the uncomfortable familiar. It shrinks the gap between training and reality to the point where, when you're on a real call and the real Director of Sales goes cold on you, some part of your brain already knows this territory. You've been here. You got through it. You can get through it again.

The AI persona isn't impressive because of the technology. It's impressive because it's the first time in history that you can dial up consequence without putting a real human on the other end. That changes what training can be - not just for cold calling, but for any high-stakes conversation where information alone was never enough.

If you want the frameworks to start with while you're building out your own training process, grab the Top 5 Cold Email Scripts or the Discovery Call Framework - both are free. And if you want to work through this kind of infrastructure with me directly, check out Galadon Gold - that's where we go deep on the full system.

The rude Director of Sales is fake. The anxiety he creates is completely real. And that's the whole point.