Sales Tip 1828: Did you notice? Adapt & Experiment

You’re listening to the Daily Sales Tips podcast and I’m your host, Scott Ingram. I ran a little bit of an experiment this week and I’m really curious to see if you noticed. Even if you’ve already listened it might be worth going back and listening again to the last 4 tips from 1824 to 1827. This will work way better if you do that before listening to the rest of this tip. It’ll take you less than 20 minutes and regular speed and it’s a pretty good set of tips too. I’ll wait …… Did you do it? Probably not, but that’s OK. Let me ask you something. Did you notice anything different, especially with my introductions on those tips? What about with this tip today? Here’s the thing. This isn’t me. This is AI audio generated by ElevenLabs and one of their Professional Voice clones. So my tip today is actually about experimenting with AI and continuing to adapt.

I actually tried this about a year and a half ago. Using the same service, though their model was simpler then, in June of 2023 I wanted to see if AI voice cloning was good enough that I could theoretically use it to replace myself and the intros and outros I do for every tip on this show.

At the time, it wasn’t even close. The voice more or less sounded like me, but the pacing was way off and it was immediately obvious that it wasn’t really me. Fast forward to a couple of weeks ago I thought I’d try again. Even the simple model that I used in my original test was dramatically better. So I went to the next step and uploaded about 90 minutes of audio to train one of their professional voice clone models. In my original test I would say it was maybe 40% accurate? This time it was more like 90-95%. In fact, I originally intended to try this experiment last week and started with tips 1819 and 1820. These are actually worth going back and listening to, especially if you actually went back and listened to this week’s tips. They’re not nearly as good as this week’s set, which I would characterize at closer to 98% accurate. The difference was that first training set I shared was only about 90 minutes, when they recommend closer to 3 hours and it consisted primarily of my solo tips and didn’t really include many examples of the intros and outros I do for all the other episodes. So I started over and retrained the model with about 3 and a half hours of samples from over 350 recent tips. Again, the difference is pretty remarkable. Depending on how long you’ve been listening, this might actually be good enough to fool a pretty regular listener. So the first lesson is that you’ve got to keep experimenting with this stuff. AI is advancing so quickly, that something you completely wrote off a little over a year ago, might be totally relevant now.

The second lesson is that when you’re working with AI the more relevant data and context you can give it, the better your outputs are going to be. Again, you can hear a pretty good apples to apples comparison if you listen to my AI intros of Meshell Baker in episodes 1819 and 1824. The last lesson is we have to adapt faster than ever. Over the course of my 20+ year career the pace of change has only increased and I’m pretty sure we’re always going to be able to say that. So if you’re not willing to adapt and flex things are going to get really hard for you. My advice is to roll with it. There are opportunities here if you look for them. In fact, tomorrow I’m going to share another pretty wild AI experiment with you. If you haven’t heard the results of or played with Google’s Notebook LLM, there’s a pretty good chance it’ll blow your mind. Now I’m not planning to let AI completely take over this podcast just yet. I’ll start recording these intros and outros and my own tips again myself, but it’s nice to know that I’ve got another option that my team can leverage if I just don’t have the time, or if I’m not feeling well and my voice is off, this can totally work. So as always, thanks for listening, and be sure to come back tomorrow for another great sales tip!