What Salesforce Admins Trust AI to Do Now (That Seemed Outlandish a Year Ago)

By

Today on the Salesforce Admins Podcast, we talk to Jennifer Lee and Kate Lessard, Lead Admin Evangelists, and Joshua Birk, Senior Director of Admin Evangelism at Salesforce.

Join us as we chat about how our relationship with AI has changed over the past year.

You should subscribe for the full episode, but here are a few takeaways from our conversation with Jennifer Lee, Kate Lessard, and Joshua Birk.

AI has gotten better at building things

For Kate, the biggest difference when it comes to AI is that she trusts it to do more things. While she used to make sure that an agent never did anything without her approval, she’s much more confident in relying on the guardrails she’s put in place.

Jen has had a similar experience. “A year ago, if AI asked me if it could create a record or touch my Salesforce org I’d say no way,” she says, “I’m going to need to see everything.” Nowadays, she just needs to make sure it’s on the right track before letting it go off and create something. But AI has been amazing for creating demos, especially for testing and creating dummy data.

Trusting a personal assistant agent

Josh, meanwhile, has been taking advantage of AI as a personal assistant. It’s allowed him to reclaim his inbox and stay on top of things. “I can use my inbox on a daily basis, without having to spend the morning deleting hundreds of emails before I can read something that is actually useful,” he says.

However, he’s quick to note that his personal assistant agent has very clear boundaries on what it can and cannot do. “If it goes wrong, how quickly can you recover from it?” he asks. An empty bank account is a lot harder to fix than an email that went to the wrong person.

Getting financial advice from an agent

For me personally, this is the year that I’ve felt comfortable having an AI look at my finances and make some suggestions. It’s helped me adjust my retirement savings and plan for the future, and I couldn’t say I would have done that a year ago.

What about you? What are you doing with AI that would have seemed ridiculous to you a year ago? Make sure to listen to the full episode for more from Josh, Jen, and Kate on how they’re working with agents. And don’t forget to subscribe to the Salesforce Admins Podcast to catch us every Thursday.

Podcast swag

Admin Trailblazers Group

Social

Full show transcript

Mike:
This week on the Salesforce Admins Podcast, we’re doing a November retro with three of my Salesforce admin evangelist teammates, Jennifer, Josh, and Kate. Now, here’s what I asked them. I wanted to know what they trust AI to do today that would’ve sounded just a little outlandish a year ago. And I think you’re going to be surprised maybe what the answers are.
We also talk about where everyone’s going to draw the line with AI, and Josh shares a simple test for deciding how much access to give an AI agent. So, if you’re a Salesforce admin figuring out your own comfort level, you’ll probably recognize yourself somewhere in this conversation. And in case you haven’t already, make sure you follow and subscribe to the Salesforce Admins Podcast, whether you’re listening to us on iTunes or on the podcast app, or if you’re watching this on YouTube, be sure and subscribe to the Salesforce Admins YouTube channel.
So, with that, let’s get my fellow teammates on the podcast. So, everybody, welcome to the podcast. It’s November and time to talk some AI with fellow admin evangelists. So, we’ll get started. Jen, could you do a brief introduction of what you do at Salesforce?

Jennifer:
Sure. Oh, I do all sorts of things at Salesforce, whether it’s blogs, presentations, videos. I’ve been doing videos galore. You probably are sick of seeing me with AIforce in Action, but that’s been my thing lately.

Mike:
Never sick of AIforce in Action. Josh?

Josh:
We all have very similar roles, but humorously I’ve had a couple of conversations this week alone about how do you keep up with the platform? And it’s like, well, thankfully a large part of our job is research, right? We get time to actually go in and research and to keep this on topic. Gosh, AI is making that so much easier. So, talking to AI and creating content, that’s what I do.

Mike:
Oh, I thought you were going to say pressing two.

Josh:
There’s a lot of pressing two, yes.

Mike:
Pressing two. And Kate, certainly last but not least.

Kate:
Oh, well, thank you. I do a lot of the same things just mentioned. I have not been doing videos galore lately. I think the thing on the team that I do that’s maybe a little bit different is that I own the workshops for our team. So, I’ve been spending a lot of time on our hands-on workshops lately and creating new AI force content to get more admins hands-on.

Mike:
Mm-hmm, fun. Obviously, it’s crazy because just a few years ago we were barely talking to AI and then we had some workshops where we had thousands of prompts that we were building. And now I feel like every day there’s at least every part of what I do, AI touches some part of it. Either it’s advice that I need or research on something or to Josh’s point, I’m just telling it to go do something while I switch a screen and go tell another AI to go do something.
I think it’s fascinating how with very little training or no training, AIs just permeate our lives and especially as Salesforce admins. So, the question that I’m going to pose to everybody, and Kate, we’re going to start with you, is when you look back a year ago, what’s something you trust AI to do now that a year ago just seemed outlandish?

Kate:
Man, and it really does come down to trust, right? I think that over the past year for me, the biggest shift has been how I interact with AI. I think I’ve gone from just asking how to do something and asking a question to actually having a level of trust to have AI go and do the work for me. And I think a lot of that really comes down to the capabilities of the headless toolkit and still being able to have that kind of direction and judgment and that control over what’s going to happen.
But while I can provide that direction and judgment, I don’t necessarily feel the need to execute every single step anymore. And that’s been a really big mentality shift over the past year.

Mike:
Mm-hmm. Wow. Okay. We’ll continue on. Josh, I feel like you’re probably doing more with AI than all of us combined, but-

Josh:
Yeah, I recently made the joke. Somebody asked if I use an agent at work and I’m like, “No, I don’t use an agent at work. I use five, some of them simultaneously.” I just had one agent, one AI proof the other AI’s work and it actually found a whole bunch of flaws on it, which I wouldn’t say is something I wouldn’t trust AI doing in the past. Although I think previously, I would force myself to be the human in the loop and not necessarily put AI, checking AI, checking AI kind of thing.
So, that’s sort of new. For me, the big revelation that I’m so comfortable with is the surprise to me is AI is in my inbox. AI is in my Slack. AI is in my calendar. And if you had gone to early AI Josh in his speeches, I would be like, “Why would you even want that?” But I use a personal assistant and I’ve reclaimed my inbox. I can use my inbox on a daily basis where it used to be I would spend a morning just deleting hundreds of emails before I could even read something that was actually useful.
Now I wake up, Slack is actually in my inbox and it’s like, “Hey, here’s all the important things you actually need to address right now.” And for somebody who sometimes has a little bit of an attention deficit problem, AI adding a little bit of focus has been very helpful to me.

Mike:
Oh, sometimes.

Kate:
I remember those early Josh AI speeches about that.

Josh:
I have been the warning sign in the past. It’s true.

Mike:
I was, as you were explaining that, Josh, thinking of all of my friends that easily a year ago were just tired of everything, every application having an AI component to it. And I think it’s because it didn’t do anything, anything better than the search bar already done. But that was a year ago. Now, now it’s almost the starting point for every application that you launch on your phone or on your desktop. Jennifer, how about you?

Jennifer:
It’s really been a productivity booster for me. I think a year ago, if AI asked if I could create a record or touch my Salesforce org, I’d be like, “No way, dude. I’m going to need to see everything.” You make the recommendations like Kate, I will go in and do it. And I’ve been using a lot of Claude lately to do some of my demo work. And first I had it set to show me what you’re about to do, get my approval.
And then after just reviewing it, I’m like, “Okay, you know what you’re doing. I’m going to let you go ahead and do.” And it’s just been amazing having it just thinking about the idea, sharing it with Claude, and then it just doing things. Sometimes, it’s not perfect. You still have to go in and make some tweaks, but it’s pretty close and I love the fact that it goes and does testing for you and it tells you what it’s going to test.
And then in terms of your demos, you want to make it look real, full of data. And I love the fact that it goes and creates the data and loads it in for you. So, that’s been a huge saver for me and something that I wouldn’t have done a year ago at all. No, no way.

Josh:
Right. Yeah. I actually got called out by the community on a video I did about hooking Claude web browser, the web version to Salesforce. And somebody was like, “Why isn’t that dangerous?” And the first speech I ever gave earlier was, I don’t know if it was a user group that I did with Mike here in Chicago, and I gave the example of people putting in internal code and then OpenAI basically just owned that proprietary data. And yeah, that was true at the time.
And I think part of the reason trust with AI is so hard is because it is moving so fast and the goalposts do shift drastically. And all of these big players, OpenAI, Facebook, Anthropic, they knew they had to get the enterprise market, right? And so, they also knew that they’re not going to get those customers if you don’t trust them with hooking something up like to a Salesforce org.
But honestly, it happened and I didn’t even see it. It wasn’t until I was told, “Hey, go make a video hooking it up to Salesforce,” that I double checked that I’m like, “Is there privacy guidelines? Are there guardrails?” And there absolutely are these days.

Mike:
Yeah. I had to think about this question because I also wanted to answer it, and my answer is probably not what everybody expects, but this year, and I would say since probably June forward, I felt comfortable having an in AI, I won’t mention which one, look at my finances, like my pay stub and help me adjust my retirement savings amount and the amount that I’m saving to hit different targets.

Josh:
Interesting.

Mike:
And I say that two caveats because the second one is a question, will be a question. First of all, I 100%, here’s the pay stub, it’s a PDF, and you don’t know a single checking or any account. I am not there yet. The second that leads to my second question is I feel like I’m seeing more and more commercials where people are asking an agent to find this hotel and book it or find this trip and book it. Or one of them is they take a picture of a bill and they ask the agent to pay it.
In realistic terms, because we live in the real world, how comfortable do you think you are? And do you think maybe a year from now we’ll look back and be like, “Wow, I can’t believe we said that”?

Jennifer:
I’m just shaking my head. Nope, no way.

Kate:
We were just talking about how worried I am in general about just everything in life. So, that’s a no for me at this point. Talk to me in a year. We’ll do a recap and see where we’re at, but-

Josh:
Yeah. I have a very good friend. He’s a technical architect and we’re very like-minded when it comes to things like applications and AI and security and stuff like that. And I told him I started using this personal assistant and he thought it was insane, but it does not have access to any of those things that you just… It does not have access to my credit card. It can’t buy anything. It can’t book anything.
It is a wonderful little research assistant and helps me on my day-to-day, but it can access my email and send emails on my behalf and also send texts on my behalf, but it does follow the right guardrails, right? It does not do it without my permission and without me reviewing and all of that kind of stuff. I think it’s a very slippery slope. And the thing is what you’re describing, Mike, is that’s very much where the big players are going.
That’s how they want to really get a personal AI into your day-to-day life is to make these things so easy that it’s like using a calculator with math, right? You’re not going to book a trip anymore because the AI is just going to do so much of the day-to-day for you and so much of the to-do list. My two thoughts on it is, first of all, I don’t think they’re there yet when it comes to actually have access to a banking account and actually spending money and having things that would be if it goes wrong, put this on the list.
If it goes wrong, how quickly can you recover from it? If my AI was like, “I am sorry, I sent an email to that person, I thought we were talking about it,” I can send an email follow up and be like, “I’m so sorry, my AI screwed up.” If my bank account goes empty overnight, that’s a little bit harder to fix kind of thing. So, I think right now it really is, there’s a personal question to it of is this going to make your life easier versus the risk involved if the AI actually does make a mistake?

Mike:
Yeah, it’s interesting because you think of you want that to happen. It’s such a time saver, but the trust factor of the money is real and does the AI feel like the money is real? I’ve also run into, I feel like we’re at this convergence point where a lot of people are safeguarding against AI’s hitting AIs, makes it sound like I’m 72 and setting the VCR clock, the AI’s hitting their website.
So, I asked one, I have this rug in my living room that I love, but it’s starting to get worn out and I really want to replace it. And I asked to find me another one and it searched for three hours and it came back and it said, “Look, I can’t get to any of these sites. They’re blocking my ability to pull up the website.” And I feel like it’s a push-pull, right?
We need to safeguard that and people rightfully so with the website, I don’t want this type of traffic, but also there’s people like me that I really want this to be able to search this site and crawl and it’s not. So, it hits that kind of barrier for me.

Josh:
Yeah. And it’s strange because they’re different in their approaches and I won’t call out specific AIs, but earlier today I was trying to get cloudy into an image and one AI told me I cannot recreate brand and logos for you now that you’ve told me this is an official mascot conversation over. And so, I went down the street to one other AI and I was like, “Hey, can you do this?” And it was totally no problems.
And remember, there’s all these guardrails that are totally human created, and there’s a line in one guardrail list and there’s not in another one. So, it’s an interesting mix of possibilities right now.

Mike:
You went down the street, you mean you just turned down the dark alley where they just relaxed their guardrails.

Josh:
Exactly.

Mike:
This will be the example I use for a long time, but I still remember Josh showing up on a team meeting and I forget what you were using, but you were like, “We can just ask it to generate a cat as a pirate on a ship in the middle of the ocean.” What? That was the coolest thing. So, I was all ready to go for AI. Great. Well, it’ll be interesting to see a year from now if we’re telling a bot to update our checking accounts or not.
I had somebody in the grocery store pay with cash the other day, and the poor lady at the register looked at it as though she was being handed Mexican pesos, Kate, because I’m old enough to remember being at the grocery store and begrudgingly writing a check and being like, “Oh, if you just had cash, this would go so much faster.” And now it’s just-

Kate:
Even better, being the person at the cashier stand trying to run the check through the little scanner thing.

Josh:
Right. Oh my God.

Kate:
I remember that.

Mike:
And you had to make the cross, fill things out.

Josh:
Yeah. Yeah. Or the [inaudible 00:16:09] credit card readers.

Jennifer:
Yes.

Kate:
Yes.

Mike:
I was part of the [inaudible 00:16:12] credit era.

Josh:
Thank you. Thank you for that.

Mike:
Yeah. Now, Josh, you just tap it. It’s a chip.

Josh:
You tap it and you go. Sometimes, I just tap it.

Kate:
Here’s your watch.

Mike:
Oh yeah, I know. Yeah. That to me still feels a little magical to tap a phone.

Josh:
Yeah. I moved back in time over the weekend because I was running around doing chores and realized one of my tires had almost like 10 PSI, dangerously low pressure. And so, I went to the most Chicago experience of getting this fixed because the actual garage was an actual garage behind the office building that they reside in. You have to go through their Chicago Bears paraphernalia in order to get into the garage, and then the whole thing turned into just a cash transaction.
There was no cash register. I’d have to go all the way back to the other building in order to do it.

Mike:
Wow.

Josh:
So, I just gave them 40 bucks and got 15 bucks back. And I was like, “When was the last time I got change? Actual change.”

Mike:
Yeah. Well, there’s no pennies now.

Kate:
It costs that much. You need your own air compressor.

Mike:
There’s no pennies. You do, Josh. I can recommend some for you.

Josh:
Nice.

Mike:
They’re not that bad. Well, appreciate you all coming on the podcast and sharing what you do and don’t use AI for, and of course, building really cool AI projects. So, thanks all.

Josh:
Cheers everybody.

Jennifer:
Thanks.

Kate:
Bye. Thanks.

Mike:
Thanks again to Jennifer, Josh, and Kate for joining me. And I think the idea I keep coming back to is the one that Josh questioned about recovery and that’s if your AI gets something wrong, how quickly can you fix it? An email to the wrong person is just one apology, but a bank account being emptied overnight is a different story. And that’s a useful way to think about how different requests could get handed over. And I think it’s also why we saw four people can have four very different answers about what they thought AI would do for them now versus a year ago.
So, if this conversation sounds like something you enjoyed having, thanks for watching. And if it sounds like something someone else would enjoy watching, be sure to share that out on social. We’d love to have more people listen to the podcast, whether you’re listening on iTunes or you’re watching this on the Salesforce Admins YouTube channel. So, with that, until next time, I’ll see you in the cloud.

 

Love our podcasts?

Subscribe today on iTunes, Google Play, Sound Cloud and Spotify!

Salesforce Admin using Agentforce Builder and Agent Script to design AI agent workflows

How Agent Script Is Redefining the Admin Role

Today on the Salesforce Admins Podcast, we talk to Joshua Birk, Senior Director of Admin Evangelism at Salesforce. Join us as we chat about how Agent Script helps admins build more predictable and reliable AI solutions. You should subscribe for the full episode, but here are a few takeaways from our conversation with Joshua Birk. […]

READ MORE

How Are 2025 Admin Predictions Holding Up So Far?

Today on the Salesforce Admins Podcast, we talk to Jennifer Lee, Joshua Birk, and Kate Lessard from the Admin Evangelist team at Salesforce. Join us as we revisit the team’s predictions from the beginning of the year for how Agentforce will change the game for admins in 2025.  You should subscribe for the full episode, […]

READ MORE

Real Talk for Admins on Content, Conferences, and Agentforce

Today on the Salesforce Admins Podcast, we talk to Joy Shutters-Helbing, Senior Manager of Salesforce Practice at Captech and Salesforce MVP Hall-of-Famer, and Mike Reynolds, Senior Product Marketing Manager at Slack. Join us as we chat about creating community content, navigating conference submissions, and their new podcast, The JAM. You should subscribe for the full […]

READ MORE