How to Use AI Agents For Business

Paula: [00:00:00] Hello, everyone. Welcome to another episode of Chatting With The Experts, where I speak with fantastic, phenomenal women from Africa, from the Caribbean, and in the diaspora. Today is no exception.
Paula: And our title is Onboarding Your Digital Employee. We’ll be talking about security and guardrails for AI agents My guest, who will join me in a minute, is the founder and CEO of Nexora Strategy Consulting, and she builds AI operating systems.
A fun fact about her is that she is a 2024 Guinness World Record holder, and she’s had 15 years of experience driving innovation and transformation as a software and [00:01:00] cloud engineer. She’s also an AI consultant and technology infrastructure leader. Wow. She has led global engineering teams and has helped several small to mid-size corporate business organizations leverage AI, cloud, and emerging technologies for scalability and efficiency.
Oh, my gosh, her bio is so impressive. So what I’m gonna do is invite her in to
Paula: the show. So welcome, Jennifer Ponder, to Chatting with the Experts.
Jennifer: Wow. Hi, I’m so excited to be here with you, Paula.
Paula: Oh, my gosh. I could read your bio over and over again. It’s so impressive. I love the part about being a 2024, Guinness Book Record holder. That is fantastic.
Jennifer: Well, thank you. Yeah. It’s definitely been [00:02:00] something I will say, you know, as a little girl who has enjoyed tech all her life and made it to that goal, and just going through, you know, writing that cloud script and going through that particular, um, you know, just competition and just reaching that, that height, it was something phenomenal.
Um, it really stretched me. So excited to be here with you today and sharing um, great information with your community.
Paula: Oh, my word. Did I leave anything out? Because I told you your bio is impressive.
Jennifer: Well, I will say this. Other than, you know, my love for technology, also too my background has been in, you know, the nonprofit space as well, helping to, you know, really excite and empower teens and, you know, just youth and young adults with technology as well, uh, because they are the next generation. So one of the things that I have been happy to do, uh, as a program director and also as a facilitator for software engineering courses as well is to [00:03:00] help young people. Because that’s what it’s all about, to support them. Because, you know, as we know, the technology landscape is changing, and we want to help to build that talent pipeline.
So that’s something, you know, that’s been near and dear to my heart.
Paula: I love that you’re helping young people because young people are our future. The young adults, as I like to call them, are our future.
Jennifer: Yes.
Paula: And, you know, as we talked about offline, AI is here and it’s the future. Yes. So helping them understand, grasp it, and know about it is something that I’m happy that you’re doing.
Jennifer: Definitely. It’s a joy of mine.
Paula: All right. So one of the things I know we’re gonna talk about is AI agents because offline I was saying to you that as a small business owner, um, huh, AI is something that I’ve had to learn, embrace. And every day I’m hearing something new about AI, and then I remember reading up about AI agents and I was like, “What?[00:04:00]
Yet another thing?” So, you know, when I met you and realized what you do, I was like, “Can you come talk to us a bit about, especially as small business owners, about what to do, how to feel safe, how to feel empowered as we surge along in this AI-enhanced world, and to know what we’re doing or what we’re not doing right?”
Jennifer: So yes,
Paula: help us. Help me.
Jennifer: Yes, definitely. Um, you know, Paula, one of the things I told you about my company, I consult, you know, small and mid-sized businesses. Mm-hmm. One of the things that I learned early on in my software engineering path was that there were a lot of small business owners back when I was doing my apprenticeship, you know, they may not could have afforded, um, you know, an entire IT department of you know, um, team of developers or a CTO.
And so I thought, “What if I lent my services to them?” Mm. And so that’s what I started to [00:05:00] do, and that’s no different from what I do in my company, you know, because AI, think about it, it just came onto the scene. A lot of people are still trying to figure out how to
Jennifer: deal with it, how to- Mm … move through this time, because I know there are so many worries out there, but as we know, AI is not going anywhere, and we have to work with it.
But the one thing I want people to know, we can be responsible and ethical AI drivers and
Jennifer: leaders of change, and we can also secure our systems. We just have to have the right guardrails in place. And, you know, now with all, uh,
Jennifer: the digital employees that are on the scene and reducing time and building revenue into your company, you still wanna add that extra layer of security because, as we know, Paula, there’s been so
Jennifer: many things on the news about AI agents going rogue, what’s happening.
They’re getting out of control in certain places. But I don’t want that to scare anyone. I just want to empower you today to think about, what do you have in place? What are the applications you’re currently using? [00:06:00] How are you using this technology? But share with you some ways that you can help to secure your applications.
And really help you to get into the mindset, let’s think about safety and guardrails first. We can use this. Let’s not be scared of the technology, but let’s also empower ourselves to stay safe, um, and use our best judgment when using AI agents in our companies.
Paula: All right. So I hear that. So as a small business owner, even someone who’s not a small business owner, as a regular user, because non-business owners, people at home, people, you know, retirees, everyone’s using AI.
So, like, let’s say I am at home, I’m a caregiver, whatever I may be doing. What, well, okay, I’m gonna roll back on that. Let’s start with a small business owner and then we can expand to anybody who’s probably at home, retired, or whatever. As a business owner, what would you say the first thing to do for an AI agent?
Would you say give it a job [00:07:00] description, um, before you turn on an agent?
Jennifer: I would say yeah…
Paula: would that be the first thing?
Jennifer: Yeah, that would be the first thing. Let’s make sure we’re giving it a job description, because think about it, this is a digital employee.
Paula: Okay.
Jennifer: You know, with a new employee that comes on, think about a human employee.
Paula: Mm-hmm.
Jennifer: You wouldn’t give it access to your passwords, give it the company credit card or all of that first, now, would you, and just leave it alone? No, let’s give it a job description, right?
Jennifer: Let’s put down what we want it to do, what it cannot do, and things that it totally cannot access without human approval.
So let’s put it down on paper first so we can see that, and make it a part of our standard operating procedures and our guidelines first. So you can write down a quick two sentences, what do you want it to do, right?
Jennifer: What if it’s your receptionist and concierge? Okay, it can book appointments on the calendar and send the client a appointment reminder.
It cannot take any payments. You know, I’m just thinking through this with you, just to give you an example [00:08:00] here. And it cannot do anything such as refunds or, you know, just going into any type of protected information without a human in the loop. We wanna make sure that always a human is in the loop. So job description, first thing you wanna do.
Paula: Okay. So give it a job description, just like you’d give a human employee.
Jennifer: Right, yes. Mm-hmm.
Paula: Very good then.
Jennifer: And I was gonna say too, onto that job description, make sure also within that you are letting your team know. Because if you’re not a solopreneur, we wanna make sure that we are including everyone, because this is how your team is gonna get best trained with dealing with digital employees, you know?
Jennifer: What I find is there’s a lot of, you know, misalignment when you have some team members know about AI, some team members can handle a agent. We wanna make sure that all team members are aware and on the same page of the [00:09:00] security measures and of the job descriptions of the agents that we’re putting in place.
Paula: That’s helpful. So no matter how small your team is, whether it’s one or two or more, let them know about the new… Well, yeah, like onboarding a new employee. Do the same thing with an AI agent. Let your team know. Introduce them, the AI agent to your team, in other words.
Jennifer: Yes. And, and that way, if only one person’s using it, you know, what if that one person gets locked out?
Mm. That password, something happens, and only one person has access. That’s not gonna be good for the team. So we wanna make sure everybody knows. Now, only certain team members now have access to the agent as well. Someone who may be, you know, not dealing with, you know, operations, you may not wanna give them access to the agent, but they still need to be aware of agents on the team because there’s [00:10:00] gonna be some people who have, you know, those permissions that are gonna deploy the agent, look after the agent, monitor the log, see what’s going on in the background.
You know, we wanna make sure as well that we are setting permissions for, as we do employees, our human employees, we’re setting permissions. And that’s the next thing we’ll get into, I’m sure, shortly here about, you know, really what’s the next steps as far as protecting your other information and applications.
Paula: Right, right, right. ‘Cause that was gonna be the next question. What about privileges? You know, because every agent, I mean, every, like an employee, we may all be employees, but some people are gonna have the key to my office, and some are not. So does that, I mean, apply when it comes to the AI agent as well?
Jennifer: It does. It does. Because what I always like to share with business owners, you know, you don’t wanna give your agent access to everything. As I used the example [00:11:00] before, if you have your agent that is an agent that books appointments on your calendar, okay, this particular agent has only access to a certain calendar and maybe certain information that could be a client’s name and email only to send them appointment reminders.
Now, when it starts getting into you know, the settings, you know, where you go into settings to use connectors to connect your AI agents, make sure you are careful about that. Do you wanna have certain emails or certain, like if you have QuickBooks for, you know, managing, you know, your accounts receivables and things like that, do you want your AI agent to access those things?
Let’s think about that first, um, even as a next step under the job description, but we wanna make sure that we’re always sandboxing our agents as well. And what I mean by that, you wanna make sure that you have a, a folder set aside that only your agent [00:12:00] can you know, work in. Also, you wanna have a helper folder, so if there’s any data or any, you know, um, forms or documents that the agent needs to work through, like your company SOPs, because your agent may be writing some type of documentation, blogs, or PR statements for you, but you wanna make sure that you are checking behind the agent as well.
Jennifer: But I always say you wanna separate your agent so that it’s only working in sandbox folders, because if you’re giving your agent access to email, calendars, you know, um, tax information, you know, it can really get away from you and start to doing a whole lot. So this keeps guardrails in so you can monitor it in the folders where you need it to be, and only granting access to the applications that you want it to have.
If you’re using, of course, I know, Paula, we’ve talked about it, Calendly and Zoom and setting appointments for you, but, you know, you wanna make sure that you’re always checking behind it. And so I always empower [00:13:00] business owners, do a two-week pilot program, right? So when you’re giving it these permissions, this is helpful for you to know, okay, what did the agent miss?
What is it doing? What did it process? How did that work out? Okay, if this worked well, let’s add one more thing in, or let’s add another application. We don’t have to give it a whole lot of work at one time. We can monitor it progressively.
Paula: Okay. So now we talk about privileges, but for most small business owners, you know, we talked about this offline how COVID made a lot of small business owners, it created a lot of small business owners, and it created, made small business owners their own IT departments.
Jennifer: Mm-hmm.
Paula: We chuckled about that. Without any training, you know? Mm-hmm. And so now we hear about giving least privilege access. Some of us, I’m gonna include myself in it [00:14:00] just for the sake of making things all equal. So we have, we may have, like, software. We don’t even know, you know, the back ends of the software.
We’re assuming that, ’cause there’s so many things you have to give access to, privilege to. You don’t really know the fine print. How do we know that we are given least privileged access? How do we know? Is it through a prompt? Is it through, like, actually going into settings of the particular um, software that we’re giving it privilege to?
Am I asking too many questions? Does that make sense?
Jennifer: No. This is great. This is great because what you wanna do, you wanna make sure you’re going into the settings, right?
Paula: Mm-hmm. Mm-hmm.
Jennifer: You wanna make sure that you are only checking the boxes for what it needs to do. I always say have a password protection, a password manager.
So if you need to keep safe your API keys, how your applications talk to one another, any, you know, like I said, private, uh, passwords or anything, you can keep that in that [00:15:00] password manager. But I look at it like this, and if some of your audience may be cloud users, you know, IAM is where you protect these type of things.
You can give it its own password to access certain applications if you’re in a cloud environment. Even if you’re just, like you said, a small business owner that you’re doing everything on your own, make sure in these applications you’re going into settings. You’re only using these applications, using that particular agent with the application and what you need it to do.
Because least privilege, that’s gonna help you and help that agent to still have guardrails and still stay grounded in the particular work that it needs to do. So I’ll give you an example. A lot of times if we are going into, uh, I’ll just use Cloud Cowork, for example, and you look at all the connectors now, and MCPs are everywhere now.
And think of that as just kind of like an [00:16:00] adapter, how AI you know, can talk to, you know, the different applications these days, the plugins, um, to be able to work with your information, right? And so when you think about that, you’re like, “Oh, wow, I can connect up to all these applications,” but what data is getting out there?
Jennifer: So when you think about the applications, you wanna make sure that, okay, if we go into settings, let’s only check the boxes for what it needs. Also, the agent has its own particular password. It’s not using passwords of an administrator or any, you know, ones of your human workforce. That way, this is another way that provides you with a step-by-step log and monitoring of when the agent logged in, what did it do.
It’s giving you logged activity so you can keep that step by step. So we always wanna make sure zero trust you know, um- is the first line of defense. And what that [00:17:00] means, it’s not saying you can’t trust your agent, but let’s just make sure at first we are giving it the least amount of privileges first so it’s not going into different folders, um, that you don’t know about.
And in the world of software, you know, we know zero trust as this is our least amount of privilege. This is the height of what you can do and what you can’t do early on. So we wanna make sure we keep that top of mind, is always just give it its own password, give it its own set of instructions, and also when we’re going in the applications on the back end, we’re making sure within settings that we are only checking the boxes for what the agents, they can and cannot do.
And keeping that log available for your team members as well. And if you’re a solopreneur, just keep it as a guide, you know, one of your SOPs when you’re working with whether it’s a chief of staff agent, your sales [00:18:00] agent, marketing agent, and I know we’ll get into that later, but just all of your agents need to have separated security standards and procedures.
Paula: All right. Those of you in the audience, when she says marketing angels, agents I mean, and sales agents, she means AI marketing agents.
Jennifer: Yes …
Paula: AI sales agents. She’s not talking about humans now. We’re talking AI. So you’ve got to listen carefully, because a lot of small business owners are getting into this, and you’ve got to be vigilant.
You’ve got to pay attention. This is not to scare you, this is just to inform you, because I’m learning too. I am learning too. And as I mentioned a few minutes ago, um, and as you said, a lot of the AI models are getting updated regularly more frequently than before, and they all have all these plugins.
Even the plugins, you need to check, [00:19:00] as you mentioned, I think, a few minutes ago, you need to go check the settings of those plugins before you give your agent access to it. Because some of these plugins, when you look at the settings, they are training. They’re using your data to train.
Am I correct or not?
Jennifer: Yes, but they’re in settings in ChatGPT and Claude, and if even if you’re using other models, you can go into the settings and click the lever, you know, you can toggle it off to not train on your data. And here’s another best practice I always tell anybody. Personal health information, any Social Security numbers, date of birth, just please leave that out of your AI.
I know some people have been going in and getting information about their healthcare and things of that nature. Now, do that at your own risk, but I don’t really share that that is a best way to go. So far you still, you know, this is [00:20:00] still your thought partner.
Jennifer: We always wanna know that AI is still AI.
It has the ability to hallucinate. It has the ability to get things wrong. Let’s make sure that we are using it in that way as our thought partner. Now, you can give it information about your business and, you know, different things that you are working on. But let’s make sure that we’re not putting our address in, any personal tax information, date of birth, Social Security numbers, things of that nature.
That also helps you to stay safe, other than, you know, toggling it off not to train on your data.
Paula: Okay. All right. What about, okay, some of um, the plugins that we’re connecting with, that we may be connecting with Google Drive, we may be connecting- Mm-hmm … with LinkedIn, we may be connecting with Salesforce, et cetera.
It’s imperative that we go look at the settings of those ones as well.
Jennifer: Definitely.
Paula: ‘Cause a lot of them have their own AIs as well.
Jennifer: Right. And it’s one thing that you can think, “Okay, [00:21:00] I just turn on the toggle for this and it’s fine.” No, you always wanna go back periodically and look at those settings also.
And if there are reports or different things that you’re running, uh, that are, that are current, you know, make sure to keep checking those. But if you’re no longer using them, disconnect it. I would say disconnect it from your agent. Make sure your agent is logged out of it. Make sure you’re logged out of it.
And anything that, you know, that can clear the path. Because if you, the last time you used the application may have been a year ago, and you notice, you go back and say, “
Paula: Hmm,
Jennifer: just got a email update,” or, “I got pinged about something that happened with my agent,” and you’re like, “I haven’t used this application in over a year.” that’s gonna be quite scary because you don’t know what happened and what didn’t happen. And it’s not to say that something… I don’t wanna scare anybody that something happened. It’s just you wanna always keep top of mind, what is your AI agent accessing? And just even before you get to the agent piece of it, if you’re not [00:22:00] currently using agents, but what is that application accessing about me?
What data is it finding out on the back end? So we always wanna make sure that if there’s any applications that we’re connected to or we’re not using anymore, we’ve discontinued our subscription, let’s just make sure we cancel it, make sure we go into our own settings in C Drive, and make sure that any connections have been disconnected.
Paula: Good point. Good point, Jennifer. Yes. Keep checking because things are getting smarter.
Jennifer: Yes.
Paula: Things are getting smarter, smarter. Okay, let’s go back again to our agents. As a small business owner, you talked about a probation period for AI agent. So as you said, test it out for about two weeks and see how it works.
Jennifer: Right.
Paula: Or whatever. Um, Do you have… Just thinking probation, yeah. Like we would do with a human employee, [00:23:00] correct?
Jennifer: Right, yes, definitely. Because what you wanna do is fine-tune that agent, right? You wanna make sure, just like how you would with a employee, when you’re looking at, okay, when they start, they may need some guidance, right?
They may need some things that, “Okay, I see you did this well, but let’s work on refining your process here. Let’s work on it.” And it would be the same for a AI agent. Let’s think about, what are we giving it access to. You know? And that’s why I said, you wanna go through, when you’re giving it the job description, let’s write out what it can actually do, what it cannot do, and what it should never touch.
And so the things that you give it access to during this pilot, and test probationary period, okay, look at how well did it work. You know, what did it do wrong that you didn’t want it to do? Okay, let’s change the trajectory. Did it access anything that you didn’t want it to access? And also look, check the logs.
What is it accessing? You know, Was it correctly, you know, did it deploy well? You know, [00:24:00] how it’s working out. And so that doesn’t mean after the two-week probationary period, if you want to stop using it, you can. But if you want to, you know, keep using agents, because now we know, and in my company as well, I set up multi-agent systems, multi-agent operating systems now, and we know agents are a thing now that’s in the workforce.
Jennifer: Whereas you had maybe 10 to 15 years ago, we have workforces of 10 to 15, uh, 20 or more, um, human employees on a team, now you’re gonna see there’s agents and human employees on a team working together simultaneously. So we wanna always keep that top of mind that, you know, we don’t have to just totally do away with the agents.
We wanna make sure we keep refining them, but putting those guardrails in safely first. And so that’s why I say it’s good to do a pilot with them first, see how they’re working. Make sure they are accessing the correct things and they’re not moving, you know, out of the, out of the scope of what you want them to do, [00:25:00] because the agents are getting smarter, right?
Because the models when they first came out, and I’ll just give you this example here, um, you know, ChatGPT got released November of 2022, right? And so along that time when models were coming along, they may have had updates every three or so months, you know, and that started to speed up to two months, and one month, and every week.
And I like to give the analogy, it’s like a tennis match now. There’s an update here and there, and I’m sure most of you all who may have seen or may not know about the OpenAI you know, agent that recently came out, and you know, that Meta Muse that’s out now. Yeah. And it’s so many agents out here, Hermes agent that you know, MIT, its open source agent by MIT.
It’s so many agents and so much going on that the models are getting smarter and smarter every day, and we just have to know how to work with them.
Paula: And we are helping them get smarter, ’cause we’re giving them so much information. I know people who sometimes upload a whole [00:26:00] video to the AI.
Jennifer: Right. Yeah. Yeah. But it’s definitely, you know, as I say right now, you know, we all know we’re in a time where AI’s not going anywhere. Mm-hmm. You know, we just have to learn better ways, safer ways to work with it. You know, I know it can be a little scary when you’re trying to figure out, “Oh, I don’t know if I’m ready to take that next step”, but it’s where we’re going.
But I want people to know that you can take the right step, and you can do it safely. You don’t have to employ five or 10 digital employees in your company if you’re not ready. Just pilot one first, see how it works out. Keep training it, you know. Um, keep discovering what’s possible. Put those guardrails in place, like I said, you know, what it can do, what it can’t do, what it should never access.
You know, have that written down for yourself, you know, that you’re always going back to, so it’s kind of top of mind as well, because it’s one thing that we have to continue to train ourselves on as well. Let’s be our own, you know, AI ethical leaders that we’re [00:27:00] putting those guardrails in first so we’re not worried about at the end of the day, “Oh my God, did this company do this?”
If you’re a solopreneur or a person who has a rather small team and you’re still managing the technical piece, uh, for yourself or maybe for a small team, like one to five people, you know, let’s empower ourselves to make sure we are doing that security measures first.
Paula: Let’s empower ourselves to make sure we are doing the security measures first. Yes. Important. Don’t be scared, just empower yourself to make sure that…
Jennifer: Right …
Paula: you’re doing those security measures first. Thank you, Jennifer. Thank you, thank you, thank you. So of course, we’re gonna open up the floor for those who have joined us, but for those listening to this later or viewing this on YouTube, how can people get in touch with you?
Jennifer: Yes, definitely. You can follow me everywhere. Um, Of course, Jennifer Ponder Speaks, and I am on Instagram. Certainly, you can [00:28:00] follow us, Nexora Strategy Consulting. Of course, we’re on Facebook and Instagram as well. Um, certainly happy to speak with anyone on LinkedIn. You can find me under Jennifer Ponder CSM.
Happy to address any questions and follow up and, and just speak to anyone. I love helping people and, you know, my love for helping people to understand technology, to make it, you know, just like building blocks. I wanna make this easier for people because technology is not going anywhere, and just for me to be able to…
If I can change one life, if I can help someone to implement and make their day easier with AI and technology, you know, I feel like I’ve done my job. So I love this, and I love helping people to, you know, just empower themselves to use these tools because technology in the right hands, it does the best things, but we all know we wanna be good actors, um, and stewards of this technology too, AI.
Paula: Absolutely. Absolutely. And for those of you who [00:29:00] will listen to this, just like Jennifer has been on this show, you can reach out to me if you’d like to be a guest. My website is chattingwiththeexperts.com. I am on LinkedIn. My business page there is Chatting with the Experts. I’m also on Instagram. My handle there is @chat_experts_podcast.
We have a YouTube channel now. Please subscribe to us there. You’ll be able to see this, view this episode that, um, Jennifer and I have done. It will be there, as well as all the other past episodes with women who are just as phenomenal as Jennifer, talking about a variety of subjects matters that are interesting and are very relevant to those of us women in the Diaspora in Africa and in the Caribbean.
And we just recently opened up a TikTok account, so subscribe to us there as well, and [00:30:00] it’s @- Um, our handle there, I don’t know if it’s handle because I’m so new at this, is chat_experts_podcast on TikTok. Thank you, Jennifer. This has been so, so, so interesting and of course very, very helpful. And now we’ll open up the floor to all those who joined us so they can ask you more questions.
And those of you who are watching this on our live stream on YouTube, we invite you to use the link that I’ll drop in YouTube and the live stream so that you can come into the room and speak with Jennifer, because she is amazing and very, very knowledgeable. So thank you, Jennifer. Thank you. Thank you.
Thank you. Well, thank you, Paula.
Jennifer: It was my pleasure being here with you and the community.
Paula: Thank you.