S1. Ep42 - Why Only 6% of Businesses Trust AI Agents (And How to Fix That)

Katie: Welcome to the AI Automations for Business Podcasts. We're your hosts, Katie and Null, and we'll be discussing how you can use AI for your business along with the latest news, updates and automations to help you stay ahead of the curve, allowing you to grow and scale your business more efficiently. Please be sure to subscribe and I hope you enjoy this episode. Hello, welcome back to another episode. Hi, hello. I'm Katie. And as always, I've got Noel here with me today. How are you doing, Noel?

Noel: Oh, amazing as always.

Katie: That's good to hear.

Noel: Yeah.

Katie: So, no, we were having a conversation about something that we read from Harvard Business Review, and we thought this is going to make a really interesting podcast episode. Definitely, yeah.

Noel: Yeah, I couldn't quite believe what I was reading, although I could in some ways.

Katie: Yeah. So in a recent Harvard Business Review survey, only 6% of organisations say that they fully trust AI agents. So that's what, 94% of businesses don't fully trust AI agents.

Noel: Yes.

Katie: To be honest, I didn't think it was going to be like only like 6% of organisations that fully trust them. I for some reason I thought it was going to be like a lot higher not like over 50% or anything but I did think it was going to be a lot higher than 6%

Noel: yeah I was hoping for at least 3040 yeah that's that sort of area maybe 20s but yeah six I was like hey yeah crazy yeah but I guess before we kind of delved the deeper into the episode. I guess we kind of need to take a little step back and then, you know, figure out what an agentic AI is for anyone that's brand new to this. So some people will maybe just use in chat GPT was I think we just need to discuss what an AI agent actually is. And essentially what it is is a large language model, which has access to tools. And it makes up its own mind on how it does things. So you could give it access to like a Google sheet, for an example. And you could have tools there, which would be search for rows or add rows or anything like that. So each of those would be a different tool. And then when you give it a plain language prompt, so you say, could you add in a row that, you know, for X, Y, and Z, and then that agent would then figure out how it does that task. So it's quite fluid. So you do need to be kind of strict on the tools you give it access to and, you know, your system prompt for that agent so it understands. But essentially an agent or gentic AI is, you know, figuring it out for itself, which is a little scary. You're letting it loose to do things. And, yeah, I can see why there is that distrust. But when you do it right, do it properly, then, yeah, you're golden. But that's a rough overview of what agentic AI is.

Katie: Okay, so let's talk more then about that trust gap because I think it seems that the organisations that were surveyed for this Harvard business review only really trust agents for like routine limited tasks or like supervised tasks.

Noel: Yeah, yeah. And that's a good place to start. you've never had an agent within your business, then that's a good place to start. You'll give it something that you know that's routine, you know, and you know what the expected outcomes should be. So when you start seeing those outcomes from the agent, you can easily spot whether it's doing right or wrong and then go in and correct it or whatever if it needs to. So that's fine. That's a good starting point. But yeah, you do then need to start then, you know, building upon. that so you know you're giving it a little bit more access to things and then you know giving it a little bit more trust than you would normally but obviously you only do that in like areas where you know it's not going to destroy anything within the business it's not going to you know send stuff out to clients that shouldn't do and all that sort of stuff so yeah

Katie: it's kind

Noel: of starting small and building that trust step by step

Katie: yeah it is expected um that investment in agentic AI over the next two years is to increase by 86%. So people are really putting money into this, but somehow the trust isn't following.

Noel: Yeah, that's also an odd stat. So the first stat that's shocking me was 6%.

Katie: And then it's 86% more.

Noel: I was like, okay, I mean, I get why most businesses, you know, big and small, want to have AI agents within there. They see the benefits, but if you're not implementing it in the right way, then you're basically going to waste that investment or

Katie: not get

Noel: the most out of that investment.

Katie: Yeah. You know,

Noel: it could increase your productivity by 10%, but if you did it properly, it could have done it by 60% or

Katie: whatever. Yeah.

Noel: Yeah. Interesting, there's a lot of money about this.

Katie: Yeah. Yeah, it's almost this like stat and, you know, you know, like the 86% increase of investment going into it and, you know, people are not trusting it. It's almost given me like NFT vibes and I don't really know why.

Noel: That's very true. I get that, yeah, yeah. I can see that.

Katie: It's given me that sort of. buy you.

Noel: Yeah. Do I want to buy this thing of a monkey? Yes, I probably will. I don't quite trust that I'm going to be able to sell it, but I'm going to buy it anyway.

Katie: Yeah.

Noel: Yeah. But this is going to be around a lot longer than NFT.

Katie: Let's hope so. Let's hope so.

Noel: Yeah. Yeah.

Katie: Okay. So I just thought we could talk through some of the, concerns that the report made so people were saying you know who was surveyed for this Harvard business review that their concerns were cyber security and privacy so like is this going to leak my data or like do something rogue

Noel: yeah that that's a very very common question when you do anything with AI but the majority of the age of if you've built them on like a no-code platform or you've built them in code, then you're using the API for those particular AI models that you want to use for that use case and, you know, obviously check their terms and policies and all that sort of stuff. But the big ones like Anthropic and Open AI, they're all fully secure, they're fully encrypted, they don't look at the data, they don't train from the data. So it is, it's fine to use, it's just that there's not as much education. in that sort of area for people to fully trust these AI providers. Okay.

Katie: Another concern was data output quality, so hallucinations and wrong answers and business workflows.

Noel: Yeah, and I think we could probably attribute most of that to either feeding the agent the wrong context. So your system prompt might be wrong or where it's called the wrong tool and it's bought in data that it probably shouldn't have done to give the answer, even though it thought that was the right thing to do. So, yeah, that's where you're going to get those sort of hallucination and errors. But Wi-Fi now with the later models that we've got now at the end of 2025, the hallucinations is better.

Katie: Okay.

Noel: But it's just making sure that we've got the right context going in. Otherwise, you're just never going to get the right answer. Yeah, okay.

Katie: And then process is not ready for automation. So maybe the process itself isn't messy. They haven't got like a SOP. It's undocumented.

Noel: Yes.

Katie: It surprises me a bit, if I'm honest, with that one. I don't know. I don't expect most businesses or organisations to have some sort of organisation. I don't know.

Noel: Yeah, I think some businesses have the business processes, but is that, it's kind of like, is that ready to add an agent into it? You know, is it properly defined enough? You know, if your team is like going through a process and then they often go off and do their own thing halfway through, then that's probably not going to be ready for automation because your AI agent is going to do the same. and then you'd be like, what's it doing?

Katie: Yeah, that's a really good point to make. Okay. And then the other top concern from the report was tech limitations. But this is not like using tech, it's their systems not able to talk to one another.

Noel: Yes, yeah. If your system is, you know, completely away from the internet, it's not a cloud service and things like that, then, that's when it can become kind of difficult to then start, you know, adding in agents, things like, because you're, you know, you're going to have to, you know, either host a large language model in-house or you're going to have to find a way of getting around business firewalls. It gets kind of messy. So, you know, I agree, you know, yeah, there is technical limitations, but if your stuff's already on the cloud, then there are, you know, secure ways of getting around that sort of thing. So, yeah, that does negate that.

Katie: I find as well, like, when it comes to tech limitations, a lot of what people are advertising or promoting agents are actually just automations with a chatbot.

Noel: Yeah. I do see this quite a bit. Yeah, they'll say that something is, an agent and then yeah you look at it you think well that's that's an end-to-end

Katie: automation process yes it's got

Noel: an AI in there but the AI is taking something in it's giving the same thing out every single time so that's not that's not an agent it's not there's no thought process in there for it to say well I'm going to choose this route instead of that route it's got one route

Katie: yeah and I know we talked about the difference between AI agents and automations in episode 14.

Noel: But can

Katie: you just kind of do a brief overview to define what is an automation versus what is an AI agent just so if anyone is listening who is new, they've not listened to episode 14, you know, they've got a clear understanding because I feel like those lines, can get blurred very easily.

Noel: Yes, and because it also doesn't help that the word, or words AI agent, are kind of like massive buzzwords. So everyone thinks that they need an AI agent because that's what everyone says.

Katie: Yeah.

Noel: That's not true either, really. So if your process is, you know, a repeatable process. So let's say you're going to take like a transcript from a call, and then you're going to then go off and then create, you know, like transcription notes and then maybe send like a quotation back to wherever you had that call with. That doesn't need to be an agent. Yes, you'll have AI in the middle of that, but, you know, you're going to have the same thing coming in and the same thing coming out, and that's what you want every single time.

Katie: You know, it's quite important.

Noel: That bit's right. So for that, that would be an automation. whereas if you had like an agent you could say let's say you attach your agent into slack and then you have like a database in air table or notion you could then say to well you know what's on my to-do list in notion and then it would then use tools that it's been given access to to then go right well how can I what where can I get that information from that's best going to answer this question so I'll probably go you're going to use the search tool that probably a good place to start and then it might then go off and get information and that sort of stuff so it's it's thinking about what it's going to do but it's never really 100% repeatable in some cases sometimes I would say like 99% of the time it's going to be fine providing your system prompt is good if it's not good then yeah you're probably going to run into trouble but yeah if it's good it's going to be it's going to be perfectly fine But if you need something where you need to talk to something, like you would do with like Chuck GBT and have a two-way conversation with a bit of history in there, or you know, you're happy for it to go off and do things in its own way, then that's where you need an agent.

Katie: Yeah. Yeah.

Noel: But if it's easy, same thing in, same thing out, automation every day of the week.

Katie: Yeah, so automation is fixed steps. It's predictable, input, output, great for when X happens, do Y and Z exactly this way. Where is an AI agent, it has judgment, interprets, contact, chooses actions, may use tools, yeah.

Noel: You can also put in some logic as well into your automation. So you can, so like within making an NAN, you can have different routes within your automation. So if something is mentioned within that call, it might send it up one route or it might send it down a difference. So there's different logic that you can build in,

Katie: but it is fixed. So yeah,

Noel: whereas an agent would just figure it out itself.

Katie: Yeah, okay.

Noel: Hopefully.

Katie: The AI agent, you're kind of letting it loose a bit more.

Noel: Yes.

Katie: Whereas the automation you have complete control over it.

Noel: Yes. Yeah, absolutely. Okay.

Katie: So I guess that's why, like, a lot of people don't fully trust it, isn't it? It's, it's not the fixed outcome. It is kind of letting it.

Noel: Yeah. Yeah. They've probably put an agent in where they should have an automation.

Katie: Yeah.

Noel: It was mentioned in that report, actually, where they were saying, look, you know, a lot of repeatable processes, we couldn't trust it to do that. And it's like, well, you shouldn't be using an agent. for that. But then all comes out, this is all new to everyone. You know, this is something that's been around for 20 years. So, yeah.

Katie: It's new. And I guess a lot of people are learning all at the same time as well, aren't they?

Noel: Yes, exactly.

Katie: Okay, so let's have a look at, like, why agents fell.

Noel: Mm-hmm.

Katie: So what would you say is, like, the biggest thing, like the biggest reason why, like, most agent projects would fail?

Noel: I guess one of the big ones is, you know, garbage in garbage out. You know, that's, you can't expect an agent to be a miracle worker. Yes, it's very clever. You know, it might be able to figure things out. But, you know, you've got to make sure that the information that you're giving it and the instructions that you give it are, you know, absolutely spot on. You know, you really need to get in and test it as well. So you need to think, so if you've got like a team, you might have somebody that will put lots of thought into the question it asks, and then you might have someone that's just going to ask a little five-word question and they expect great things, you know. So yeah, you've kind of got to test all of those bits out. But yeah, making sure that you're giving it the right context, the right tools as well. You know, you don't want to be giving it access to the wrong things or giving it. access to the wrong like database and things that so you might have multiple and you'd be thinking well why is it talking about this i want to be talking about that but yeah if you've got that wrong then you will get it will get the garbage out unfortunately yeah

Katie: yeah and we say this a lot as well even with like the prompt engineering like if you're putting in like you know a really basic prompt then you're only ever going to get a very basic answer yeah

Noel: it's also important important to put it within your agent prompts is like expected outcomes.

Katie: So if somebody asked this,

Noel: I want you to go and use this tool, follow by that tool, and then give the response. So if there's things in there where you think, well, that should be fairly predictable, then bake that into that prompt. Then yeah, it should. Depending on the AI model you've chosen, of course. It should follow those instructions. Yeah.

Katie: Okay. What would be another reason why AI agents fail?

Noel: I guess it comes back to the no clear process. So if you're expecting it to, you know, go off and do something for your business, but really you can't or you and your team can't do that process effectively already, then, you know, it's never probably going to, you know, meet your expectations. You're going to be expecting to do, you know, produce great. things, and then, you know, it's not quite going to fulfill that need. And you'd be thinking, well, why isn't this work? But really comes down to, you know, understanding exactly what you want it to do.

Katie: Yeah.

Noel: So, yeah.

Katie: So I guess as well, like when people say, oh, you know, these AI agents are going to take my job, they're not really because someone still needs to be able to understand, you know, what the outcome is to be able to even set them up in the first place.

Noel: Exactly, yeah.

Katie: Yeah.

Noel: Definitely. You've got to have someone there to tap the keys to ask a question.

Katie: Yeah.

Noel: Or talk to it, whatever. Yeah, yeah.

Katie: Very good point.

Noel: Yeah.

Katie: Any other reasons, no, why AI agents might fail?

Noel: They also, there comes also down to, like, guard rails and things that. So you can, like to say, you can put guard rails within, like, the system prompt. So you could say, well, when someone asks this sort of thing, only do this. So, you know, that's really, really important. You don't want it to go off and use a ton of tools instead when it could have just used one. So I had an example of that when I was testing an AI agent in Make.com. And I was using an open AI model. and it was just going off and it was using like every single tool in its arsenal to answer the question. I was like, what's it doing? And I thought, I didn't want to blame make for this sort of error. I thought, well, maybe it's something decoded. And I switched it out for Anthropic and then used one of their models. And it just went, yeah, cool, yeah, I'll just use that tool there. And then that gave you the answer. And I was like, so yeah, having those sort of guardrails and things out is really important. also helps with like prompt injection as well. So, you know, you could have people with nefarious means would want to maybe, you know, get the agent to do things that it's not supposed to do or access things that you shouldn't do and things that. So, you know, really important to have that sort of thing built in. And I know NAN have just brought out a module within their tool that does guard roads.

Katie: I've not

Noel: tested it out, but it is in there. And I'd expect, I'd expect make to do something fairly similar, although theirs is pretty strict on what it does already. Yeah.

Katie: And I guess like with Maya coming out as well, maybe that would help with the guard rails?

Noel: So yeah, Maya's just going to help you build the automation with the agent in it. But, you know, they do have tools to help you out with like the system prompt. So you could say, well, rewrite it and then add in there, I want it to only do this and this there.

Katie: Yeah.

Noel: And then it will, you know, they produce like really, really good system prompts when you use their AI to do it. Yeah. And I guess the other things of why they fail is you might give your agent too much. So you might say, well, look, you know, I'm going to give you access to, to Notion. I'm going to give you access to Slack. I'm going to give you all of this information from all these different locations. I'm going to dump it in one big super agent. And at that point, you would have to really nail down the guardrails in your system prompt. But at the same time, you're kind of giving it too much. So if somebody comes in with a vague question into the agent, then it's going to go down all kinds of crazy different routes to try and answer your question as best as it can. So, yeah, what it would say is, you know, have agents that are specific. So what you could have is like a classification agent. So that's the one that the user talks to. But then that classification agent then has access to like three others. So one could be going off to notion. One could be going to Slack. So when you ask a question about Slack, that first agent talks to just the Slack one. And then it then gets the response. then it's fine so yeah yeah definitely if you're getting it too big think of ways that you could probably cut that down and then link them together yeah okay top tip that

Katie: was great top tip thank you no you're

Noel: welcome um

Katie: okay so only six percent of businesses fully trust agents today not because AI is doing something wrong or just, you know, going off and doing its own thing, but it could be that, you know, maybe the processes aren't in place. RELs are missing. People are jumping to agents thinking that they need an agent and, you know, sometimes actually all they need is in automation.

Noel: Definitely. That's the big one. Yeah.

Katie: Yeah, for sure. Like I said, if you haven't listened to, episode 14 where we go through the difference between an AI agent and AI automations then definitely check that out and of course we will leave the link below for you as always we appreciate you so much for listening if you like this podcast please leave us a review or you know leave a rating because that really helps us in creating more amazing content for you if you've got any questions that you would love for us to discuss or answer on the podcast, then please get in touch. You can email us, hello at makeautomations.aI, or you can come over in our free LinkedIn group, which is called AI Automations for Business. Thank you so much again, and we will catch you next time for another episode very soon.