Skip to content

HSBU - EN - HubSpot AI - Explain it to me like I'm 5

18 mins read

AI & HubSpot: The Future of Smarter Marketing in 2025

Artificial Intelligence is revolutionising the way businesses operate, but how does it apply to HubSpot? In a recent episode of Avidly Talks, Emily Yates and Annka break down AI in simple terms, exploring its practical applications, common misconceptions, and the importance of high-quality data. They discuss how AI enhances automation, predictive analytics, and personalisation—while emphasising the need for human oversight. Whether you're just starting with AI or looking to refine your strategy, this conversation offers key insights to help businesses navigate the evolving AI landscape. Let’s dive into the major takeaways from this discussion.

Listen to the Full Episode here

Listen on Spotify

Takeaways from this episode:

  • AI is essentially automation with a brain, trained on vast data.
  • Understanding GDPR is crucial for companies in the EU when using AI.
  • Data quality is foundational for effective AI implementation.
  • AI tools should be tested over time as they evolve.
  • Human oversight is necessary to ensure AI outputs are reliable.
  • AI cannot replace the need for human expertise in specialized tasks.
  • Companies should explore existing tech stacks for AI opportunities.
  • AI can automate various sales processes effectively.
  • Training is essential for teams to feel comfortable using AI.
  • AI adoption requires addressing fears and misconceptions among team members.


Chapters 
00:00 Introduction to AI and HubSpot AI
02:32 Understanding AI: Simplifying the Concept
04:21 Practical Applications of AI
06:40 Getting Started with AI in Business
10:27 Identifying Suitable Tasks for AI
13:43 The Importance of Human Oversight
17:08 Common Misconceptions about AI
19:55 The Future of AI and Its Implications
23:21 Encouraging AI Adoption in Teams
25:52 Leveraging HubSpot AI for Sales Processes

Demystifying AI in HubSpot: Insights from Avidly Talks Episode 149

Artificial Intelligence (AI) is transforming the way businesses operate, and its impact on marketing and CRM systems like HubSpot is no exception. In a recent episode of Avidly Talks, Emily Yates and Annka explored the evolving AI landscape, breaking down its applications, misconceptions, and best practices for implementation. Their conversation provides valuable insights into how AI can enhance business processes while emphasizing the need for quality data and human oversight.

1. Simplifying AI: What It Really Means

For many, AI remains an abstract and intimidating concept. Emily and Annka take a practical approach, explaining AI in simple terms and illustrating how it applies to everyday business tasks. AI is not about replacing humans but rather augmenting efficiency by automating repetitive tasks, analyzing large datasets, and providing actionable insights.

2. Practical AI Applications in HubSpot

HubSpot is continuously integrating AI-driven features to improve marketing automation, customer engagement, and data management. The discussion highlights several key areas where AI is making a difference:

  • Chatbots & Conversational AI: Enhancing customer support and lead qualification.

  • Predictive Analytics: Helping marketers forecast trends and customer behaviors.

  • Content Generation: Assisting in drafting emails, blog posts, and social media content.

  • Personalization: Customizing user experiences based on data-driven insights.

3. The Importance of Data Quality & Human Oversight

One of the biggest challenges in AI adoption is ensuring that the data feeding these systems is accurate and reliable. Emily and Annka stress that AI is only as good as the data it processes. Businesses must invest in maintaining clean, well-structured data to get meaningful outputs. Moreover, human oversight remains crucial in interpreting AI-driven recommendations and making strategic decisions.

4. Addressing AI Misconceptions

Many businesses hesitate to adopt AI due to myths surrounding its complexity and reliability. The conversation debunks common misconceptions, including:

  • AI will replace human jobs entirely (reality: it complements human work, not replaces it).

  • AI is infallible (reality: it requires high-quality data and human validation).

  • AI is only for large corporations (reality: AI tools are increasingly accessible for businesses of all sizes).

5. Encouraging AI Adoption in Teams

For AI to be successfully integrated into an organization, employees need to feel comfortable using it. The episode outlines strategies for fostering AI adoption within teams:

  • Education & Training: Offering workshops and hands-on experiences with AI tools.

  • Start Small: Implementing AI gradually in small, manageable ways before scaling up.

  • Demonstrating Value: Showcasing AI’s ability to reduce workload and improve efficiency.

  • Encouraging Experimentation: Allowing teams to test and explore AI-driven solutions.

6. GDPR & Ethical Considerations in AI Implementation

AI comes with responsibilities, particularly when handling customer data. Annka underscores the importance of understanding GDPR policies and ethical AI practices. Businesses must ensure compliance with data protection laws and maintain transparency about how AI is used in their operations.

Conclusion: Embracing AI with Confidence

AI is reshaping the future of business, and HubSpot users have a wealth of opportunities to leverage its capabilities. By understanding AI’s practical applications, maintaining data integrity, and fostering a culture of responsible adoption, businesses can stay ahead of the curve.

As AI continues to evolve, embracing its potential while ensuring ethical and strategic implementation will be key to long-term success. Are you ready to integrate AI into your HubSpot strategy?

 

 

Transcript:

Emily Yates (00:02.463)
So yeah, this will be an exciting topic that I think everyone is very, very keen to hear more about. But yeah, we've already had bits of chats already, haven't we, about how it's really changing all the time. So.

Yeah.

Annka (00:23.34)
Definitely. It's been a big few years for AI for sure.

Emily Yates (00:28.821)
I think we're just going to wait so I can get the go ahead from the team. We're definitely online and then we will get going.

Emily Yates (00:44.693)
trying to do a quick check as well.

I'm always cautious of hearing myself back through the headphones as I go to check the events page.

Annka (00:53.422)
something.

Echoes are the worst.

Emily Yates (01:01.519)
It looks like we are live.

Emily Yates (01:08.797)
not had confirmation from the team that they're seeing it yet. Looks like it could be about to land. And it is bang on 12 o'clock as well. So we might as well get going because we do like to keep these to 30 minutes and include opportunity for questions as well. So welcome to our third LinkedIn Live of the year. These are happening on the third Thursday of every month. And today's topic is

AI and HubSpot AI and trying to really simplify that, remove some of the jargon and just strip it back because it's become so populated and such a heavy topic in some cases that we want to really simplify it. So I am joined today by Anka who is Director of AI Strategy and Transformation if we're going to go with the very full title.

Yeah, great to have you with me. So I guess the first task I'm gonna give you, we obviously said that this was trying to explain it like you're five. That's not meant in any patronising way. It's just to really make sure we keep the simplified way of doing it. Can you explain AI to me like I am five?

Annka (02:07.692)
full fancy title.

Thank you for having me.

Annka (02:32.598)
I think there's, there's a lot of overcomplication of this topic and, and fancy jargon that's just being thrown around, but in super simple terms, essentially, in terms of LLMs or these large language models like Chat GPT, they're more or less machines that are trained to pretend like they're human. Essentially it's automation with a, with a brain and it's trained on huge volumes of data. But the average of all of that is kind of average for answers as well.

But these kind of models, they are covering a lot of the day-to-day needs. But yeah, it's essentially like a machine that's trained to pretend like it's human.

Emily Yates (03:15.221)
and guess we've discovered that it's pretty good at pretending to be human as well as time has gone on.

Annka (03:20.78)
Yeah, absolutely. Especially when it's kind of settings or scenarios where it has training data that kind of covers that scenario. I mean, there's a lot of scenarios where it doesn't have training data and it doesn't understand what's going on necessarily. And one important thing to know is that it doesn't know all of the answers and it's trying to be really helpful. So if it hasn't come across something similar before, it very often makes up

things. It's trying to be super helpful and it has a tendency to just pad lists or just add answers that things will be useful for you.

Emily Yates (03:57.641)
I've often found myself in my prompts asking it not to make stuff up. I think that's one of my biggest tips for people, like, telling that you don't want it to make stuff up.

Annka (04:01.678)
Yeah.

Annka (04:08.296)
That's definitely some of the guardrails. That is a useful tip for adding to any prompt. And please don't give me any answer if you don't know the answer. Seems like a given, but yeah, they're very helpful, though.

Emily Yates (04:16.447)
Yeah.

Emily Yates (04:21.301)
So, what would you say are some of the most useful parts of AI? What is out there that should probably not be used and what should we be using?

Annka (04:36.568)
So think there's so many different types of AI functionality. You have a lot of AI, that's more the conversational AI, but not only differences in what kind of tools are available, there's also a lot of differences in terms of what kind of tasks you should be and should not be looking to use AI for. So think maybe the answer that will help you find value or the...

question that maybe will help you find more value is asking what tasks should be like prime for, or should be good for using AI for. But there's definitely a lot of hype out there just to cover some of the kind of heights and not like what is worth it and not worth it. I think there's a lot of, or a propensity or a want to just start implementing a bunch of like point solutions like

You see some fancy demo and you just want to try to do what's in the demo and you sort of, you get a little bit sold on fancy demos and you come back home and then you try to implement it into your, sometimes these are hard to implement and it ends up making more chaos than structure. I maybe trying to find AI functionality that could cover.

a lot of use cases in your existing tech stack, like HubSpot AI, there's so many good functionality there. A lot of undiscovered functionality that could cover a lot of those kind of pain points that there's a tendency to look outside of HubSpot for. And I'm thinking a lot of the way we're using a lot of these large language models definitely could be covered in some of the functionality that we find in HubSpot.

Emily Yates (06:24.071)
Yeah, are there any places that you would definitely start with when it comes to a company getting started with AI in general? And then we'll touch on the HubSpot side of stuff just in a little bit.

Annka (06:36.266)
In general, yeah.

So one of the things, especially for companies that are placed in the EU, it's important to look at your GDPR and privacy policies because sometimes they can shape your opportunities for using AI. Trying to figure out how you interpret GDPR. If you have any data processing region restrictions, that's a big one in AI. Most of the AI processing is happening in the US. So if you have a strict

protocol for not wanting to process any data in the US, then you should be pretty clear on that because it maybe gives you different opportunities than maybe just reaching for some of the solutions that would be maybe more obvious. But yeah, when you've got a good sense of your privacy policy and your data processing policies, looking at your data quality because

We see all of these fancy demos, but none of them really work in practice if you come back to a data set that is really dirty or you have lot of duplicates or you have a lot of missing data. So yeah, think getting your data quality right, getting a good sense of where you are with your data because AI readiness, it comes down to data quality as the first step for sure.

And then I would look at content. Suddenly you have a lot of content out there that's super outdated, but now chatbots and AI tools might not be able to evaluate necessarily what is the useful content. So think a lot of people have been kind of looking up their own company in in ChatGPT and finding a lot of the answers are either off old or straight up just made up.

Annka (08:27.502)
And in the same sense, if you're going to be implementing AI tools like chat bots or other AI features in HubSpot or other software, then making sure that your content is kind of maintained well. That's a big one also, for sure.

Emily Yates (08:49.833)
Yeah, you mentioned just a little bit before about not looking outside of the existing tech stack that you already have. And Valerie on the LinkedIn Live has posted about the content remix being a really good, easy starting point with HubSpot AI. She's actually put that it's not perfect yet, but it's really good to turn blogs into social media posts and emails. And recently there was another update to that as well where you can have like...

bring it from multiple data points as well. So, guess again, always expanding. Yeah.

Annka (09:27.576)
Yeah, I think that's a I think with a lot of AI tools or functions, the word yet is super important because even if we are testing something today, don't discard it. If there is something that seems promising, keep testing it. Give it a couple of weeks. Give it a couple of months. Retest it. And I think that has been the case with our experience with these remixed tools as well that at first they might not truly make a big impact in the terms that we work or in the ways that we're working. But

retesting them and we see that we're getting more more value out of them as they develop. And that's something that goes for most AI functionality.

Emily Yates (10:07.335)
Sure. So just thinking about different parts of the business and specific tasks that may be suitable for AI and implementing AI in that part of the business. Have you got any sort of guardrails for that that people can consider?

Annka (10:27.266)
Yeah, so when I'm considering whether or not a task is suitable for AI, I like to categorize the tasks. I usually think about them in terms of three segments where a huge chunk of our day-to-day tasks falls within sort of an objective category. These are tasks where outputs can be easily checked if they're right or wrong. There usually is a right or wrong answer, like processes with an expected outcome.

where there's not a lot of gray areas, not really subjective, just like objectively is this right or wrong? These could be a really good place to start. And it requires a little bit of a mapping and an understanding on how to evaluate whether or not a task is objective or more subjective. But there tends to be a high return on investment in the segment of more objective tasks.

Yeah, so I think I would have started there maybe. One another segment of tasks that are maybe easier to implement solutions for are the more subjective tasks where there are bigger room for what is acceptable and things that are not like high stakes necessarily like emails, a lot of internal communication, more generative tasks.

might be easier to implement and can save a lot of time and give a lot of opportunities for personalization or other things that you wouldn't necessarily consider if you'd have to do them more manually. not only would you save time on the things that you're actually doing, but it opens up more opportunities for doing things that you wouldn't consider otherwise.

There is also a third category that I would say it's more in the expert category. This is where you really need the domain expertise and a lot of experience to be able to get this right. And I think I would have left that one as like the third thing to start picking up with AI. And an advice for all of these is making sure that when you were implementing things, keeping humans in the loop.

Annka (12:32.974)
automate everything to the point where it's automatically set up and sent. It ends up being a little bit of a black box and it's hard to evaluate whether things have gone wrong. Yeah, that's maybe.

Emily Yates (12:45.333)
Yeah, you see lots of discussion, don't you, on social media about how people can tell that content is AI produced and that people talk about the dash and things like that and the overuse of emojis and things like that. That's where the human element comes into it and we're obviously big advocates for making sure the human side of it stays. So yeah.

Annka (13:13.326)
Yeah, it's getting really easy to see and spot where it's obvious. And I think that's with a lot of the spam that we're getting and the spam that we are ads that we have been getting more and more used to in the past 10, 15 years, we're getting like that filter in our heads where we disregard when we see something feels very advertising. And I think that filter now extends more to the AI generated text as well. So keeping things human, keeping things authentic, super valuable.

Emily Yates (13:43.125)
I think we spoke last week and we touched on the point of when you're reviewing AI content, it's quite a subjective point of view. Like I could read your work and think that needs to change or you could read my work and think that needs to change. I guess it's about applying that same lens to AI content as well in some regards.

Annka (14:04.733)
Yes. Especially in that subjective category of tasks where we could be utilizing more generative tools, creating text and images and things. mean, yeah, as you said, if I was to read somebody else's work, my preferences would probably make me think this, this, could have changed. My sort of outlook on some of the

Outputs from AI changed a little bit when I started thinking about it as a colleague more than what I have produced because if I criticize it in terms of what this is something I have produced, I would maybe not look at it as subjectively, but looking at it as if it's a colleague that's produced it. You understand that there's a bigger like gray area of what is like subjectively high quality and this it's a way of thinking about it that maybe makes you understand that there's.

more of a gray area around the quality output.

Emily Yates (15:03.285)
Yeah, so I guess one thing that HubSpot is really good, the AI in HubSpot at the minute, there's a lot of generative AI opportunities in there. mean, everywhere you look in the platform, if you look for that little diamond shape, there's an opportunity to use Breeze. I guess you've got the copilot with the list generation and you've got the, you never have to write

Annka (15:29.25)
I find the summaries, the summaries, yeah.

Emily Yates (15:33.171)
you never have to write a description or a subject line or anything like that ever again if you didn't want to. And then you were just mentioning the summaries.

Annka (15:42.638)
Yeah, there are so many. These tools are popping up almost day to day. feels like there's so much development happening and only this morning, as you mentioned, there's new features popping out in new beta So it's definitely new features all the time. But when you're sort of just day to day working in HubSpot, when you see those little stars, there's always opportunities there for

for utilizing AI in there. And some of the features that are easy on the GDPR policies are definitely the generative ones, like the content assistant. So even if you have a fairly strict GDPR policy, in most cases, the content assistant and AI in terms of generative doesn't touch the CRM, even though we've turned things off, doesn't access your CRM.

you would still be able to use a lot of those tools. And that is essentially what I mean by learning your policies, GDPR policies to understand what you can and cannot, because I think a lot of companies are kind of fearful. They don't want to breach any GDPR or internal policies, so they stray away from it. But I do think that there's a lot of opportunities for companies that really truly understands what are the leeway within our policies.

Emily Yates (17:04.991)
Yeah.

Annka (17:06.764)
That's

Emily Yates (17:08.565)
for sure. We have actually got a question in the comments and what are some of biggest misconceptions about AI in HubSpot? What do people often assume it can do but it actually can't?

Annka (17:23.542)
I think maybe easily the first one is create something without having any humans in the loop. I would not recommend that. think a lot of, if you're looking at the most efficiency and you're seeing this as a tool that can replace people or replace a lot of processes that takes a bit of time, I think that is where I would pause and

reconsider, I think there is still need for the human in the loop for a lot of these generative tools. So that's maybe a misconception. Another misconception that we talked about is that we have to stay away from all tools in there because we have strict GDPR policies. But while there are so many functions that you could still be getting a lot of use out of and find a lot of value in that doesn't touch your data. And that kind of

to different uses either. Of course, there are so many opportunities to do good things with while it's accessing your data, but there are so many valuable things like the content remix, if there's no personal data in there, like the generative as well. I mean, just figuring out what tools you can use and start using them has a lot of value.

Emily Yates (18:46.227)
Yeah, I guess one of the things for me, if I have to answer that question as well, is that when you do have it linked to your CRM and you are using your CRM data, it's actually only as good as your CRM data is. And you mentioned very early on about the...

Annka (19:04.76)
data quality.

Emily Yates (19:05.247)
data quality and I guess that's a big point in this.

Annka (19:10.028)
Yeah. mean, if your result won't be better than the quality of your data for sure. that's usually when we talk to customers, when we go in and consult, we see that is usually our first kind of blocker. The data quality just isn't good enough to give the expected outcomes using some of these tools. So yeah, that's the AI isn't going to be better than what

the AI has to play with or to learn from and to process. That goes for content as well for chatbots, for example, it will truly go off the rails if you don't have the quality data.

Emily Yates (19:43.978)
for sure.

Emily Yates (19:55.177)
Yeah, for sure. So in terms of touching on the future of AI and where we believe it's going and what sort of things people should be looking out for, I know we've mentioned literally 20 minutes before this live happened, Kyle Jepsen had an Orange Hat moment on LinkedIn announcing new prospecting agents, which shows that HubSpot AI is going to keep continuing to grow. But what about AI in general?

where we see that going.

Annka (20:26.434)
Yeah, so I think there's this thought that AI has to keep the pace of development for us to be getting new tools. But I want to just say that if the world of AI just paused or the development just paused right now with the existing capabilities that we have, with the tech that are the existing models that we have, we still would have spent

10 years implementing all of the useful functionality that we could get out of that. So even if it paused right now, we would still have 10 years of tremendous tech growth, I would say, in front of us and a lot of opportunities. But it's going so fast. And now I see the predictions for general intelligence is getting...

Like they're moving the bar closer and closer by the minute and now they've come up with a new term, super intelligence that is going to be even more revolutionary. there's just a, there's a lot of a very techy talk coming up here. I think this is where I would go down the rabbit hole for real. But no, think it's, it's gonna, in terms of the market, I think that's maybe where we need some time to really adjust.

and need some time to start using it properly. People need to get used to using it. And there is a lot of opportunities that will come once people are getting more used to it, like live translations in meetings or things like that that we're just not ready for. And humans have been through kind of these big shifts before, like the digital change and it's...

getting used to the internet and the fast-paced things. So we can adjust. And I think once people get more used to things, once it's normalized in society, I think there is even more opportunities. So I think it's gonna be up to individual companies and organizations to decide whether or not they wanna be early adopters and push for things, or if they wanna sort of wait and let the market get used to some of these features before pushing them as well.

Emily Yates (22:48.745)
Yeah, guess there's always going to be people at different stages of the AI adoption journey. I have a question that's a bit off of what we were talking about, but if someone had a colleague that wasn't up for adopting AI and the rest of the team were, what sort of tips do you have for encouraging them to start on that journey and why should they?

And what would you say to put the mind at ease?

Annka (23:21.07)
Yeah, I think that's a lot of people have a colleague or feel this way themselves where it feels like things are moving too fast and they can't really see any practical solutions or any benefit to having the machines take over some of our tasks. I think there's a fear of being replaced, a fear of having our tasks devaluated or having our tasks just taken over essentially. I think one thing is

finding good AI solutions that has an impact on their day-to-day that they find value in. Like giving them tools where they see that there is an actual positive impact for their day-to-day. And maybe just making sure that they have the right training as well. Because a lot of people are afraid of using it because it feels like this weird black box. You don't really have control over it.

Training really helps and also setting safeguards in place so people don't feel like they're doing something wrong, having good understanding of what to do and not to do, like in terms of privacy and so that they feel more secure in using AI. But I mean, there is a time of maturation to get to the point where everybody feels comfortable in it. But I would say just training and the...

AI isn't going to come in and take all of our tasks. Our tasks are not done evolving. Some tasks might change, some roles might change, but our tasks, our roles has been evolving since the dawn of humanity. And even if our tasks has to change, there's not going to be any stop to... Our tasks aren't done. They're not done developing or evolving.

they will evolve into different things as we're getting AI tools to help us with some of the more, yeah, the older tasks. So finding new tasks.

Emily Yates (25:25.681)
Amazing. Yeah. I've done that rogue thing where you start a LinkedIn Live and now my battery is telling me I've got 9 % but it's been plugged in all morning. So this is a bit touch and go. Bit nervy. So yeah, I guess we've got one other question. So is there a way to use HubSpot AI to elaborate on the processes behind...

Annka (25:37.87)
All right.

Count down.

Emily Yates (25:52.379)
e.g. a sales process on how a lead goes from MQL to customer within HubSpot.

Annka (26:01.437)
Right, so there are, I'm just thinking on the fly here, but there are probably opportunities for setting something up in terms of workflows that could document something. I think this question needs to be paused on to a CRM expert, I think. So maybe we can reach out after the live and have some answers out there. I don't know how this works.

Emily Yates (26:22.099)
Yeah, I mean, I mean we can always reply in the comments afterwards. That's absolutely fine. Are there any other comments coming in? Or is there anything you wanted to share, Anka, that I've not asked you about?

Annka (26:36.65)
yeah, could. I think one of the one of the things I try to cover when I talk about AI is what to use AI for and not to use AI for. And one of the one of the key advice that I always give is that in terms of expert competency or domain expertise in the beginning you got chat GPT or you got all kinds of AI and then

there was a little bit of a tendency to think that, okay, now I can be a lawyer or now I can be a, like you got all of these free expertise that you felt like you had at your fingertips. And then an advice is that if you don't have the competency to actually write it yourself, you won't have the competency to qualify or check the quality of the output of AI. So that's maybe a advice that.

Please don't pretend like you're an accountant or pretend like you're a lawyer. If you don't have the competency to complete the work, you don't have the competency to quality check the outputs.

Emily Yates (27:41.875)
Amazing. So just in response to the question that was asked just before about HubSpot AI, we've actually had a response in the chat. So yes, HubSpot AI can help explain and automate the sales process from MQL to customer by tracking leads, assigning tasks, and sending follow-ups. It can also create step-by-step guides and reports to train new team members.

plus AI chat box can answer questions and provide insights in real time. So the answer is there. And if you do need any more, we can still get someone in the CRM team to come back. That's absolutely fine. Are there any more questions in the chat or any questions that people want to know about?

Annka (28:16.449)
Amazing.

Annka (28:36.692)
How's your battery doing, Emily?

Emily Yates (28:38.405)
and what what you don't all realizes I've got someone in the room right now trying to discreetly get get me plugged in so it doesn't die I'm on seven percent and we will look at the comments and questions in the chat afterwards anyway so and if any come in after this or you think of any after that we're always happy to go in and answer there and and you can always res reply to us directly and

Yeah, it'd be great to have more questions. I've just got notification that there actually are no other questions in the comments, which is good. And we are just approaching the 30 minute mark, so I guess we've hit the time. We've shared some top tips. My battery is surviving. And that is the end of our LinkedIn Live for March. But we will be back in April. And at that point we'll...

It might have to change from the third Thursday. It depends on when HubSpot release their next spotlight because that is going to be the focus and the updates in HubSpot. It's what we're going to chat about at our next LinkedIn Live. yeah, if no one's got any other questions, I'm going to stop going live and thank you for joining us everyone and there was some great interaction. Bye bye.