The Uncomfortable Truth About AI in Sales

Recently, Collin Cadmus (Founder of Collin Cadmus Consulting) and Yuriy Zaremba (CEO of AiSDR) teamed up for a podcast where they discussed peoples’ most burning questions about AI in sales, including:

• The role of AI in sales strategies and operations
• Hidden costs of AI
• Challenges and risks of AI in sales
• Sales AI functionality and limitations
• GDPR compliance

Here’s a recap of what they said.

How is sales data collected? Is it scraped in real time?

(Yuriy)

There are two types of data. 

First is the contact data, and for contact data, AI SDRs usually rely on third-party providers. At AiSDR, we use Apollo, ZoomInfo, RocketReach, and even our own database. 

Then there’s the research data. For research data, we analyze LinkedIn, the internet, hiring intent, and news so that outreach emails are more relevant to the person being contacted, which increases our chances of conversion.

Is the future of sales an AI-native CRM?

(Collin)

During one of my recent podcasts, Jeron Paul from Spiff said that no one should be starting an SaaS company anymore. He said that everyone should be starting an AI company because you’re going to be building software that has an AI component to it. And if you’re just building SaaS, you’re probably going to get leapfrogged by someone that’s doing the same thing with a bit of AI mixed in.

But does that mean we’re going to get rid of Salesforce and HubSpot?

Probably not. If anything, they’re going to become AI-driven in many ways, and I think we’re starting to see that.

You see startups coming out with this sort of single-purpose product like using AI to analyze your calls and update your CRM. Then in two or three years, they probably get acquired by HubSpot and Salesforce, and they’re built in natively. 

So is our CRM going to be largely powered by AI? I think that it’s inevitable that everything is going to be largely powered by AI, but that doesn’t mean humans won’t be tinkering around and doing stuff.

There will be areas where there’s work that we used to do that we no longer need to do, but that just means we can focus on other things. If we look at an AI SDR that’s drafting and sending out emails, it just gives us more time to actually talk to people, get on demos, or get on discovery calls. So all of a sudden, you can do more with less, and sales teams will be smaller and more efficient.

(Yuriy)

We already see that HubSpot and Salesforce are making a huge AI push. HubSpot announced their Breeze AI which is about data analysis at inbound, and Salesforce has agents. I would love to see another CRM market transformed with AI. It’s yet to happen, but I’m very curious about that.

As AI takes on more responsibilities, how can sales teams ensure they don’t become over-reliant on automation and lose critical thinking skills?

(Yuriy)

This is something I’m thinking about every day as we automate a lot of stuff inside AiSDR. I always keep coming back to this, and honestly, I don’t have a good answer yet other than people with great critical thinking should be operating AI to their benefit.

I think there are two categories of workers. “Creators” – people who analyze and build the process. And “Executors” – people who execute the process. I think the biggest risk lies in people who just execute the process because that part will be automated more and more. And then there’s truly a risk of being too reliant.

But then they can approach it with a curiosity-driven mindset where they say “Hey, let’s learn how to use this thing” and then be mindful they shouldn’t rely too much on it because it’s a tool to assist and help us grow faster than a crutch that will make us less sharp.

(Collin)

This reminds me of conversations about the future of education. People will say, “Now that kids are typing their essays, they’re not learning how to write in cursive” or “We’ve got spreadsheets and calculators so no one’s exercising their mental math.”

I think there’s a couple way to look at this. 

On the one hand, we can look at it where we start to compare ourselves to the next generation and we’re going to think “They’re not going to be able to do something that we can do” and everyone’s been saying that since the dawn of time. 

If you go back long enough in history, you’ll find people that every single day they spent their entire day walking to where they could get fresh water and carry enormous buckets of water to their farm or community. And then they’d go back and forth, back and forth, and back and forth. Then one day, someone invented the wheel, and all of a sudden, one person could pull multiple buckets. 

Now people don’t know how to go fetch fresh water. But it’s also not a problem because we just don’t have to do it.

So we have to think about things like that – Are people going to become dependent on tools and not know how to do certain things they used to know how to do? Possibly, but the bigger question is “Will they ever need to know how to do these things without the tool?” The whole purpose of using tools is so we can do more.

AI can streamline sales, but what about hidden costs like training the team to use these tools effectively? If the learning curve is too steep, are we really saving time?

(Collin)

I think it’s too soon for me to have a clear answer yet because we don’t really know how it all looks. But this is a concern that should be had when adding any tool to the team. You have to do a sort of cost-benefit analysis and understand how critical it really is.

I’ve made the mistake in the past of signing up for everything that sounded like it could add value because I completely overlooked the implementation, the training, getting people to adopt it, and the simple annoyance of having to bounce around using so many tools all day. Over time, it adds an inefficiency, which is why I’m such a big fan of HubSpot since it brings everything under one umbrella. 

And I do think that there is a problem of having too many tools. You just need to really make sure that any new tool you add makes sense and that you’re committed to it. And if it’s not going to become a big part of your process, it probably doesn’t need to become a part of your process at all.

(Yuriy)

I agree. If we focus a bit more specifically on tools that use AI, I think there are many flavors of AI tools. There are different co-pilot tools that sit in your mailbox or on Zoom calls.

The learning curve can range from installing the app and letting it run in the background to training sessions on how to use it, do the research properly, and understand my voice.

I think when you’re talking to any vendor, even AI vendors, you need to ask “What is your average time to value?” Then ask them in detail what the onboarding process looks like. 

In the AI SDR space, many of them take time to work. You have to set up the email infrastructure, you have to train the model for specific messaging and frameworks, and you have to budget for it. At AiSDR, we try to take a different approach so that the customer sees results in the first months. 

Is AI changing the way businesses buy and sell software? Will sales outreach as we know it go extinct?

(Collin)

On the one hand, there’s sales and outbound, like how do we drum up demand. And if you remove the human from it, it’s just sophisticated marketing.

But then we have to think about the buyer side of things. It’s not just about how tools are going to change the way we sell. Tools will also change the way we buy. We’re transitioning from a hardcore sales model to a ‘buyer enablement’ model where sales teams try to help buyers make a decision.

We’re making tools that make it easier for us to do the job we were doing yesterday, but there’s going to be companies that are looking to make ti easier for buyers to find the right product. Today’s version of that is G2 and other product review sites, but what doesn’t exist yet is a completely self-guided sales process.

There will be AI SDRs, and then there will be AI like demo rooms and deal rooms. There might even be an AI buyer tool where if I’m looking for a CRM, I could go to this platform. They’re not affiliated with Salesforce or HubSpot, and I can tell them what I’m looking for and it will give me my options, break down the pros and cons, and make specific recommendations based on my criteria, sort of like what a sales person does today. 

At the end of the day, there are still going to be instances where sales people are needed, and maybe they’re go from running 4 demos a day to 10, or maybe even none because by the time people talk to you, they’ve already gone through most of the sales process.

(Yuriy)

We are still in the very early stages of AI automation and outreach in sales. It’s going to take time and the buying process will start transforming once all the outreach is AI and people need something to start analyzing it.

People want to buy on their own terms without being sold to. And as AI advances, AI will enable larger deals to be handled this way. Imagine being able to jump on a demo with AI knowing that it’s prohibited from selling to you. If you can simply get your questions answered and have a consultation, maybe you’d be more interested. Essentially, you get a very personalized buying experience without actually be sold to and talking to a real person.

If AI is already being used in sales and marketing, is it safe to say it’s being programmed to be manipulative?

(Collin)

Yes, and I’ll tell you why. Because humans are biased, and it’s getting all its info from us. You can look up a healthcare question and it’s going to aggregate with a bunch of other info from doctors. So the bias is innately in there because it comes from us. Is it programmed to be biased? That’s another question, and I don’t think we’re going to get answers unless they open-source the code.

(Yuriy)

Let’s be honest. The number of vendors that have their own proprietary large-language models in the sales and marketing space is close to zero. Everyone is using foundational models with some infrastructure and fine-tuning on top of it. Then the question is “Is the foundational model manipulated?” No. 

(Collin)

I agree with that. It’s more the data it gets from humans. That’s where the bias seems to creep its way in. 

(Yuriy)

The goals that humans are instructing it to achieve. So are humans manipulative? Absolutely. And AI is just a piece of software that we use again and again for our purposes.

Is AI for outreach only?

(Yuriy)

If we’re talking about AI SDRs, it’s not about what it’s going to do – find leads, create emails, and so on. It’s about fully automating the process end to end. When we think about outreach in the past, what essentially happened is you have a data vendor, a sequencing vendor, people to execute another part of the process that could not be automated, and people to write emails. Then when somebody responded, you jumped in and closed the deal.

It still works like that. And the promise behind AiSDR is that now with AI, you can automate that workflow and do a really good job. You can pull data from a data provider, then the AI will analyze that data, research the prospect, and create a unique email for every single prospect based on the research and the details you gave it about your business. 

It then takes care of the email infrastructure and gets the email into the inbox, which is not obvious nowadays. Then it follows up and follows up again, and when it gets a response, it corresponds, answers questions, and handles objections until the meeting is booked. In the future, it will use multiple channels and it will have more signals to act on.

(Collin)

I think you nailed it.

How can sales AI be GDPR compliant?

(Yuriy)

All databases are somewhat GDPR compliant, but not entirely. First, data has to be acquired in a legal way. But most importantly, people have to have the ability to remove their data from the system if they don’t agree or don’t want it to be there. You then have to remove that data and report to the person that their data was removed.

From the perspective of sending emails, you have to make sure the person receiving the email can opt out. So if you look at AiSDR or other AI SDRs, you should be asking the question “Where are you getting your data?”

Usually, it’s a major data provider, like Apollo, ZoomInfo, and RocketReach in our case. We also have our own kind of database.

From there, you should check “Is the provider GDPR compliant? Do they have indemnities in their terms of service?” But this is just common business practice. Companies that claim to be GDPR compliant need to have a track record of how their data was acquired.

At this point, you need to see if there’s a process in the company you’re working with and the data provider you’re working with to remove data from the system. This is a requirement.

The last piece is whether the AI SDR solution you’re using allows unsubscribes to be included automatically in the links. The link itself isn’t always enough. You sometimes have to have it in the header of the email.

How can you avoid becoming a target for a phishing email or other cyber threat when receiving AI sales emails?

(Yuriy)

My general piece of advice is to not click links in emails from unknown recipients. Check who you’re receiving the email from. Check the domain. A common sign of phishing is a domain that’s not aligned with the domain of the company reaching out to you.

Now if you’re using throwaway domains like “tryaisdr.com”, it’s a bit more complicated. I’d say though that if see the domain doesn’t match but you like the email, try copy-pasting the domain into the main search bar and see where it takes you.

Will AI firms pivot or persist? What’s your take on the state of AI sales tool adoption?

(Collin)

I think it’s both. There are some companies that are going to fall off a cliff, mess up, and burn all their money or piss of their customers.

Then there are those who are going to figure it out and crush it.

But there will also be situations when maybe a few years in or many years in that at a certain point, it’s easier for someone to just build from scratch something better than you and innovate or iterate their product. As a first mover, people are watching everything you’re doing and figuring out your mistakes, and they have the advantage of having learned from you.

Can you discuss word choice and the importance of vocabulary as it relates to perception, sentiment, and the feeling of being duped when engaging with AI agents?

(Collin)

I think you need to call it what it is. You shouldn’t use AI and try to call it a human. If it’s something like a chat agent, I think it’s very obvious when those aren’t human, and that’s probably why they don’t work so well. But as they become more human-like, I don’t think it matters.

The negative undertone to AI is probably a temporary situation. It’s probably something that’ll fade over time. Currently, we’re dealing with just the first iteration, which is never great. They’ll only get better over time. We’re also going through the whole phase of people are scared of losing their jobs, so there’s a sort of threat element to it that paints a darker picture.

But I think as we become more friendly with AI and more reliant on it, it will become our best friend and I think we’re not far away from preferring to speak to AI. 

(Yuriy)

I think there will be normalization and basically AI will be a fact of life, like a personal computer or smartphone. 

(Collin)

It’s just a new tool to make us better and smarter.

(Yuriy)

When we’re talking about today, for most AiSDR customers, the AI pretends to be human. It doesn’t disclose that it’s human. Emails from a good AI SDR are indistinguishable from emails from humans. 

Does the near future consist of companies hiring people to do full-stack sales who can run everything by using tools like AiSDR?

(Yuriy)

Yes would be my first guess. I know that people who are doing full-stack sales are generally not happy about it because they don’t have enough time to prospect from their closing deals and vice versa. When you stop prospecting the deal flow drops and AI SDRs really fit well into that because they prospect and let you close deals.

(Collin)

I think we need to see how everything evolves and how successful it becomes. I think there’s potential where a full-cycle AE uses this as their companion on the side, but I also see a world where you have someone else doing it the same way you would with an SDR team. But instead of 10 or 20 SDRs, it’s like 1 or 2.