How to maximize efficiency with AI-driven sales enablement solutions?

Oleg

Hi St John! Thanks for joining Devico Breakfast Bar. I'm sure that we'll have a nice episode today. Could you please start by telling us a bit about yourself and your professional background?

St John

Absolutely. Hi, Oleg! Good to chat with you. So, I'm St John. I'm the Founderand CEO of Perlon AI. Perlon AI was originally started as a strategy consultancy and advisory business by myself and two partners. In 2023, we've subsequently moved to being more of a product business that's focused on AI sales enablement tools. And we've launched our first commercially available product just in December, which is called GenBD. Prior to that, I worked at global AI and SaaS businesses. For the past six years, I was mainly Head of Growth, Head of Strategy, Head of Sales in roles at companies from seed stage through to series D. Before that, I worked in Alternative Investments. And I've always been interested in building and scaling companies. Given that I’ve already had the experience in AI, I don't think there's ever been a better time to go out and start a company. So, that's what I did last year.

Oleg

Sounds great. Could you provide an overview of your current project with Perlon AI and the one you are actively involved in and elaborate on the specific challenges or issues it addresses?

St John

Yeah, sure. So, we started out by doing a range of different projects for enterprise AI deployment use cases and have subsequently, like I mentioned, moved into a very specific product focus because we spoke to a lot of businesses who were struggling with the same thing, and that's generating new business from outbound prospecting. There were kind of three reasons for that and three components to it. And the first was that templated emails just don't work anymore. Everyone is very sick of receiving the same trash in their inbox, and they don't get good reply rates. The second thing is that personalization of outbound prospecting really does work, but it's not scalable. It takes a long time to sit and compose an email, do the research, and write something that's actually effective. And then, thirdly, the new rules that are coming into play from Google and Outlook in February make things a lot harder. When it comes to actually landing in inboxes, if you're sending templates, you start getting your domain ruined and start going to spam at a far higher rate of regularity than has been historically.

So, we kind of built those three things into a product. The product GenBD uses fine-tuned AI models that incorporate data, tone of voice, sales style, statistics with up-to-date web scrapers on the target companies and the target prospects that each individual business we work with is going after so that they can send hyper-personalized cold emails that are relevant and that can be conducted at massive scale. So, it's kind of all relating to the fact that the world of prospecting is really changing and has changed dramatically. And we're building the tools that enable those sales teams to do more with less and make the most out of some of the advances in LLMs and AI in general to be able to really optimize the selling process and the prospecting process more specifically.

Oleg

Okay, great. And what are these rules that are coming in February from Google? What are those rules exactly? I mean, especially in terms of templating, because cold emails, it's all about templating.

St John

Yeah, exactly. So, what happens is if you start to get more than 0.3 percent of the emails you send marked as spam by the recipients, you risk ruining your entire domain and having your domain blocked. They've already started to really clamp down on templated emails and what they perceive to be spam by a flagging mechanism, essentially, that is able to determine based on the volume of which you send emails and also the content, whether those emails have been written by a human or not. So, one of the first things that any company can do is just be conscious of how many emails are being sent out per person per day. You ideally don't want to be doing more than 150 outbound emails, regardless of what you're sending, because you'll start to, again, trigger these flags that will put you into a bucket of someone who is a spammer. So, this is why for a lot of sales teams, open rates of emails, open rates of outbound emails in particular, have really dropped to, you know, often sub 20, sub 10% in some of the use cases that we've seen, because those domains have just been harmed to such an extent by the practices that they've been operating with for a while.

Oleg

That's interesting because we're also doing email outreach in our company, and our open rates are 60 and 70 right now.

St John

Nice. That's good. Yeah, that's a really good open rate. And again, depending on your domain health, you've clearly got quite good domain health, whether that will be the case in February, and again, I don't know exactly how much templating you guys are using and at what volumes, but ideally what you can do with something like GenBD, the tool that we've built, is you can get generally around the sort of 80 to 90% open rates from the emails that we're sending. Because the content of every single one of them is genitive, and it's different in each instance, and it's very specific to that user or that company that we're going after. So, they're able to beat the spam filters in a way that some of the templates won't be able to in the next couple of weeks.

Oleg

Thanks for the advice. What inspires you in the field of AI and technology? And what innovations do you see on the horizon?

St John

So, I suppose there's a lot to inspire us when it comes to AI. I think the pace of change in the industry and the fact that we're really only just getting started is in itself just pretty incredible. I'm personally excited by the development of some of the multimodal technology. I really like the intersection of robotics and AI, and I'm looking forward to seeing ways in which that can really sort of take an effect in people's homes. I mean, for example, when I'm useless at cooking, I really want a robo Gordon Ramsay in my kitchen or something like that. And really sort of seeing how that develops is going to be really interesting. And again, it ties into this multimodal nature. Also, it ties into, obviously, what OpenAI are trying to achieve, which is AGI, Artificial General Intelligence. It will be really interesting to see how that develops and quite what form AGI actually takes. It's kind of difficult, I think, for anyone really to define it at this moment. But it seems to be that that's the one thing that is getting a huge amount of investment and attention from what is certainly the most innovative company in this space. So, I think the progression of that over the next 3-4 years is super, super exciting, and can't wait to see exactly how that manifests and what those benefits are to everyday people.

Oleg

In addition to your professional life, do you have any personal interests or hobbies that you're passionate about, and how do they complement your work or provide balance in your life?

St John

Yeah, my main real interests are sport and travel. I play a lot of tennis. I exercise a lot. I think that helps with two things, mainly. Firstly, it really gives you a place to switch off from the phone for an hour or two, which I think is otherwise almost impossible these days. And then secondly, it really gets the endorphins going and just keeps you sort of healthy, which I think obviously ties to your mental health in many ways and your ability to work well. I mean, one of the things that I've really been trying to get back into this year is reading as well. I found it's not going that well. I'm trying, but what happens when you're starting a business, as you know, is that you'll find yourself on your laptop until lights out most nights anyway, which isn't great. I mean, I think the other thing that's very important to me is travel. I love going abroad. I love going abroad for work, I love going abroad for pleasure. So, being able to do as much of that as possible, particularly if I can do it and actually do it for a work reason too, is also something that's really interesting to me, and I've loved doing. I've been very fortunate in the places I've been able to go over the last few years with my jobs. So, that's definitely helped with keeping me safe.

Oleg

How do you stay on top of industry trends and events? Are there any particular valuable resources or sources of information that you find useful?

St John

Yeah. So, I'm very big on podcasts. I commute into the office every day from my house in West London to our office in East London. So, I usually have a good 30 minutes there and 30 minutes back where I can really get into a podcast. Typically, I listen to business podcasts. I listen to the AI breakdown on my way to work every day. That's a short sort of 20-minute update, daily update on exactly what's happening in the world of AI, any latest developments. Then, I listen to things like Pivot, I listen to Saster's pod, IdeaCast from Harper Business Review, All-In Pod. A range of different generally sort of business-focused podcasts is where I tend to get most of my information. As I mentioned, reading has been something I've really wanted to get back into, but it can be tricky, and it's often the thing that goes out the window first when you've got a million other things to do.

Oleg

Given that we can no longer imagine our life without AI, obviously, let's focus and talk a little bit about it. In your opinion, what challenges do companies commonly face when integrating new AI technologies, and how can they overcome these obstacles to stay competitive?

St John

Yeah. Okay. I mean, I think there's probably three core things to address with that. The first, when those companies are trying to implement AI, it's knowing where to start is the first one. Have they got a clear strategy around what they want to do and how they want to do it? Often the companies that I've spoken to, both when I was consulting and also when I was deploying enterprise solutions, was that big businesses just don't have any idea of what they want to do and what they can do. So, we spent a lot of time advising, uncovering those opportunities for AI to be implemented. And that's the first one. It's just understanding that map of opportunity that you might have within your company.

I think the second thing that's really important is alignment across teams. You need buy-in, and you need buy-in from the top. That CEO, you know, I think it's often important for them to have mapped out what those company OKRs might look like and where AI might fit into each of them. I think everyone needs a mandate to be able to explore and experiment how to implement different solutions because I think it's all about trying different things. No one knows exactly what's going to be right for them from the off, because we are still at such an early stage of development for a lot of this technology. So, I think it's just about having an open mind, having that alignment, and having that strategy. The third thing, of course, is bandwidth. Do you have a team who is able to implement these kinds of solutions? Often, the case is they don't have machine learning engineers in-house. They don't have a team, and it's not feasible for them to hire in a team of very expensive ML engineers to work on projects. So, that bandwidth question is also one that they need to address. It's like, how are we actually going to implement something if we do want to run with it? I think those are probably the main things to consider.

Oleg

Thanks for the answer. Do you think there are specific industries or sectors where AI deployments in large enterprises have demonstrated significant impact or transformation?

St John

Yeah, I think so far, again, from some of the enterprise projects that I've worked on in the past with companies like Saudi Aramco, HSBC, other banks in the Middle East, banking and financial services can be big beneficiaries of a lot of the technology. Whether that's document processing, whether that's legal services processing, I think sales organizations as well are ones that we've clearly seen can really benefit from automation of top of funnel. There's a lot of work that goes into sales teams daily, you know, daily business, which is prospecting, which is researching, which is kind of the way we see it as like very top of funnel. The bottom-of-funnel work for a salesperson is that part of the funnel where they're most effective, so that's having conversations, meeting people, and doing that sort of close. For those sales teams, that part of their job is usually only sort of 20 to 30%. So, where automation can come in and take away 70% of what they would typically do manually, it can be hugely beneficial. I think there's kind of the LLM side of things. And for deploying LLMs in an enterprise environment, you've kind of got three buckets. You've got information, discovery, and synthesis how to get sort of deeper insights. You've got the hierarchical summarization side of thing and the use cases that can come out of that. And then you've got chatbots and the automation of customer engagement that you've also got there. So, I think those are some of the three buckets that mainly work into what a large enterprise looks at. But again, there's plenty of different opportunities. I think any company that has manual processes or automation can obviously have a huge impact there.

Oleg

What steps should businesses take when developing a strategic plan for deploying AI within their operations?

St John

I think when we've looked at these questions in the past with companies, it really makes a lot of sense to understand what does success look like, and how does that success relate to the production and deployment of the technology. Everything needs to be tied into a business's overarching goals. So, we need to look at what those are from the fundamentals: what sits in their OKRs, how can we build back from what the end goal is for each company, do they understand what they want? Much as I was touching on earlier, do they have any idea of what's possible? How can we work with them to enable some solution that is going to address those things that they want? I think what we also really try and do often with companies and what we've done in the past is show them what's working for others. Are there comparable scenarios within their business? If they're not as far along in the journey of understanding, we can give them those examples and hopefully talk to some of those to give them inspiration. But I think this roadmap planning for AI deployments is super, super important and needs to be addressed from the top down, basically.

Oleg

Could you highlight any common misconceptions or pitfalls that businesses should be aware of when formulating their AI deployment strategies?

St John

Yeah, I think there are probably two sides of this. The first is that AI is just going to work from day one out of the box. It's often not the case. You know, this is software. You need to work out the plan for it. You need to really put something together and put a strategy together that is going to be custom to your business. And I think there is going to be an element of experimentation that a company needs to be prepared to do with testing solutions. That's probably the main thing. I think the second misconception is AI is gonna take all of our jobs. You'll get quite a lot. And I noticed this in a couple of the use cases that I've had in the past. You'll get quite a lot of middle management being quite reluctant to implement AI because they think medium-term, it's gonna affect the size of their team, it's gonna create a lot of shakeup. And this idea that AI is gonna take all our jobs is something that we really need to try and get around. We need to understand and be able to convince people within organizations that AI is there to supplement people's jobs, not necessarily replace them from the off. I know that there are implications of deploying AI solutions for some businesses, but in many cases, it's not simply that we're just going to sack 50% of your team and put in a bot. It's that we're going to use AI to supplement the workload of people in a positive way. So, I think those are the two things. There are some roles that are going to be in trouble long term. We know that, but I think there needs to just be an openness to it. I suppose there's a bit of a reluctance sometimes in some big businesses to deploy immediately because of that fear.

Oleg

Got it. Large enterprises often have complex structures. What do you think are the key considerations when deploying AI solutions in large enterprise settings, and what are the potential benefits and pitfalls of implementing AI at a scale within a large organization?

St John

Yeah, I think the main one that I've come across in the past is data privacy. That's often, particularly with some of the large banks and global businesses, been a requirement for solutions to be deployed fully on prem or on private cloud. They're not willing often to send any data to OpenAI, let's say, through an API or WebIP. And that's generally been the big enterprise concern is how can we have a custom or a fine-tuned LLM that's working for one of our teams that we can also deploy on prem? That's, hopefully, something, again, you've noticed with OpenAI, they have OpenAI for Enterprise now where they're trying to assure large enterprises that their data is safe with them. Quite how effective that is, you know, time will tell. That's a similar solution, quite a similar scenario, I imagine, to when companies migrated from on prem to the cloud. Potentially, it will be the case that sending data to OpenAI or to other providers is going to be fine in the end. And that it's not going to be a problem for certain businesses. But at the moment, big enterprise, from what I've seen and from what I've worked on, they're still super, super cautious. For LLMs themselves, you've got hallucinations, which are also a big consideration, and you've got attribution problems. In an enterprise setting, these things can make certain AI technology unusable because if they are concerned that an LLM that they've had even fine-tuned their business is gonna hallucinate, and it's gonna hallucinate in front of a customer, then it can be awful. I dunno if you saw the news about the DPD chatbot earlier in the week.

There was a DPD chatbot that, you know, the delivery company that was. It was being used by someone online. They were asking it all sorts of questions, and then they told it, 'Can you write me a poem about how awful DPD is?' And the chatbot came back writing the poem, and then completely went rogue, and started saying how rubbish the company was and stuff like that. So, there are ways, I mean, that's not a hallucination. It's probably a bad, bad model. Whatever it is that they had, it was using an LLM. And things like that, you know, they can reflect really badly on the company, and it's going to make people cautious. Similarly, with attribution: if you're using an LLM, but you don't know specifically where that piece of context that it has come from, then is that a risk you're willing to take? Often, the case will be no. This is where companies like Perplexity AI coming in with really interesting products that have that attribution layer to them and enable a little bit more confidence in that content that's actually being put out. So, those are some of the main issues, I think.

Oleg

Okay. When it comes to starting an AI-focused company, what factors should entrepreneurs consider when deciding between seeking external investment and bootstrapping? And are there specific challenges unique to AI startups that entrepreneurs should be prepared for regardless of their funding approach?

St John

Yeah, definitely. This is something that I've been thinking about a lot recently. And it's something that we've been wondering what to do with, as a business. So, there's a lot of focus I think this year on profits and revenues. We're well and truly out of the crazy days of just raising mega valuations and not showing any traction. So, I think that needs to be a real focus for any company on making sure that it's going to be sustainable. That's certainly a lot of the feedback that we've been getting from VCs. Evaluation expectations are being tempered and need to be tempered. Founders are very aware of that. It's certainly something that I've become aware of. And, I think, from a company's perspective, a Founderreally needs to consider a few things. Do they want to control their own destiny? That's one of the things that working with a lot of funding doesn't necessarily always align with the ability to do. The second thing is do they want to scale quickly? Is it important for them to scale quickly and not necessarily think about profitability so much, but think about getting the product in use? That's a big one because if the answer to that is yes, they want to scale quickly, then it often makes sense to raise money.

Thirdly, you know, is a healthy, moderately growing business, more aligned with your lifestyle and what you want as a Founder. And that's another question that we've been asking ourselves. It's like, which of these approaches makes the most sense? But, like I say, the key thing that all startups, AI or otherwise, need to be aware of when they're going into these VC conversations is what's the revenue track, and how much are you thinking about that from day one? Because it seems to be very important. The other thing, actually, is AI startups in particular need to be able to show differentiation. There are a lot of AI startups, and there have been more AI startups created in the last year than ever. So, there's a relatively saturated market. You've got to be able to show differentiation, you've got to show traction. But you need to really stand out if you're going to appeal to companies from a funding perspective, and you need to show how that could be defensible as well. Obviously, the technology itself has enabled the creation of thousands of companies, which is great, but which ones will last and which ones won't is still very up in the air, and we'll see how that progresses over the year.

Oleg

Many of those startups have nothing to do with AI in reality, and they don't bring in value.

St John

Yeah, definitely. There's a lot of startups that have just stuck AI after their name. I think that that will play out. We'll see a lot of those come to nothing in the next few years.

Oleg

Could you please comment on the problems associated with the lack of qualified AI engineers in the IT sector, namely in connection with the industry you operate?

St John

Something, I mean, it's often mundane that companies or CEOs will say, 'You know, we have to do something with AI.' Building a team of internal machine learning engineers is not cheap, and it's not possible for a lot of companies. But companies like ours as well, we're a relatively lean team. We also have to think about how can we bring in talent in the best possible way. Finding good talent is particularly difficult. In 90% of cases for us, particularly in the past, we've had a lot of project work, it hasn't made sense to hire full-time machine learning engineers to support projects. So, we've had to look into our networks and establish who can help with various projects. So, I think it's really important. It's hard to find good ones. It's something that really needs to be addressed in terms of what are the benchmarks for quality, how can you be introduced to somebody new who's actually going to be able to do the job. And I think that's it. Finding a few trusted partners in that field is really important, but certainly for someone like myself, who brings in quite a few different people from ML at different stages.

Oleg

Could you share how the development team is structured in your project, and have you ever outsourced your tech needs to an external vendor?

St John

Yeah. So, we're a small company. So, our dev team consists of CEO plus several ML engineers that we work with on a project basis. They're typically people that we have known and worked with before, and will be brought on depending on the complexity of an individual project. We have worked with outsourced businesses before for very specific things, particularly recently around UX and UI design. Again, we're looking to optimize the skills of the engineers that we have in-house. And often, because of the way that our company is structured, because we are an AI-first business, those skills don't sit with front end. Those skills sit with deep machine learning expertise. And we need to be able to hire and bring people in who are able to fill those gaps, basically. And it's not going to make sense for our CTO to spend three days doing a bit of UI when we can have someone brought in who is much more of a specialist in that area.

Oleg

What were the precise factors that prompted you to consider IT outsourcing?

St John

So, the key one is capacity. Capacity and budget. Those are the two main things.

Oleg

What are the benefits and drawbacks of IT outsourcing?

St John

Yeah, I think benefits are the flexibility that it enables. And the cost implications of that are obviously great. The drawbacks, for me, it's not having necessarily someone always there to sit in the office with you, and be able to really go deep on a project, and talk things through in the same kind of way. Obviously, sometimes if it's outsourced, and they are in the office, then that's great. But I think that long-term cohesion of vision can sometimes be a little tricky. But again, if the project is very specific, they know what they're doing, you know what you need them to do, then I think that cohesion is generally quite easy to get.

Oleg

How do you measure the success of your collaboration with an IT outsourcing vendor?

St John

So, I suppose it will just be specifically on the project itself and what's actually being delivered. This is typically something that sits mainly with our CTO prom, but it's what can be shown in some tangible results after a period of engagement with an outsource vendor. Has the project been achieved well? Is it working? Do we have the results that we want? So, it's usually fairly straightforward. We'll set specific goals ahead of the project so that we know whether we've hit them or not.

Oleg

Great. And finally, what advice would you give to other companies considering IT outsourcing right now?

St John

Yeah, I'd say, firstly, do your due diligence and just understand who you're working with, what kind of talent they have, and how that talent then fits in with specifically what's on your roadmap and what you need to do. Again, I think recommendations are always a good thing. If you can get a recommendation to a specific outsourcing company, then that's always great. But then, just understanding what that talent makeup is and how that fits in is the key thing.

Oleg

Thank you. Thank you, St John. The information you shared is definitely valuable. Thanks for sharing insights about Google plans to implement additional filtering. I hope it will not break cold outreach industry much. I'm sure there will be overcomes in the future, even with those new algorithms. Thanks for your time. Thanks for having this episode with me.

St John

My pleasure. Thanks, Oleg. Good to chat with you. Speak again soon.

Oleg

Bye-bye.

St John

Bye.