In this session, John Arnott, CEO at C1M, will share how AI can revolutionize your business strategy and significantly increase profits. We will explore the transformative power of AI, demonstrating how it can streamline operations, enhance innovation, and drive measurable financial results across various aspects of your business. Join this session to gain a solid understanding of AI technologies and their applications in optimizing processes, improving decision-making, and boosting overall profitability.
TRANSCRIPT
Greg Alexander:
Hey, everybody, this is Greg Alexander. You’re listening to the ProServe Podcast brought to you by Collective54. If you’re new to our show and new to our community, we’re a community dedicated to helping founders of boutique pro serve firms make more money, make scaling easier, and make an exit achievable. On today’s show, we’re going to talk about artificial intelligence, in particular, how it’s applied to the boutique pro serve firm in an area that everybody’s focused on these days, which is margin expansion. With me, I have a Collective54 member. His name is John Arnott. John, would you please introduce yourself and your firm to the community?
John Arnott:
Absolutely. I’m John Arnott. I’m the CEO of C1M. We are a professional services organization that specializes in helping services businesses grow with great marketing and artificial intelligence.
Greg Alexander:
Okay, very good. So there’s a lot of hype around AI right now. When we try to be relevant to our members, we’re trying to pull it home to the things our members care about the most. In the pro serve space, the biggest expense for a firm is labor. It’s somewhere between 80 to 90 percent of the expense structure. Therefore, I believe that AI has the potential to help our members make a lot more money by either eliminating headcount or increasing the output per billable hour of the existing headcount. That’s where I’d like to spend our time together. So John, let me tee up the first one. What are you seeing in terms of AI’s ability to eliminate headcount?
John Arnott:
The way I see it is we in professional services need to optimize the output of each person. In many cases, what happens is headcount doesn’t get eliminated as much as those people get moved to a different position or they grow into another position. There are certain roles that tend to go away. We started out our lives 12 years ago in this company doing digital marketing. Over time, as we adopted AI about five years ago, we have had roles that we no longer have in the company, but we also have people that evolved and changed with that. Now those people over the past two to two and a half years have become significantly more productive. We’re getting a lot more done with fewer people. I can talk to you more about those particular things during this conversation.
Greg Alexander:
Yeah, so let’s jump into those. I mean, that’s music to my ears and to our members here. So keep going on that thread, please.
John Arnott:
Great. If you want to think about how you’re going to take your pro serve organization and make everybody more productive, right? Get to that ideal blended bill rate that you’re moving towards. Don’t think about it as something you do all overnight. It’s not boiling the ocean. It’s a maturity model. There are certain steps that you should take. Some of them are super easy and some of them take putting in infrastructure to support it. When I think about that maturity model, the first thing is just building a culture around using generative AI and frontier models to do basic work. So that everyone in the organization is comfortable with things like Copilot or Claude or ChatGPT. They’re comfortable using those tools. The way you build that culture and that level zero of the maturity model is just let them tell you how they’re using it or how they think it could be used. We even ran a contest about a year and a half ago to come up with new ideas. Fast forward today, everybody’s using these tools for their basic work. So that’s like level zero and getting going. But when you really think about how you get the most out of this generative conversational AI, it’s moving into the next levels. It’s talking to your documents, talking to your backend systems, talking to your data and doing that in a natural language and doing it in a way that makes you much more productive. With that in mind, once you kind of work on getting the culture so they at least accept it and more than accept it, they come up with ideas on their own. Level one is talking to documents, right? Collective has their own trained AI that helps with some of the content that’s created, speaks in the right voice and structures things in a certain way. So there’s ways in ProServe that we can do that really effectively. I’ll give you an example of one that we do on a regular basis, and that’s proposal reconciling. For some types of clients, I’ll give you a scenario. We have a client and their deliverable was to do a digital marketing strategic plan, a long-term strategic plan with all the tactics that they need to do in that plan. When we presented it, the CEO was like, I’m not sure that’s everything I was expecting. So how did we handle it? Well, the old school way would have been I assign a staff person to go look at the deliverable. That staff person goes, looks at the proposal. The staff person works with the senior and manager that were creating that. They all get together and have a meeting and they put all this time into trying to figure out the gap analysis of the deliverable to the proposal. This is a scenario where I trained a very simple model. In fact, in this case, I used Claude, a publicly available one. I gave it the proposal. I gave it the deliverable. I fed it a prompt, not a super long prompt, but a prompt that described in addition what I wanted it to understand and what I wanted it to do with these two artifacts. See, the thing is sometimes people get a little confused when they’re using these simple chat type AIs and they think it can do everything. It can’t. You should think of them as an intern that you hired with a master’s degree. They have all this knowledge, but no real-world experience. So what I did is I created a prompt that told it what I want from it. Within moments, I have a gap analysis between a proposal and a deliverable. It literally gave me three items that said we could expand on these in our deliverable. Turn that over to the manager running the project. They added those things that only added a couple of hours of labor just to that piece of it. Boom, that client’s happy now. They got everything they wanted. And I’m able to say we delivered what we promised. I didn’t spend a lot of time doing that.
Greg Alexander:
Yeah, this is a great example. I mean, what you just described, that use case, which I love talking about use cases because it brings these ideas to life. In a normal scenario without Claude, that would have been the dreaded scope creep. I mean, you got that done in two hours. It could have been 20 hours. And that’s how the profits get destroyed. Because when the client’s unhappy, you’ve got to fix whatever they’re unhappy about. It’s tough to say to the client, yeah, I’ll do that. Give me some more money, right? They’re reluctant to that. And they’re saying, hey, you missed the mark. So that would be labor costs that you would have to eat. As a result of eating that labor cost, your margins would have plummeted. Instead, you were able to get that done in a couple of hours. That’s a fantastic example of leveraging the tool. Now, let’s go on to another example if you might have. I think that would, so in your maturity level, you got level zero.
John Arnott:
So now we’re on level one. We’re on level one. We’re able to use, we’re still using external tools. In this case, I use Claude, but that could have been Copilot. That could have been ChatGPT. Could have been an external tool. Now let’s move to level one, or level two rather. We talked to your documents in level one. And we can go a long way on that. We can have an entire conversation for an hour just on that. Let’s talk about level two. That’s talking to your dedicated back-end systems, right? That’s systems that you use in your everyday business that have some form of API. The example I can give is we use a project management system called Asana. And in Asana, we run everything. Every project has hundreds of tasks. There’s so much in it. We have templatization, automation. We have all that. But sometimes I just want to know what is the status of this project? Now I’m sure I could go see if it’s code red, but I want to ask a question or I want to know, do we have any overdue tasks? Because I’m about to go into a client meeting, as essentially the CEO of the company and I’m not in the project. So I don’t know all the details. I want to know what does that look like? So we created a simple conversational interface that I will ask the question, how many overdue tasks do we have for this project? Or what are the tasks that we’re waiting on the client for, or which one of my team members has taken the longest on getting things done? And I can ask these in natural language. And we created a simple bridge. The bridge now bridges a language model that’s been trained. It hasn’t been fine-tuned, but it’s been trained to work a certain way. We created a simple bridge to talk to the API. So it knows whenever you talk about these topics, it uses these types of API calls. And then it’s just using what Asana already has built in an API. And that interchange, when it pulls all that data back from Asana, that goes through a foundation language model and converts it into answers to my questions. So I can interact with Asana that way.
Greg Alexander:
Can I ask you some questions on this? Because this is also fascinating. Because in that scenario without AI, you’re heading into a meeting, you’re the CEO, you’re not in the project. You know you’re going to get asked a bunch of questions. You don’t want to look stupid. So in the old world, you would have had to have called a staff meeting. That could have been three, five, 10 people in the meeting. For an hour or two to give you a quote-unquote update. So just do the math on that. That’s what, 20 hours of cost associated with it? Now you’re not doing that. You’re talking directly to your system and it’s giving the answers. I mean, just another example of a massive time save. Now the word talking to, I find intriguing. Conversational. Are you typing? Are you talking? Or does it matter?
John Arnott:
Fair question. At this stage in our own company’s maturity, we’re doing this through typing. At this moment in time, it can be done through spoken word, like pick up a device and talk to it. But I haven’t deployed assets in our own company to go create that piece of it. But it’s totally doable that way.
Greg Alexander:
Okay, very good. All right, let’s go to the next level of maturity.
John Arnott:
All right, so after we’re talking to our dedicated systems through some sort of language to API interface, let’s talk to our data. So for some companies, they have a data warehouse, a data lake. They have some sort of repository that aggregates data from different sources. And this is where you want to use natural languages. And I really got excited about your readiness checklist because your readiness checklist is essentially now what I would say is our specification document for our own internal model, right? Because I want to be able to ask those questions. What is our trending margin? What is our client retention rate over the last three years? Things like that. So how do we handle this? Well, that’s now a two-step process. In fact, we’re doing this for an organization right now. They were all excited because they wanted to be able to simply ask, what is our success rate? Which is a metric. Over the last three months. Or what was our injury percentage year over year? Things like that. These are simple KPIs that are important for their business. But when we got in there to talk conversational AI, we realized they didn’t even think through the data aspect, which is building a data lake or a data mart or a data warehouse, however you want to perceive that. So for them, what we did is we worked together and said, okay, these are the conversational pieces. This is how we would solve that. These are the KPIs we need to do. Let’s go build that. For some companies, they already have it. Once you already have that, instead of having a data scientist answer all of your questions for, you can build a layer that yes, you might still have your data scientists doing the really heavy lifting. But for the simple things that are already residing as KPIs in your data warehouse, ask it those questions in natural language, typed or spoken.
Greg Alexander:
Another great example. You know, we at Collective. When we talk about scaling a pro serve firm, that can mean a lot of things, but at the very basic level, what it means is decoupling the rate of revenue growth with the rate of headcount growth, which gives you operating leverage. So if I’m growing revenue at 30 percent, but I’m only growing headcount at 10 percent, over time, if that compounds, now all of a sudden I got an extremely profitable business. Years ago, before these tools and those that were the predecessors of these tools, revenue growth and headcount growth went in parallel to one another. So it was incorrectly believed or rightly then, but incorrectly now, you couldn’t scale it. Professional services firm. Yeah, you might be 30 percent bigger, but you also have 30 percent more headcount, probably more expensive headcount. So you really can’t make that much more money by having machines do the work as opposed to people doing the work. Then you can really scale these suckers because the productivity increase, which is what John is talking to us about today, is you’re using technology tools to make existing staff that much more productive. So you might not be eliminating headcount, but you’re capping the need for future headcount. Which can have just an incredible impact on the pro serve firm. Okay, so we talked about talk to docs, talk to systems, talk to data. In your maturity model, once I get through those three stages, what level am I at?
John Arnott:
Well, that’s essentially your level three. And now what you move towards is, once you’ve got to this, is being thoughtful of other ways that you can use these types of tools. So what these tools are really good at, and when I say tools, I mean Frontier models, foundational models. I mean, tools that we are accustomed to already now today. What they’re really good at is translation and summarization. Translation means I have a concept here and that concept has the same meaning, but it’s different here. Anything that’s comparative in nature, these tools do really well. And summarization means it takes large volumes of data, that could be numeric as well, and crystallizes it into something. So whenever you have a person on your team and they do a repetitive task, that’s taking something over here and fitting something to another place over here, which is either concept to concept or simply summarizing to something smaller, these tools can do it. So that’s where you really can start saving the labor for resources, is all of these administrative functions of what we call translation and summarization.
Greg Alexander:
That sounds complicated. I know it’s probably not, but to me, the layman, that sounds complicated. Like in order for me to get my organization from level one to level five, does this take five years? Does it take a million bucks? Like how do I go from point A to point B?
John Arnott:
Yeah. So, you know, level zero is first, right? Just get it, building a culture around it and just making it acceptable. You know, there are some companies that in some cases just tell their employees flat out, don’t use them. Well, guess what? Their employees are still using them. These are level zero tools that they’re using for free. When, if you paid 20 dollars, you wouldn’t have to worry about as many security issues, right? That’s getting through that. So once you’ve gotten through that, or you as a company are saying, we accept that this is the way everyone will do business, in the future, now and in the future. Once you get past that, then it’s not a very long process for talking to your documents, right? So talking to your documents is only as good as the documents you have or the people you have or partners you have to create good documentation. And documents are proposals, right? So we write proposals all the time. Oftentimes they’re built off of templates. Well, I’ll give you an example where you take a templated proposal that you’ve worked on. Then you have a meeting with a client before you do the proposal and you’re getting lots of information from them. Maybe you have two meetings. You record those, you transcribe them, and you use these tools to take your template and take everything you talked about in that meeting and make sure it’s tailored for them. Those are simple things that you can do within a couple of weeks, a couple of months. That doesn’t take very long. Talking to your systems, it depends on the system. Systems like Asana, you can get going in a matter of weeks. A much more sophisticated ERP system may take a few months. Talking to your data, if you don’t already have some sort of centralized data repository, that’s going to be much more complicated because you want to go through the proper task of doing a data lake and essentially building that together. That could take months. That could take a year. It really depends on what you need to create in your organization.
Greg Alexander:
Sometimes I hear from my members, they say, hey, I want to use AI. I want to expand my margins. I want to increase my output per billable hour, et cetera. My clients won’t let me use it. I laugh when I hear this because I say, listen, do you ever talk to your client via Zoom? And they’re like, yeah, all the time. I’m like, you realize you’re using AI. Do you ever talk to your client through Outlook or your email application? Yeah, then you’re using AI. Do you ever communicate with your client on your mobile phone? Yeah, then you’re using AI. It’s almost a ridiculous statement when the client says you can’t use AI. Give me a break. What isn’t penetrated by AI right now? Everything we do. I mean, do you listen to music on Spotify? AI. Do you watch movies on Netflix? AI. I mean, it’s just a fear. And it’s our job as the experts in professional services to explain to the client what’s in it for them. And the scenarios that you laid out today, you could say to the client, listen, fine. You want me to do everything manual, I’m going to double your price because I got to pay these people to do these manual tasks. Or I can use tech and still be cost effective for you. What do you prefer? And it’s our job to overcome those objections. But John, this was a really great overview of what’s possible. You gave us a great structure to think through it. I love maturity models. In particular, talk to docs, talk to systems, talk to data and how that moves you up the stack, so to speak. We’re going to have a much deeper conversation with you when we have a private one-hour member Q&A session. And for those that are members that are listening to this, please look for the Outlook meeting invite that will come your way. And I know for a fact that our members are going to be able to or are going to want to ask you much more deeper questions than a podcast allows for. These are only 15 minutes in length. So on behalf of the membership, thank you so much for being here today. And we look forward to the upcoming session.
John Arnott:
Thanks, Greg. Looking forward to it.
Greg Alexander:
All right. And a few calls to action for listeners. So if again, if you’re a member and you want to attend John’s session, please do so. We look forward to that. If you’re not a member, you want to become one, go to Collective.com and you can book a call with one of our reps or just submit an application. We’ll get in contact with you. If you don’t want to do either of those things, but you want to read more and learn more, I would encourage you to check out my book, The Boutique, How to Start, Scale, and Sell a Professional Services Firm. Written by yours truly. You can find that on Amazon. But until next time, I wish you the best of luck as you try to grow, scale, and someday exit your boutique.
Note: This transcript was generated by Gong.