5. Our Pricing Framework - SCAI017

All right, hey guys, today in this video, we're gonna be talking about pricing and it is obviously a topic that has been interesting and it's been in high demand in this space. Everyone wants to know about pricing, how it works, how you should do it. You know, there's no like baseline really to work off of. Every agency's pricing is so different right now. Clients don't have a great expectation of what they should expect to set aside for a specific project. So just gonna talk about three main things today. So first one's gonna be the current way that we're pricing and we've kind of wrestled with this for a while. So right now, this is how we're doing it. That's not to say it's the best way, that's not to say it's how I'm gonna be doing it in a month, but it's what we're doing right now and right now it's working. So if it ain't broke. Then gonna talk about key factors to consider because that's sort of the more evergreen mindset here is what you consider, how do you get to that number and yeah, because that's kind of more important to know, I guess, than one standard framework. You know, same thing with like, what's the best way to build this type of workflow? Well, that really varies. So what you wanna do is you wanna understand the skills that go into building workflows so that you can build all types depending on the use case. So just trying to understand that evergreen stuff, right? But anyways, and then I'm gonna at the end show two projects that we've done, talk about the functionalities and what we ended up charging. So let's just dive right into it. So first thing here we're talking about is our current pricing framework. So I just took this little snippet from one of our proposals and I just cut out the actual numbers from this one. So anyways, the objective of this one was just kind of a basic AI powered personal assistant. Before this, what goes into it is, you know, discovery call, second call, third call. It ranges from two to five calls really. I've had a client that we did like six calls and then we didn't end up getting the business. But anyways, the whole point of those calls is because you really have to narrow in on exactly what are we doing here? And not only what are we doing here, but what will we be doing in the future with this client as far as deployment, support, project two, project three, project 10? Is that even a possibility? All that kind of stuff goes into play and you really wanna be careful about who you end up partnering with. You know, you'd rather have five quality partners than, you know, 20 kind of pain in the butt ones. Anyways, so the point I'm trying to make is we don't like to throw out a number or pitch a proposal until we're pretty sure that we like the goals of the company. We are very clear on the scope of work because you never know if you get down near the end and it's like, I'm not paying you this last payment. This isn't what we agreed on. So having a very clear scope of work, of course, check boxes that you can go down as you're developing and knock them off and say, there's no way they can argue this point that's in our agreement. They can't say that this isn't what we delivered because we have it right here. So check boxes that we can say, this is what you get, this is what we agreed on and this is how it works. And then, you know, things come into play with long-term opportunities. And there's a couple of ways I mean that, you know, sometimes it's like, hey, we kinda just wanna get a quick win. We wanna establish trust with your agency. We know that AI is huge, we need to start using it. So we wanna get this quick win under our belts and then we'll bring on bigger projects and bigger projects. And so sometimes it's just about, okay, let's just lock in this first build. And then it's also like, okay, you know, I have this need and I think that because I'm a subject matter expert in my industry, I know that everyone else that has the same role that I have also has this problem. So let's work together on you build this for me. If it's good, then let's go sell it to hundreds of other people as well. So then there's also that type of long-term potential as well. But it's been four minutes. Let me just start looking at the actual document that I put together here. So anyways, what we do is kind of like a menu of options. We like to do three implementation tiers. Sometimes it's two, sometimes it's three, depending on the actual scope of work. And then at the bottom, what we like to do is have additional add-ons that can be added to any tier. So let me just discuss this real quick. Two reasons why I think this is a really powerful move. The first one is because people come to you with a pain point or a couple operation, a couple processes that they want optimized and they don't know where to start and they don't know what's exactly feasible. They don't know how much this stuff's gonna cost. So when we can come in here and understand their problem and pitch to them, okay, hey, we thought about it. We've talked about it internally and we have these three tiers that you can look at. It's gonna show our expertise in knowing functionality. And usually what we end up doing is we end up making this middle tier kind of what we discussed on the call. We make the basic tier a little bit less. This is just in case they didn't think we were gonna charge them 10,000 and then they can pay five or 6,000 to get at least some basic functionality. And then this one is like, oh, wow, I didn't even realize that it was possible to have these features. But if that's what we're doing, I'd love to just go all in on this one and we'll take these extra features as well. So that's why we do that. And so this is just a rough outline of a personal assistant that we have these tiers for. So I didn't even hit on the second reason. I guess the second reason was really just saying I think it shows expertise. It shows, okay, wow, these guys really listened to me on these past two, three calls. They understand the business. They understand my goals. And then they develop this sort of system where they're showing that they put time and effort into this. Anyways, so basic tier functionality for a core AI assistant. We've got vector database ingestion. And this would be an example where we have like, okay, we're gonna give you this manually updated vector database where we'll build you a pipeline, but you're gonna have to go into NNN and hit test workflow if you wanna keep updating it. And you're gonna have to manually delete the outdated vectors. And we've got like basic task management, to-dos, reminders, calendar scheduling, email and message drafting, of course, with generating responses based on the information in the vector database or different, CRM information, whatever it is. Slack integration is your communication channel and basic CRM integrations. We don't always do this ideal for, but sometimes if we feel like they may not be able to understand the differences, we'll throw that in there. Man, I had something good I was gonna say. I'll get back to it, I'm sure. But something else that sometimes comes into play is like, how rigorous are we gonna do your testing with your vector database? Are we gonna test different embedding models as well as different vector databases? Or are we just gonna say, hey, we're gonna use Pinecone? Because sometimes in the advanced one, we'll say, okay, well, we're gonna test Pinecone, Quadrant, and Suidbase for you. Or we're gonna test embedding, we'll test OpenAI's embeddings, we'll also test like Azure, all that kind of stuff. So that's another way you can increase the value is because we're saying, hey, we're just basically gonna spend more time up front testing, but also maybe even more time on the backend. Maybe not on the backend because we always wanna do pretty rigorous testing with the prompting and making sure that this thing is working for them. Because obviously, what's the most important is that they're happy with the solution. But yeah, data in, data out. So we really try to say, we wanna be able to really do some advanced testing here for you with the metadata tagging and the chunking and the splitting and I mean, it's kind of the same thing. Anyways, advanced tier. So you can see it's an upgrade from this tier, of course, because we've got multimodal and voice support, we have enhanced CRM capabilities, full read-write access into, and database manipulation compared to just that read and write functionality. Web search and content creation, it can generate summaries, reports, create content, and then we have the automated dynamic data pipeline, which is usually a big one, because people were like, well, I don't wanna have to go in there and I don't wanna worry about it if my information's up to date, I just want it to be up to date. So here's where we build out the automated data pipeline, which can pull either a folder, a Dropbox, emails, really whatever it is, it's just about understanding that upfront. And I know this video isn't meant to be really about gathering requirements, but that comes into play with the pricing. But it's so important, you have to understand everything about what goes into this build, and I'm not even saying I'm great at it, I've gotten better over time with doing all these discovery calls, but sometimes my team will be like, Nate, we can't propose anything yet, we don't understand anything about these API calls or the schemas of this database, we don't understand the cadence that we're gonna be running this workflow on, we don't understand the expected traffic, we don't understand all that kind of stuff. So that's why it can be multiple calls that you're hopping on until you can really confidently throw out a price. And then finally, that Premier tier would just have real-time voice integration, a custom analytics dashboard, this is usually us just throwing out, okay, how can we make this thing even cooler? And sometimes they'll bite on it, right? So anyways, with the additional add-ons, sometimes this is going to be features that we grab from maybe right here. Maybe they want the basic one, but they just wanna be able to add in the automated data pipeline and switch that out for this. So we would just say, okay, that'll be an extra $1,000 or something like that. But then there's other things that sometimes they'll choose from if they want us to help fully containerize everything and deploy it for them self-hosted, or voice model can be added to anything, continuous AI model if they want us to get in there and fine tune and be there for that type of support. So all that kind of stuff, custom front-end UI. So anything that maybe they can just add on, which I think they also feel like, oh, wow, I didn't realize that there's all this stuff that can be done. I see that they're probably delivering projects for 15K up here, but I'm still probably getting a steal if I go with this basic package and maybe just add something on. So, okay. Hopefully I didn't spend way too long there. Okay, but key factors to consider while pricing. So I've got some high level points right here, right? So the first one's really about like the project scope and complexity, obviously. So what type of solution are we delivering here? Is it gonna be a simple AI chatbot system, or is it going to be an architecture of many agents working together, many active workflows? Do we need, is it gonna be custom data pipelines that are gonna hook up from multiple different sources? Another thing to consider is how much data upfront do we have to ingest? And is it really clean or are their internal database systems pretty messy because that directly affects us trying to get in there and vectorize it and split it up. And if it's really messy and if there's just a ton, then that's going to make the process for us to vectorize it and build that pipeline and test it with like rag validation. That's gonna make that process much more painful. So understanding what their data currently looks like and how they plan to maintain it, because a lot of times, sometimes it's not actually changing like the workflow solution. Sometimes it's changing the data upfront before you even get into any of this. Like sometimes the requirement is, okay, if this is gonna work, once we're like while we're developing the solution and once we deploy it, you can't keep storing data like this. You have to do it in this cadence under the schema, something like that. So it's also setting the framework, the ground, setting the rules, the ground rules sort of about like how that's gonna work with their data. Obviously, probably the biggest thing that we think about is like how much time will this take us to get in there and develop? Obviously, the testing integration goes into that. Thinking about maintenance and that always isn't saying like pricing that in, but it's just making sure you have an open conversation before you sign off anything, before you get into the build of what does maintenance look like? What does deployment look like? What does the process of developing look like? Like what is the client's touch points in there? A lot of times testing and iteration is so important. And obviously, the performance of the agent based on the prompt is gonna be directly impacted by the quality of the prompt and the quality of the prompt is going to be directly impacted by the subject matter expertise that we're able to get from the client and put it in there. So the client's really hands-on in this process. And the more frequent the touch points, the more time you're gonna save at the very end when you're like, okay, here, test this, because then they're gonna have way too much feedback at the end, but getting feedback all the way through. And then this point about infrastructure and API costs also just goes back to like understanding and being very transparent about where are we hosting this? Do you need this stuff to be locally ran? Like what are your requirements with data security and privacy? And then what is this looking like when we hand this over? Usually the assumption is that they will be paying, the client will be paying for all of the subscriptions. So any APIs, anything that we needed to connect, we'll do that during testing. But then when it's over, they have to plug in all of their credentials and integrations if we're handing it over. If they want us to host it, then of course we try our best to forecast. And sometimes it's tough to forecast. I mean, they can say, yeah, I mean, we're gonna be running this thing twice a day, every weekday. And in that case, we can kind of estimate based on the system prompt and based on the data that's gonna be passed through, we can sort of estimate the token usage and the costs, but something like a personal assistant, that can be way more variable and you really won't know until maybe they use it for a couple of weeks. Anyways, licensing and compliance, obviously. Factoring in those costs, factoring in what that looks like from a compliance standpoint with different organizations, different industries, what the regulatory compliances look like for your data. And then the client's budget and expected ROI. So obviously you could have the same build and it will provide a different amount of value based on the size of a company. So, you know, like let's say it's a sales agent and you're building that for a small business that only has one sales guy that it's going to replace, it's not gonna be as valuable to a company that spends a couple million dollars a year on the sales team. So understanding the value add, and it's a tough thing to try to navigate, it's a very delicate dance because when it comes to pricing and that discussion, the client's always wanting to, you know, they want you to give them an estimate, they want a ballpark. Because a lot of times they're gonna be shopping around and getting quotes from other agencies, but I don't know, I don't like to be the one that throws out a number first. I try to get a sense of what their budget is, what their mindset is around AI. Are they bullish about it? Do they plan to invest in it over the next multiple years? Or do they just wanna get like one thing up and running? So trying to understand where their head's at with AI and understanding like how much time is this gonna save them? How much value are we really coming in here and adding? And so one of the first things we do on our proposals is we throw at the top like a value prop. And we say this solution is going to do X, Y, and Z for this company and save X, Y, blah, blah, and just try to show the value up front because like right here, like if your AI system saves a company half a million a year, your pricing should definitely reflect that value if it's gonna be like a one-time cost. It's insanely valuable because not only is it gonna be cheaper than salarying someone with also like their benefits and everything like that, but you've also got the aspect of like, it's working 24 seven, it's never gonna ask for a raise, you kind of own the solution once it's built out and it's on your system, you own it, you can tweak it, you can upgrade it, all that kind of stuff. And then obviously like the long-term goals and value, I touched on this and I'll touch on this again kind of in that last example, in the last example on this video that I'll show, but sometimes you have to make us kind of just bite the bullet and say, okay, I mean, if they want to do all of this, but they're looking for a quick win and wanting to build trust, let's just drop it by like three or 4K, let's really just secure this first build over or under promise, over deliver, lock in this relationship. And then from there, so much we can do with these guys. So yeah, but common mistakes to avoid, and this is the first one is undercharging. So these solutions, like once you understand the functionality and once you understand what it's gonna be doing on a day-to-day basis, they can provide so much value, so much time saved. And I've talked about the fact that like productivity tools, which a lot of times these AI things are, sometimes it's hard to see that value upfront before you actually have it integrated into your day-to-day life. Because something like a lead gen system, where you're explicitly saying, I'm gonna get you 100,000 leads per week, or I'm gonna increase your revenue by X amount per month, because I'm getting you X amount more sales. And so that's like a very clear, direct relationship of return on investment. But when it comes to productivity tools, like a personal assistant, these quoting agents, sales briefing agents, you may not understand the value in terms of the dollar amount until you have that system. And then if you were to have that system taken away from you it would feel like night and day. So if someone today were to take away my chat GBT, even though I had operated like that through school and through work for years, I would be like, this is the worst thing ever, just because I'm so used to now using it as a productivity tool. So that exact thing, right? And so when it comes to infrastructure costs, maintenance, mistakes to avoid would just be like getting into an agreement before you understand what that looks like, because then you could really be shooting yourself in the foot there. And then I don't think that we'll ever move away from a pricing tier model now, now that we really have seen an increase in conversion rates when we pitch out an option, rather than just saying, hey, it's this, or it's this deal, or you can walk. So we give them flexibility and then they feel way more in control with like, oh, okay, I can customize this however I want, I'm still in control. So I think the pricing tier is a pretty good way to go around that. So, okay, project one we're gonna be talking about. So obviously I messed up this stuff, so it's gonna be a lot more vague than the actual scope of work slash proposal that we sent to the client and that they signed. But here, what we're looking at here is a project, we wanna implement AI driven features to modernize a procurement platform. So developing a natural language query interface where they can talk to it in natural language and then access their data, right? So and then also an automated risk management and escalation system to assess and mitigate supply chain risks. So obviously huge value here, it's going to enhance decision-making efficiency, user experience, it's also going to give this company an edge because they're able to, you know, save a massive amount of money if something went wrong in the middle of the night and they needed to get alerts and they needed to reach out to people to make sure that they were getting what they needed to and that everything was working as it should. So the functionality overview of like this interface, right, is AI driven chat interface. Obviously it's gonna have RAG where we'll be pulling back information from that structured database because we're able to use natural language to query it in SQL. Role-based access control, automated notifications, custom views and insights, seamless integration, and then with the mindset of scaling on top of this. And so for this one, we were able to get the client at 18 to 50 and then we usually always just do initial payment upfront and final payment. And what really came down to this one was, this one was a lot of calls. We had to understand, you know, timeline, we had to understand the resources going into this, we had to understand all of their data, where it's all hosted, like, you know, the schemas of all that, if the tables are dynamically updated, if, or like the fields are dynamically updated or if that's at least standard. So obviously my team really helps when it comes to getting into the technical questions that need to be asked about all this kind of stuff. But yeah, like that's really all I can share about this one as far as like the actual build. But, you know, that's how we were able to price it. I think we could have gone more just knowing what we know now about the company and what they're looking to do here. Because also like, this just goes back to, for some people, for some companies, 18,000 is a lot, but this guy seems like this number didn't really phase him. So we could have gone more, but this is also another example of, you know, there's a lot that this company wants to do and we wanna be there, we wanna be their pick. So establishing that trust and just making sure we lock them in. And then this project, so project two, a little bit similar. We've got an 11 labs voice model with a RAG system for dynamic knowledge-based updates. The solution here is to have an automated, or yeah, an automated data pipeline that they are able to chat with through voice. And here's the value probably said, you know, by reducing manual workload and enabling faster decision-making, this innovation will enhance operational efficiency while ensuring compliance with GDPR and scalability for future expansions. So scope of work here, we have, you know, an automated knowledge base, of course, that's updating. We've got a RAG enabled voice model. We have integrations with external APIs to actually like create bookings and check availability, that kind of stuff. And then we have, you know, obviously the long-term goal in here, because this is one of the examples where we came in here and we wanted to do more than 42. I think we wanted to go like in the eights, but this particular client was a good name to have under the belt. And also, you know, he was pretty transparent about, we think that the solution could be pushed to tons of people, and we already know a ton of people that would probably want this. So yeah, that's why we did that, right? But we wanted to just lock them in and we knew that the customer lifetime value here was gonna be a lot more than if we were to pitch out eight, nine, or 10K. And then they said, oh, you know what, we're gonna go a different direction. So yeah, and it's just kind of a, sometimes you kind of just gotta go with your gut. I mean, I hope that this was helpful. I think that, you know, we saw massive improvements when we started doing this implementation tier system. So like I said, I know I can't share everything that we're doing and exactly how it works, but I hope this was helpful and at least interesting a little bit. But yeah, definitely let me know. Let me know your guys' thoughts. I would love to hear some of the other things that you guys take into consideration and maybe if you're willing to share some other projects that you've taken on and how you navigated pricing and all that kind of stuff. But cool. So I'm looking to do more videos like this in the future where I'm obviously wanting to share some insights with you guys in the space, but I'm gonna call it here for today. So appreciate you guys.