You're going down the right path. Too many people are focused in getting something in place now and don't understand the importance of building the foundation. Impatience now, will hur them later if they don't take the time in setting this up. Appreciate your content.
Thanks man, this is good to hear. We'll continue to dial in our process more and more over the next year. Excited to see how good we get at building and consulting on these systems
@ your one of the few who saw the death of SaaS way back. Also grateful that you Guided me to learn N8N now my entire is build on it. Keep pushing brother
this was so great, devin! I'm curious: how do you approach testing these agent solutions? do you hook it up and throw real customer data at it? do you have an internal testbed that you work through scenarios? my intuition is that you and customers share some unknowns which are hard to anticipate in testing - how do you prep/mitigate for those?
Great question. We're standardizing our QA & Testing process right now. What we've been doing in the past is set up demo data in the clients environments to essentially create a dev environment, then switch to prod when testing is complete.
Your approach is good; building up the foundation is the key for future implementation and AI adoption. Automation processes stand alone for specific use cases can't be scaled up, but they may be an good entry point for the company to start with AI adoption.
How do clients feel about token usage. I’ve had a lot of agent setups absolutely eat through credits when I put them into a hierarchical structure with a supervisor agent.
I think your approach is good. But I also think that doing a full blown discovery out of the gate might be too much from a sales perspective. Why not get a quick win, learn more about the client, then upsell / get some MRR with all the other automation they will likely need?
Yea this is a great point and something we've been wrestling with. Here's our thought process: If we can show a ton of value up front, it helps the prospect understand how we work, what our goals are, and how long some of this might take. During the sales process, we're also able to talk about other parts of the business that we can tackle. Having this forward facing approach is crucial because it can change how we structure the workflows (especially the data pipelines and databases) at the very beginning. Sometimes we have to rip out something we've done before because it wasn't designed for the new use cases we're adding. However, with all of that said, about two days ago I decided to add a Pilot Project as one of the offers that would be that first initial gateway before agreeing to a full retainer. So we're essentially taking your approach hah.
@@customaistudio Yeah. There is definitely some balance needed. But I think the more you do it, the more reusability you can get out of it. Sure, you might need to swap out the CRM, or the trigger, or whatever, but the "guts" of the process remain the same. I like to go in with some established low hanging fruit, deliver, then sell bigger agreements on the larger ticket stuff that requires more customization. Essentially productizing the easy stuff to kick off the relationship. But I definitely get what you're saying about future proofing. Always a balance!
I’ve been wondering about the topic of SaaS. Do you think it could pose a threat to us as creators of AI agents? For instance, could a new software emerge in the coming months that makes integrating these agents into businesses so easy that it renders creating them ourselves meaningless or obsolete? Nice video by the way :)
I think that's always the concern with new tech. YC seems to think vertical agents are the right play because they serve a very specific niche with a very specific pain point. They're probably right. I imagine the big players will capture the mass market products, and well funded tech startups will capture the ai infrastructure & tooling needs, which would leave the final application layer up to developers. This application layer may be platform specific where the platform creates a developer 'marketplace' of sorts (like plugins or apps) where individual developers can sell their agents, or build custom agents for their clients. With all of that said, I don't see SaaS existing as it does today - there's just no way it will more efficient for a human to manually click around or log data if an agent can do it.
You're going down the right path. Too many people are focused in getting something in place now and don't understand the importance of building the foundation. Impatience now, will hur them later if they don't take the time in setting this up. Appreciate your content.
Thanks man!
Keep pushing. You’re on the right track.
Thanks man, this is good to hear. We'll continue to dial in our process more and more over the next year. Excited to see how good we get at building and consulting on these systems
@ your one of the few who saw the death of SaaS way back. Also grateful that you Guided me to learn N8N now my entire is build on it. Keep pushing brother
this was so great, devin! I'm curious: how do you approach testing these agent solutions? do you hook it up and throw real customer data at it? do you have an internal testbed that you work through scenarios? my intuition is that you and customers share some unknowns which are hard to anticipate in testing - how do you prep/mitigate for those?
Great question. We're standardizing our QA & Testing process right now. What we've been doing in the past is set up demo data in the clients environments to essentially create a dev environment, then switch to prod when testing is complete.
@@customaistudio thanks! keep up the good work, man - you always bring an extremely thoughtful and insightful approach
Your approach is good; building up the foundation is the key for future implementation and AI adoption. Automation processes stand alone for specific use cases can't be scaled up, but they may be an good entry point for the company to start with AI adoption.
Thanks man
How do clients feel about token usage. I’ve had a lot of agent setups absolutely eat through credits when I put them into a hierarchical structure with a supervisor agent.
Usually the ROI makes the usage cost irrelevant but sometimes it becomes an issues. I don't have any real clear thoughts about it tbh.
another banger 🔥
Thanks brother
Great vid - are you doing this as a pre-sales activity or are you charging for it?
This is all pre-sales. We go deeper with metrics and Discovery calls in our paid offer.
I think your approach is good. But I also think that doing a full blown discovery out of the gate might be too much from a sales perspective. Why not get a quick win, learn more about the client, then upsell / get some MRR with all the other automation they will likely need?
Yea this is a great point and something we've been wrestling with. Here's our thought process:
If we can show a ton of value up front, it helps the prospect understand how we work, what our goals are, and how long some of this might take. During the sales process, we're also able to talk about other parts of the business that we can tackle.
Having this forward facing approach is crucial because it can change how we structure the workflows (especially the data pipelines and databases) at the very beginning. Sometimes we have to rip out something we've done before because it wasn't designed for the new use cases we're adding.
However, with all of that said, about two days ago I decided to add a Pilot Project as one of the offers that would be that first initial gateway before agreeing to a full retainer. So we're essentially taking your approach hah.
@@customaistudio Yeah. There is definitely some balance needed. But I think the more you do it, the more reusability you can get out of it. Sure, you might need to swap out the CRM, or the trigger, or whatever, but the "guts" of the process remain the same. I like to go in with some established low hanging fruit, deliver, then sell bigger agreements on the larger ticket stuff that requires more customization. Essentially productizing the easy stuff to kick off the relationship. But I definitely get what you're saying about future proofing. Always a balance!
I’ve been wondering about the topic of SaaS. Do you think it could pose a threat to us as creators of AI agents? For instance, could a new software emerge in the coming months that makes integrating these agents into businesses so easy that it renders creating them ourselves meaningless or obsolete? Nice video by the way :)
I think that's always the concern with new tech. YC seems to think vertical agents are the right play because they serve a very specific niche with a very specific pain point. They're probably right.
I imagine the big players will capture the mass market products, and well funded tech startups will capture the ai infrastructure & tooling needs, which would leave the final application layer up to developers.
This application layer may be platform specific where the platform creates a developer 'marketplace' of sorts (like plugins or apps) where individual developers can sell their agents, or build custom agents for their clients.
With all of that said, I don't see SaaS existing as it does today - there's just no way it will more efficient for a human to manually click around or log data if an agent can do it.
@@customaistudio I agree with you also.