Ep. 219 – What happens when a platform migration goes wrong, and what does it look like when it goes right? In this episode, Chris shares two real migration stories. The first is a learning platform move that left courses, audio, and video missing right before the old subscription shut off. The second is his own CRM migration that finished with no data loss and a fraction of the automations. He explains why the difference came down to how the AI was set up, and why asking AI questions beats telling it what to do. Tune in to learn how to put an AI agent to work on your next platform move, so you’re not stuck doing the copying and pasting yourself.
Migration From Hell to Heaven
What You'll Learn
- 4:19 – The learning platform migration that fell apart days before the old subscription cut off.
- 7:06 – The 80/20 flip that sank the whole project.
- 10:06 – Why “I’ll have Claude export a spreadsheet” reveals your real AI skill level.
- 15:16 – Why agents match IDs instead of names, and the access you need to give them in AWS, Cloudflare, and more.
- 18:13 – Why asking AI more than you tell it gets better results.
- 22:05 – How 16 automations in the old CRM shrank down to one.
- 24:25 – The built-in “undo button” that made this Chris’s best migration ever.
- 25:21 – How to join the 30-Day Build Your Own Agent Challenge and set up your first agent.
Resources Mentioned
- Join the 30-Day Build Your Own Agent Challenge
- Subscribe to the Automation Bridge Newsletter
- Join the Automation Bridge Community
Transcript
Chris Davis 0:10
And we’re back with another episode of the All Systems Go podcast. And let me say this. You know what? I realize I’m a man of many tales, and I hope you all appreciate the context in which I share these tech tales from.
Chris Davis 0:30
Because this podcast in totality is not theoretical, it’s not aspirational. These are practical learnings that are taking place live, and I’m capturing them, sharing them, and I’m just hoping that this is fruitful and beneficial for you and to your journey as both a builder and a founder. And this one, and just so you know, because these are so real, I will. The names have always been changed to protect the innocent.
Chris Davis 1:03
I will never put a client, friend, community member, anybody, on blast. And it’s funny, a lot of times I can record about someone and they can listen to the podcast and they’ll be like, man, that was such a good podcast. And in the back of my mind I’m like, yeah, it was about you. Right.
Chris Davis 1:17
But that’s my way of protecting, you know, don’t want to ever be messy and gossipy and, you know, all of that. But I also want to be fair and share. I mean, that’s a big part of my community is sharing, so that we can shorten. Sharing shortens the learning curve and helps you get there faster.
Chris Davis 1:38
So with that being said, there’s, there’s, we, we’ve, we’ve talked a lot about automation, AI. I actually can’t wait for you to listen to the next episode because I already know what it’s going to be where we tackle agents. But I’ve, I’ve, I’ve talked about how in just 10 minutes a day, you can learn AI. And the whole premise is use the time in your day instead of taking time out of your day to learn it.
Chris Davis 2:03
Right. And by now you should have seen either in the beginning of this podcast or somewhere else in marketing and announcing the 30 day build your own agent challenge, your own coding agent challenge, which will allow you to build, build an agent that builds other agents and deploys other agents on your behalf. So all of that to say this is one of the results of said practices. Right?
Chris Davis 2:35
This is what happens when you are actively building and learning how to prompt and build software in a way that’s effective. You’re learning how to deploy agents effectively. And that’s why I call myself a player’s coach. Yeah, because a player’s coach is a coach who played, they’re on the field and, and if you, if you catch them in the right practice, they may be playing with their players in practice.
Chris Davis 3:05
That’s me. I’m, I’m a consistent practitioner to learn so that I can then help you all learn faster or not fall at all. Right. That’s the goal.
Chris Davis 3:16
And I, I see this more, more prep. I didn’t realize just how valuable this is. And you’ll see my push for any and everybody to learn how to do AI and use AI for coding is not to. Yes, I want you to make money, but my hopes are for you to learn a skill that makes your life easier.
Chris Davis 3:44
This is why my community is of no cost because I just want to share and equip you to, to do what you know to do better. Right. And this is a big piece, understanding agents, how to use AI effectively. Because I’m seeing the whole is really.
Chris Davis 4:01
People don’t know how to build and train agents correctly to get a consistent output or to get the idea output. And it’s defined, it’s, it’s highlighted today in both the migration from hell and heaven. Right. And I’ll, I’ll start out with the, the migration from hell.
Chris Davis 4:19
What, what ended up happening is I was a client and we were tasked to migrate LMS platforms. Now, Learning management system is what an LMS is. And if anybody has ever dealt with one, you know that there’s no two that are made the same. One may have, may track progress in lessons and topics.
Chris Davis 4:45
Another one may just have lessons, they may have sections. Another one. I mean, it is. They are not compatible in any way.
Chris Davis 4:53
Well, the, the company, the company that we hired was. No, no, no, let me be explicit. It wasn’t the company that we hired. The, the client wanted to go in a new platform and the platform offered migration.
Chris Davis 5:07
And the point of the migration was to get all of the content, of course, out of the existing platform and put it into the new one. And they made all these promises. I have to give a shout out to my developer. He did flag many times.
Chris Davis 5:26
I don’t think it’s going to work as smoothly as they say. Xyz. And I’m thinking in the meetings that we had leading up to, they’re talking about, oh, we’ll use Claw to export and this, this and that. So I said, okay, they’re pretty savvy, so this should go well.
Chris Davis 5:40
That was my thinking and maybe it was wishful thinking only to find that. So let me tee this up. You have a contract with your current provider that you’re saying, hey, we no longer need your services. We’re moving on.
Chris Davis 5:56
And the current provider says, okay, you got 60 days. Everything cuts off on this date. Well, that date is approaching. And the client starts to realize, wait a minute, this is not going well.
Chris Davis 6:07
There’s courses that are missing, elements, audios, videos, whatever the case may be. So I started to look at it with, with my developer. We called a meeting and I’m looking and I’m. This is where the value of understanding how to engage with agents and understanding the output of automation, I mean, of agent, a automated workflow deployed by an agent and uses an agent within the workflow.
Chris Davis 6:42
But just understanding AI and how agentic workflows work, I was able to quickly see, wait a minute, these people. I think it was Abraham Lincoln. I used this quote, if it wasn’t Abraham Lincoln, he got the attribution. But I said, hey, the quote goes, if I’m going to chop down a Tree, I spend 80% of my time sharpening my axe, 20% of the time chopping down a tree.
Chris Davis 7:06
And I told them, I said, it looks like they flipped it. They spent 80% of the time moving, trying to move the data and 20% of the time understanding the actual data structure. And here’s what I mean when I come into a project, what I’m looking for is, first off, we know the definition of automation is moving data between one platform to another in its most simple form. I feel like I say that, and then every time I come up with another definition that try to make it even, even more simple.
Chris Davis 7:38
But for the most part, it is moving data from one platform to the next. Right. In a way that does not require humans. And in that sense, you have to understand the data structure of your current platform and then the data structure of, of the destination, the new platform.
Chris Davis 8:01
And this is where I would spend most of my time is understanding those structures. Here’s the benefit. I don’t need to understand the structures. I can instruct my agent to go and say, how is this what it.
Chris Davis 8:14
Define this database structure. Okay, go define this database structure. Where are they similar and where do they differ? Because once I know that, now I can train my agent to translate and say, okay, I need you to help me map how certain content in platform A will look in platform B.
Chris Davis 8:39
And once you do that, I’ll get to the output of that later. Okay, that’s actually the migration from heaven part. I jumped ahead because I’m excited to show you the good part. But the bad part was I started to See that, wait a minute, this person, this company, whomever did it, started the migration and then they got their fingerprints all over it because it was, it was inconsistent in what was missing.
Chris Davis 9:07
And I don’t know where you are in your AI journey, but AI is very consistent with how it’s wrong. It’s not like it’ll be wrong in one area and right in another. Usually when it’s wrong, it’s wrong all the way and all the way through. So when I saw that, I saw, oh, wait a minute, that wasn’t AI, that, that was a human, that was a human being that got in the middle of this thing and assisted AI.
Chris Davis 9:32
And by the time it got to the, to the destination, it was not, it was an idea, it wasn’t what you wanted to see, it wasn’t what the client needed. And this is what I call the, the messy middle. When you have a fully automated and automated and agentic workflow, there is no messy middle because the, the agent is handling end to end. But when you don’t understand how to fully deploy it the right way, you just maybe know, oh, I can hook up Claude to ask this and, and I’ll have Claude export a spreadsheet.
Chris Davis 10:06
The second somebody says, I will have Claude export a spreadsheet, I can tell their level of acumen with coding, with AI, because there’s no need for you to export a spreadsheet. Essentially, you’re, you’re choosing to use your human eyes to look at a database structure when you could just tell AI, hey, I’m looking for these trends. Hey, put it in a way where I can click some buttons and say, keep this and delete that. You haven’t built your relationship with the robots enough to know that’s even possible.
Chris Davis 10:34
So we get this hybrid approach where you’re asking AI to assist you to do outdated processes. And that’s what happened. So when we were approaching the deadline that this thing needs to be finished, not just because there’s a launch, an upcoming launch, but also because the subscription is about to cancel and there’s hundreds and thousands of users in these courses. So it would be beneficial, to say the least, for these courses to be able to be accessed in the new platform, right?
Chris Davis 11:18
And when I, again, when I saw that it, you know, there are times in life where you’re doing a thing, you’re learning a thing, let me say that. And you don’t realize the value of your learning until later. And it was right at that moment where sometimes I, you know, I’ll tinker around and I’m always, you know, building apps and trying to simplify the process, do a lot of teaching, explaining these things to people. And there are some times where I’m just like, it wasn’t worth it.
Chris Davis 11:45
You’re just wasting time. Are you distracted? Are you this? Are you that?
Chris Davis 11:48
But this was a highlight. I said, you know what, I wouldn’t have been able to look at the output that they gave and understood exactly what happened had I not been experienced with building, with AI. And again, that’s another reason why I’m here on this particular episode, highlighting the importance of you to do the same. I’m, I promise you, the only thing I gain from you learning how to leverage AI more effectively, specifically building agents that do tasks for you more effectively.
Chris Davis 12:24
The thing that I get the most out of that is you get better. And I know that’s kind of hard to understand because everybody has their guard up, like, oh, why do you want my information? What are you telling me this for? What’s the hidden agenda?
Chris Davis 12:38
It’s pure, I promise you. And anybody in my community can vouch for me. We share freely. The goal is learn it, do it, learn it, do it, learn it, learn it, do it.
Chris Davis 12:47
Because in that process you’ll master it. Right? So what happened? What was the problem here?
Chris Davis 12:54
The problem was one wasted time, waste of trust and a damaged relationship. Because company you told me you’d be able to do this. Since you didn’t do it, I can’t trust that you can fix it, what you couldn’t do. Now I’ve got to get developers out of the clients books to fix.
Chris Davis 13:14
Now I’m paying double, I’m paying double for their trouble and the client potentially will miss their launch and it will not go as anticipated. Terrible experience. That’s why I said migration from hell. Everybody knows because platforms are formatted so differently.
Chris Davis 13:33
That’s what makes migration so difficult. If there was a standard like every learning management system has to have this format. It’s easy, right? But when it comes to training the agent, this is the step that they skipped.
Chris Davis 13:49
What I call it, I call it end to end because I want to train an agent to be able to go from one end to the next. So that means understand the source, understand the destination and write the translation language between. You have to think about it. In languages, everybody, if someone speaks French and another one speaks English, there’s a translator in the middle.
Chris Davis 14:11
If they’re going to talk correctly to one another, that’s what we’re using an agent to do. It’s just technical language. Think of Platform A. API speaks one language.
Chris Davis 14:22
Platform B’s API speaks, speaks another language. Now you understand those languages. Agent now translate so that they can both communicate to one another effectively on my behalf. That is the most powerful use of AI in the automation era, right?
Chris Davis 14:40
To create these automated workflows that work for you and not against you. So what, what I would have done is I would have went and really studied our current setup. We’re using a SaaS platform. Hey, what is available through the API?
Chris Davis 14:55
That’s one. I need to know all the capabilities. How is the data formatted? What are the tables named?
Chris Davis 15:00
Not me, everyone. I’m deploying AI to learn this and tell me about it, right? The next thing I need to know is you, hey, client, you said there’s videos, audios, pictures and all of that. Where are they hosted?
Chris Davis 15:16
Because I need to give AI access to that as well. Because one of the things that we saw was that maybe someone named something in the course and it didn’t match the name of the actual file. So a human is looking for a match of a name, but guess what? A robot, an agent, is looking at hash numbers.
Chris Davis 15:38
They’re looking at the ID of the thing to say, oh, this course ID 256 has this video ID 8161, blah, blah, blah. So it’s matching the digital footprint. Now, there is no confusion. There is no, oops, I looked at this and mixed it up.
Chris Davis 15:55
You remove all of that, but you have to know how to give the agent access to perhaps your video hosting, your, your, your media hosting. If it’s aws, Cloudflare, whatever. You have to give it access there so it has everything that it needs to complete the job. Everyone.
Chris Davis 16:13
You’ve got to equip it with everything that it needs to complete the job. Because what you don’t equip it to do, you will do. That’s how it works. And I know you don’t want to do it, because if you wanted to do it, you wouldn’t have deployed the agent to.
Chris Davis 16:30
So remember, whatever you don’t equip the agent to do, you’re going to end up doing. And you’re doing is a degradation, is degraded, okay? It is a, a shadow. It is a cheap shadow of what AI can produce, right?
Chris Davis 16:54
So you don’t want to. And when I say do, you do not want to be doing title matching and copying and pasting content. Come on. In reality, we should have never been doing that stuff because humans are the worst.
Chris Davis 17:11
We are the Worst with copying and pasting, everybody, okay? Because I feel like there’s a little sensitivity in the air right now. For whatever reason, let me say this. I am the worst with copying and pasting.
Chris Davis 17:24
I can’t tell you how many times I’ll copy and paste. Maybe forget a period. Sometimes I copy and hit the paste button twice. I don’t know.
Chris Davis 17:31
I don’t know what I do at times. Just moving too fast, right? But you remove all of that, right? So you have to take the time to educate.
Chris Davis 17:42
You have to take. Let me, Let me not say educate, because that sounds like what you’re doing is you’re teaching the agent. No, you’re setting the agent up where it can teach itself. So when I say education, it’s self education.
Chris Davis 17:56
You. You need to put the agent in position to self educate itself, which means whether it be MCP or API, you need to give it access to the proper tools and privileges to read. Right? Once it knows the source and knows the format.
Chris Davis 18:13
Now it’s just a matter of saying, hey, here’s the destination. What matches and what doesn’t. Another thing that, that I, that I teach in my challenge is asking AI more than you tell it. I think humans get in the way of artificial intelligence far too much.
Chris Davis 18:34
When we try to be smart and then instruct when. When we should inquire and let AI instruct, right? We. You just have to flip that, right?
Chris Davis 18:47
So it’s. Instead of saying, hey, now, migrate all of the pictures over and do and make them look like this. Instead of saying that you. I wouldn’t instruct it.
Chris Davis 18:57
I would inquire and let it instruct me. And then I correct this instruction if it’s not right. Because it always comes up with things that maybe you didn’t think of and you’re like, oh, wow, that’s a really good idea. So I would instead say, based on the format of the destination, what do you recommend as matches and mismatches and what risks should I communicate to the client of things that may not be compatible?
Chris Davis 19:27
I would say that and let it do all of the thinking and then present something to me for me to react to. Say, okay, this is right, this is wrong. Oh, that was really good. Didn’t think about that, right?
Chris Davis 19:38
This is how I engage with AI. Everybody. There is a way. I know it’s the wild, wild west out here, but there is a way to do this in a way that makes where you get smarter and you can leverage its intelligence more effectively.
Chris Davis 19:57
All right, now, here’s what here, let’s Talk about the heaven piece. Because I, I had this same exact need for myself though. I was migrating from one CRM to the next CRM. Now mind you, the order, the sequence of these events is I migrated my CRM software first and then my client had this need.
Chris Davis 20:18
Which is why I knew what I knew because first off I thought it was, I thought it was going to be very complicated and I was surprisingly, I was pleasantly surprised how non complicated it was and how much was able to be achieved. So what did I do? I did exactly what I’m telling you all to do. I went and I said, hey, look, I want to go from this CRM to this CRM.
Chris Davis 20:43
Tell me how the data format is in this in the source and tell me what the data format is in the destination, what matches, what doesn’t, and come up with the plan of migration. So it comes up with this plan, right? Then I say, hey, before you do anything, let me give you context. This is the whole reason why this platform exists.
Chris Davis 21:04
This is why the fields are structured that way. Here’s some of the things that you wouldn’t know just by looking at the data set. I gave it that. And lastly I said, hey, here’s what I’m trying to achieve.
Chris Davis 21:17
So take what I’m trying to achieve as top priority. When you come up with your plan and ignore anything that I’ve said that’s prescriptive that may go against this plan being fully functional in so many words. That was the prompt, right? So what ends up happening is it does the work, it comes up with the plan, it tells me everything and it gives me use cases.
Chris Davis 21:40
I love clog because I can build these artifacts that are interactive and it just puts all the use cases in an artifact and I just react and say yes, that’s right. No, that’s wrong. Yes, yes, yes. And then tell Cloud, okay, I put my answers in the artifact.
Chris Davis 21:52
Proceed. Right? It’s kind of like I turn it into a project manager before it turns into a developer, right? And when I did this, not only did it highlight some things, but it rebuilt.
Chris Davis 22:05
Oh my goodness, everybody. Because the previous platform was limited in its automation capabilities, it I needed 16 plus up up to 20 automations to do what I was trying to do. Well, guess what? When I let AI figure it out and wasn’t prescriptive, the new, no exaggeration, the new platform needed three automations and the one function that was taking up 16 automations to complete was now condensed down to one.
Chris Davis 22:35
I lied to you not and it’s it’s what I, it’s a, it’s a term that I use. I created an external logic engine and my approach is in brevity. I take the logic that you would usually program by hand in a platform and I take that out of the platform because that logic is for my business. Then I put it in an app or plugin, if it’s WordPress and I execute the logic off site and turn my marketing automation platform, my email marketing platform to an endpoint of that automation.
Chris Davis 23:10
So it’s doing all the processing, all the stuff you would have had to build, it’s doing all of that and then it’s just responding in the actual app. But the app has to have those capabilities. So my new CRM did have that. So long story short, I’ve got the migration complete, it’s running on way less automations, it’s cleaner, it’s more consistent because it knew my use case and it was able to build to exactly what I was trying to do.
Chris Davis 23:38
And here was the last piece I built in a dashboard that every day it gave a report of because I did it over 30 days. Like I didn’t do it all at once over 30 days so I could watch it and improve it. The first few days I needed to make some adjustments. So every day it showed me a report of who moved over, why, what, what’s what, fields moved over, if any issues.
Chris Davis 24:01
So anytime I saw something, say hey, why did this look? Why, why wasn’t this moved over? It could look at the log and fix it itself. I didn’t have to figure out anything.
Chris Davis 24:11
So there was some monitoring and then after a while it was off to its races. No data loss everyone. This is the first time I’ve done any migration without a data loss. And clean, when I say clean.
Chris Davis 24:25
It even built a reverse path that in any case, if the new platform messes up, I can click a button and move all of those people back the next level. It is hands on, hands down, the greatest migration experience that I’ve ever had. Cost of time was low, the quality output was high and the level of frustration throughout was non existent. This is the promise of when you learn how to leverage AI, automation, data, workflow, process systems all in one.
Chris Davis 25:11
This is, this is, this can, this can be true for you. This can’t be true for you. This is not, you know, like pie in the sky. Oh Chris, you’re an expert.
Chris Davis 25:21
No, you can do this. You just need the right training, you just need the right hand holding and you just need the right structure. This is what I am committed to giving you all and is why I have a new challenge that you can find in the show notes. The link will be easy to find.
Chris Davis 25:39
You just click and go and you can get access to the challenge that will walk you through every day how to set up your first agent. Because here’s what I learned. When you know how to set up your first. When you get your first agent set up, it’s like a tattoo.
Chris Davis 25:54
It’s like Pringles. I think Pringles is the one where it said once you pop, you can’t stop. You know, I don’t have tattoos, but I’ve heard when you get one, you’re like, oh, let me get another, let me get another, let me get another. Right?
Chris Davis 26:05
That’s what it is. And instead of you saying, oh my gosh, I gotta. I’m gonna learn that you guys have been. I’m gonna be.
Chris Davis 26:14
You have been. I’m gonna learn that you’ve been saying that for the last six months, maybe even a year. You’re not going to set aside the time to learn it. Therefore, how about you use the time within your natural flow of the day to do it?
Chris Davis 26:30
So anyways that’s available. And, and, and of course, the community is the easiest starting point. You just join. Literally, there’s no there.
Chris Davis 26:44
I don’t, at this point, I honestly don’t know what you would be waiting for. And the reason I’m pushing this is the more people that join, the more people to share. The more people to share, the faster we can, we can move intelligently. And I’m, and I say this to say the invitation is for people who have experience doing a thing and what they want is to learn how to deploy an agent to do the thing they know to do.
Chris Davis 27:09
It’s not necessarily for somebody learning how to do something in their career. So if you’re just getting started in your career, this probably is not the, the community for you. But if you had a corporate career, you, you’ve been an entrepreneur for some years, you have proven processes, you have an approach to things and you want to take that approach and see how agents can improve it and how you can remove yourself without removing the quality. This is the community for you.
Chris Davis 27:37
So I hope to see you in there. For those of you who are joining from listening, thank you so much. I smile every time I get a new, a new addition and keep, keep, listen, if it’s not with me, please find somebody that can help you move the needle in this AI space. I, you, we cannot guarantee the cost is going to be what the cost is.
Chris Davis 28:03
We cannot guarantee the accessibility is going to be what the accessibility is. But I promise you this. The the framework and the in the route that I teach is is evergreen. I don’t care how AI changes, you’ll always be able to leverage it in this way.
Chris Davis 28:16
That’s why I’m excited about it. So hopefully something was said in this episode to help you take not just one or two, but several steps forward. And I thank you always for being a listener of the podcast. Whether this is your first one, your 200th, wherever you are, I value your time and appreciate you.
Chris Davis 28:36
And until next time I see you all online, automate responsibly, my friends.
Want to Be a Guest on the Podcast?
We’re currently accepting guests for the podcast that are SaaS owners, marketing automation consultants, and digital professionals that have produced high results with automation.
If that’s you, or you’d like to recommend someone, click here to apply to be a guest.
About the Show
Discover how to deploy automated marketing, sales, and onboarding systems to scale your business without working long hours to do so.
