The AI Automation Graveyard (Why Most Projects Die After Launch)

The AI Automation Graveyard (Why Most Projects Die After Launch)

If your automation hasn't been updated in 90 days, it's probably broken. Alane and Micah dive deep into why automation is actually software and what that means for your business. They reveal why most automation failures aren't due to poor building—they fail because people stop maintaining them. Learn why starting small beats building "maintenance monsters," how API changes and model updates break workflows faster than ever, and why the business case still makes sense even with ongoing maintenance costs. This episode will completely shift how you think about automation in your business.

 If your automation hasn't been updated in 90 days, it's probably broken. In this episode, Micah and Alane reveal why most automation and AI projects fail after launch, and it's not because they were built poorly.

Discover why automation requires ongoing maintenance just like software, how API changes and platform updates can break your workflows overnight, and why the "build it and forget it" mentality creates what we call a "maintenance monster."

In this episode, you'll learn:

• Why automation is software (and what that means for your business)
• How fast tech changes are breaking automations every 90 days
• The real cost of ignoring updates and maintenance
• Why small automation projects beat massive workflows every time
• How to avoid creating a maintenance monster in your business
• The business case for automation vs. manual processes
• Real examples of automation failures (like creating 150 proposals by accident) 

Whether you're using n8n, Make, Zapier, or building AI agents, this episode will change how you think about automation maintenance, updates, and long-term success.

Topics covered: automation maintenance, AI agents, workflow automation, API updates, software development, business process automation, no-code tools, Make.com, n8n, Zapier, automation strategy

Disclosure: Some links are affiliate links, meaning we may earn a commission at no extra cost to you. Thanks for supporting the podcast!

For more information, visit our website at biggestgoal.ai.

Upcoming Cohorts & Workshops:

• Build AI Agents + RAG
• Build Scalable Systems
• Leadership Team AI Workshop

Enjoy our Free Tools:

• Free Asana, ClickUp, or Monday.com Selector Tool
• Get 25 Free Custom Automation Ideas for your Business
• Process Mapping course

Connect with Us:

• Join our Community
• Follow us on LinkedIn
• Subscribe to our YouTube channel

Transcript

Micah Johnson (00:01)
If your automation hasn't been updated in the last 90 days, it's probably out of date. And that's how fast tech is moving these days. So in today's episode, we're talking about how automation is actually software and what that means when you start using it.

Alane Boyd (00:19)
A lot of times automation is not failing because it was poorly put together. It really is because people stopped maintaining it.

Micah Johnson (00:27)
Yeah, they just ignore we've seen this across dozens of companies. They have this grand idea, building automation, let's go do this and we'll use AI, but it's the same thing every time. Like, isn't this a one-time build? Why do we ever need to touch this again? Shouldn't it work forever?

Alane Boyd (00:48)
It is hard to understand when this isn't your life every day, and I understand that. But if you even just think about your phone, whether you have Android or an iPhone, you are constantly making updates. The software itself make updates, but then you have to have your apps also make updates to keep up with the updates that the iPhone made.

Micah Johnson (01:12)
Yeah, but that's happening behind the scenes, right? Like as a normal phone user, I just go update and like five minutes later, everything's magically updated. But what's happening behind the scenes is months ahead of time, those developers of all the apps are getting updates from Apple and Google saying, hey, here's the new OS.

Alane Boyd (01:19)
Bloop!

Micah Johnson (01:37)
start testing your app, start upgrading so that when it gets pushed to all the users, your app's going to keep working. And if the owners of those apps treated that like some of the people treat automation and AI workflows, what they would do is pay a lot of money to get it developed and then fire that development team. And when Google and Apple goes to them and says, hey, we've got a new version coming out. You should update your stuff. They go, hmm.

I'm gonna pretend I didn't hear that.

Alane Boyd (02:08)
No big deal, we're just gonna keep things as is.

Micah Johnson (02:11)
Yeah, it should work forever, right?

Alane Boyd (02:13)
It does not. And there's so many things that creep up whenever you don't do this. One of the biggest is that it doesn't work, but if you're not updating software, that's where security risks start to pop in as well, because you're not keeping it up to date with the latest software updates. And every single software has updates and it can be an API changes. It could be their endpoints change. It can be a whole bunch of different things.

A lot of times it's for the best. You want to use the latest, smartest AI model, you know, for some of these things. Like it is a good thing to make these updates and you want your automations to stay strong, stay working. also is just because the models change and that it needs to be a healthy connection point that you have.

Micah Johnson (03:00)
it is really wild when you think about it, like how much actually can change. So when you build automations, you're connecting via APIs, you've got webhooks, you now add more modern MCP servers, you've got AI Agents, which mean you have AI models, you have tools like n8n or Zapier or Make, you have the platforms that you're connecting to.

and the features that they're adding. So every single one of these things that I just listed could all be updated. And with AI powering code and development teams, that stuff's happening so fast these days. New features are getting added. New models are being created for AI. New API features are being added. New API endpoints, new webhooks, new.

everything

Alane Boyd (03:57)
it's not even just You know, I didn't want to confuse things that we're just talking about AI Agents that need automations I mean, we just Make connection with Pipedrive. they just made this huge update to it. it didn't have anything to do with an AI Agent It didn't have anything to do with AI.

Micah Johnson (04:11)
yeah.

Alane Boyd (04:16)
It had to do with the connection point between Make and Pipedrive. They made a huge update. Anybody using the old connection, those automations weren't going to work anymore.

Micah Johnson (04:26)
Well, and there were even issues when they updated it. So if you don't have somebody that is helping you maintain this and understand what is going on, for example, with that update, the Pipedrive webhook just kept sending and sending and sending.

Alane Boyd (04:29)
Mm-hmm.

We know from experience, we made the update ahead of time. We knew it was coming because they do communicate these updates. So we made sure our team went in and updated the modules in Make to make sure our connections all worked. But we had our proposal generator created 150 proposals for one person. And we followed the update. What's that?

Micah Johnson (04:46)
Yes.

That's how we make the big bucks, Alane.

That's how we make the big bucks. 150 proposals.

Alane Boyd (05:08)
For ours, yeah, we make 150 proposals. Once you sign all these new clients,

then we'll get started. But that's even us doing a healthy maintenance. We kept up to date, but they had bugs in their update, which happens. So it really is this maintenance that you're ongoing and watching for. Even if you are keeping up, those updates aren't perfect. Just like when you update your iPhone and it is buggy and they come up with a new release.

Micah Johnson (05:21)
Yes.

Alane Boyd (05:35)
a couple weeks later and you have to do another update.

Micah Johnson (05:38)
I mean, even your browser, you update your browser and then like two weeks later, they're like, ⁓ security update. We're gonna now, you know, click to refresh and update your browser again.

Alane Boyd (05:48)
I mean, if we

could just stop having to do that, I think everybody would be happy.

Micah Johnson (05:52)
I

think if we never had to update anything again in software, we'd all be happier. But to reiterate, that's why you need to think about automation and AI agents as software. If you frame it that way in your head, you're thinking through this the right way, right? Like you're, you're going, okay, I would not pay somebody or train somebody to build me software and then fire them and never update it. We all know at that point, like,

Alane Boyd (05:54)
Yeah, that's true.

Micah Johnson (06:20)
We need to update stuff. you know, that's, it's the same as like getting a phone system and then never upgrading the hardware or getting databases or servers and never upgrading it. So if you still have a server from 99, you're running on antique stuff and when you say it like that, that doesn't make any sense yet, people still think today let's automate some stuff.

I automated this in 99. 10 years from now, this shit should be running perfectly, right?

Alane Boyd (06:51)
For anybody that is thinking that at this point, go back to your phone that you had in 99 and see how that's gonna work for you. So, Micah, there's all different types of updates too. Some of them are really easy and could be done by almost anybody that has access to Make or n8n or whoever has access and you can go in and you just update it.

Micah Johnson (07:12)
I would say it still takes a little bit of, like, you've got to feel comfortable going into n8n. You've got to know what a node is. Like, there are these little things, and if it's built correctly, it gets even easier to update.

Alane Boyd (07:19)
right.

True, I mean, you do have to have some basics to know how to log in, find that workflow, find the node, things like that. And some of them are bigger updates. Like just like I mentioned, PipeDrive and Make, that actually was a bigger lift.

Micah Johnson (07:43)
Yeah, yeah, that's true. You got to go in and you're changing nodes and you're possibly changing the architecture, rebuilding where the from and goes to. But some of the other bigger areas where you'd still want somebody to help you with are like, if you have like old Zaps and you have the idea that maybe AI Agents can help me with this.

That's another one where it's a lot harder I can't just update it by clicking a button and getting my AI Agent workflow in there. All of this can be learned. That's what's amazing to me today. Alane is like, it's not like development from so many years ago where I'm going to learn assembly machine code, use three letter

Alane Boyd (08:32)
Mm-hmm.

Micah Johnson (08:33)
abbreviations for commands to make the computer do what I want it to do. You're literally connecting stuff by dragging and dropping. And then even it's like visually programming stuff. So if you have the interest in building your own, the technology is there and I would even say accessibility is there for almost anybody.

Alane Boyd (08:38)
Yeah.

Micah Johnson (08:57)
to be able to learn this and go from idea to execution. And that's going to help the maintenance and keeping this up to date as well.

Alane Boyd (09:05)
Yeah, we even last year started training people outside of our company on how to build agents and in n8n but also maintain them. Like that's part of it is learning and being able to do some debugging and understanding those things. And then we've launched some help support for afterwards so that if you wanted to do it in house, you still can. And you're going to have the basic knowledge and then have us to, "Hey,"

"I'm having a little bit of trouble with this" because it is something that's teachable without having to be a full on developer.

Micah Johnson (09:38)
Yeah. Those are some of my favorite things that we do right now, Alane is actually help when, we get like a CEO or a COO and they're like, I've got all these ideas. I want to build this cool stuff. And then we watch them build it after some training and instruction and working together. And they're like, I got this working. Like I seriously get a dopamine burst for them.

Alane Boyd (09:53)
Mm-hmm.

Mm-hmm.

sharing in the happiness, Micah.

Micah Johnson (10:02)
I'm sharing in the happiness.

Alane Boyd (10:04)
I think the biggest thing too that we really, there's two things that we are big fans of. One, no code or little low code. And then small projects, not massive workflows because that creates maintenance monsters.

Micah Johnson (10:21)
Yes, that's my new favorite term, at least today, is this concept of like, start small or you're gonna get a maintenance monster. Nobody wants a maintenance monster. I'm just gonna keep saying it because it's so much fun to say.

Alane Boyd (10:24)
And I said it first. ⁓

Yeah, starting small.

But it's also the time to execute. mean, all of this that we talk about absolutely all like works together. But when you try to do a massive automation to solve 100 problems in your business, not only is it going to take months and months to actually execute and get the output that you really want to maintain that, the more connection points that you have,

the more maintenance you're gonna have.

Micah Johnson (11:04)
Yeah, more points of failure.

Alane Boyd (11:06)
Yeah.

Micah Johnson (11:07)
And I mean, like what you're talking about, Alane, and again, when we say automation is software, we're now thinking like developers a little bit. But we have capabilities that allow us to say, let's only introduce and launch this one little bit of it, and then this one little bit of it, and then this one little bit of it. And it used to be when we're building software, we have to do the big thing. And it does create a maintenance monster. But.

Now it's like, what if we just had an agent that did this? Cool. Solve 2% of our problem. Then tomorrow let's build an agent that does that and solve 3% of our problem. Well, in two days we've solved 5% of our problem and the business impacts already like something that we can feel tangible, see, and we're building like a whole squad of modular solutions that work together and get impact together.

Alane Boyd (11:43)
Mm-hmm.

Micah Johnson (12:04)
Start small, avoid the maintenance monster. I just needed another reason to say that.

Alane Boyd (12:08)
All right, Micah.

Micah Johnson (12:09)
So I think a good way to kind of wrap up this episode, Alane, is thinking about just the cost of this versus doing it manually. You know, there's business cases behind all of this. But for the most part, it is still less expensive to build automation, even if it's small things and I guess I'm going to touch on this for a second because

we hear a lot like, that's not going to move the needle. Well, five small things that you can build in five days will move the needle a hell of a lot more than one giant thing that you can build in 12 months. The needle is not moving in 12 months. And in fact, you have a lot of costs that you're incurring in 12 months. And you've created a maintenance monster. So if you think about

I'm going to create a bunch of small things. I'm going to implement and launch those things. It's less time to build. It's less cost to build. It's less cost to maintain. You still have to budget to maintain those. And the business case gets so much easier when you look at that compared to, I'm going have somebody do this or multiple people do this manually over and over and over again, especially with training and turnover and management costs and benefits and just salary.

or hourly, or however you're paying them with soup and noodles. I don't know. But the cost difference is still huge. I guess the moral of the story, maybe on this one, is do think about it like software. Do realize that there are going to be ongoing maintenance costs, because it is software. And still, build the business case around, does it make sense to

Alane Boyd (13:31)
Hmm

Micah Johnson (13:55)
implement this in small bursts, shorter terms, easier to maintain versus continually hiring and paying people to do this manually.

Alane Boyd (14:04)
And the manual piece, I just think of like the excitement that we all in our team get or seeing the excitement from clients faces when they have an automation that works. It's like, I don't have to do that tedious shit anymore. That's the stuff that brings you down. And like now you can press a button and a proposal gets created. Like these things make true happiness in people's jobs.

that they are not wasting hours to get to the output. So it's hours in salary that you're saving that makes it worth it, but also people's happiness in their position, not having to do data movement.

Micah Johnson (14:44)
Yes, yes, that busy work, that copy and paste, ⁓ man.

Alane Boyd (14:49)
Pretty much everything is copy and paste. You're moving things from one place to another. And I mean, even the proposal, you're moving it from a transcript, from your head, into a place to get a proposal out. That is data movement. That is moving data from one place into another.

Micah Johnson (15:06)
You know what that sounds like, Alane? A manual maintenance monster.

Alane Boyd (15:10)
I'm so glad you got to say that so many times.