别再像使用聊天机器人一样使用AI——与Vern Crawford

好的，各位。

Alright, everyone.

欢迎回到商业版播客。

Welcome back to the Business Edition podcast.

这是与弗恩·克劳福德那期节目的第二部分。

This is part two of the episode with Vern Crawford.

如果你还没听过，请先回去听第一部分。

If you haven't already, please go back and listen to part one.

我们将继续讨论如何真正有意义地、深思熟虑地、有效地在你的业务中部署人工智能。

We are going to continue talking about how to really meaningfully deploy AI in thoughtful ways and useful ways in your business.

我相信你们会从弗恩的分享中收获很多。

I think you guys are gonna get a lot out of hearing, everything Vern has to to say.

我们开始吧。

We'll get into it.

欢迎收听HVAC万事通播客。

Welcome to the HVAC Know It All podcast.

在加拿大多伦多某个地下室录制的。

Recorded from a basement somewhere in Toronto, Canada.

你的主持人兼暖通技术员加里·麦克里迪，将带你深入探讨行业，聊遍暖通的一切。

Your host and HVAC tech, Gary McCreadie, will take you on a deep dive into the industry discussing all things HVAC.

从讲故事到技术讨论，尽情享受节目吧。

From storytelling to technical discussion, enjoy the show.

现在可能就说得通了。

Maybe it makes sense now.

我知道我们在开始录制之前，Vern，已经稍微聊过这个。

I know we talked about this a little bit before we hit record, Vern.

我很想看看你给我们展示一下，比如，我想你在Make上做过一些，比如，一个这样的自动化示例，因为我觉得这能帮大家弄清楚，第一，API的概念，以及两个系统如何互相通信。

I'm interested to see you show us, like, I think you work some on Make, like, an example of one of these automations because I think it will help make clear for people, one, just the idea of an API and two systems talking to one another.

还有，当我们使用任何大语言模型时，什么时候需要这个MCP桥接？

And then when do you need this MCP bridge when we're working with any of the LMs as well?

能不能请你调出一个例子，然后大致讲讲发生了什么，以及你如何用这个工具来构建这类自动化。

Can we ask you to pull up an example and sort of talk through even in a high level what's happening, how you use this tool to build these kinds of build these kinds of automations.

这个窗口是在make.com里，你在make.com里建的每个自动化，他们管它叫场景。

So this window here is in make.com, and each automation that you build in make.com, they call it a scenario.

这个场景就是……

So this scenario is...

想象一下，如果你在听的话，想象一个流程图，但这个流程图有大概三四十个不同的气泡或者对象图标。

If you imagine, for those listening, if you imagine a flowchart, but this flowchart has, like, 30 or 40 different bubbles or objects Icons.

在上面。

On it.

图标。

Icons.

谢谢。

Thank you.

这个特定的场景做的是，它会查看一家管道公司收到的邮件，只关注那些来自家庭保修提供商的新工作单。

And what this particular scenario does or did is it would look at the emails coming into a plumbing company and it would only pay attention to the ones that were new work orders from a home warranty provider.

然后它会一步步地处理一系列不同的问题，把那个工作单录入Housecall Pro，为这家公司准备好一切。

And then it would walk through step by step a bunch of different questions to get that work order into Housecall Pro and set up and ready to go for for the business.

我赶紧说一下这个画面，因为我知道有人在开车。

I just want to talk to the visual quickly because I know people are driving.

我们现在看到的是一个满是图标的屏幕，我想问个问题，因为我自己也想搞清楚。

So what we're looking at is a screen of a bunch of little icons and I want to ask a question because I want to understand this myself.

我们是不是在看……

Are we looking...

所以这些图标从左到右排列，大概有二十个左右。

So there's icons going from left to right, maybe about 20 of them.

然后从左到右的过程中，它会分叉到不同的图标。

And then as we go left to right, it branches off to different icons.

那我就猜一下这个是怎么运作的。

So I'm just going to make a guess on how this works.

你告诉我我是不是完全搞错了，还是说对了。

So you tell me if I'm out to lunch or if I'm right.

我们是不是从左到右一直走到最后？

Are we going from left to right all the way to the end?

然后每个分支都是我们需要决定的不同选项，决定完之后再回到主线上，一直走到最后。

Then each sort of branch is something different that we have to decide and we come back to the main line to get to the end.

是不是差不多这样运作的？

Is it working similar to that?

你说得完全对。

You're absolutely right.

没错，就是这样。

So that's correct.

这里有好几种不同的颜色，每种颜色代表 make.com 里的一个不同工具。

So there's a couple different colors and each color represents a different tool in make.com.

然后那个……

And so the...

当你横向看这些不同的节点时，图上的绿色圆圈就是决策点，它会根据流程中前面步骤得出的值来判断：这个条件成立吗？

As you're looking across at these different points, the green circles on this diagram are the decision points, where it's gonna look at the the values that it's determined from the previous steps in the process and say, is this true?

这个条件成立吗？

Is this true?

如果这两个都不成立，我们就走第三条路。

Or if neither of those are true, we'll take the third route.

所以这里发生的是，这个工作流在问几个关键问题。

And so what's happening here is this workflow is asking a few key questions.

当我打开这封邮件时，这个客户是否已经存在于Housecall Pro账户中？

When I open this email, does this customer already exist in the Housecall Pro account?

如果客户不存在，我们就创建一个新的。

If the customer doesn't exist, we'll make a new one.

如果客户存在，那我们就得检查地址是否存在，如果地址存在，我们就可以直接添加工作，准备就绪，等等。

If the customer does exist, now we have to check if the address exists, and if the address exists, we can just add add the job, we're ready to go, etcetera.

所以这……

So it's...

这就是为什么它有点长，而且有不同的分支，因为它涵盖了当这些工作订单到达邮箱时可能发生的所有不同情况。

That's why it's kind of long and that's why there's different branches is because it's covering all of the different possible scenarios that could happen when one of these work orders hits the email.

有意思。

Interesting.

好的。

Okay.

所以每一个绿色的图标都有一个……

So each one of those green icons has a...

我看不出来它是紫色还是黑色之类的。

I can't tell if it's like purple or black or something like that.

每一个绿色的图标都有一个比较暗的图标，它们会从……移出来。

Each one of the green ones has one of these darker icons that sort of move out of...

那些是什么？那些符号是什么？

What are those what are those symbols?

那些深紫色的图标其实是Slack消息。

The dark kind of deep purple icons are actually Slack messages.

我把这个自动化设置得很容易调试，在每个决策点，如果有什么不对劲，或者我想弄清楚系统在做什么，每个决策点都会给我发一条Slack消息，告诉我上一步发生了什么。

So I set this automation up to be very easily debugged where at every decision point, if something is is not right or if I'm trying to figure out what the system is doing, each of those will just send me a Slack message with what happened in the previous step.

哦，好的。

Oh, okay.

明白了。

Gotcha.

我把那些关掉了，因为太多了。

I turned those off because it was too much.

Slack消息太多了，简直没完没了。

It was way too many Slack versus shits.

不过，确实对调试很有帮助。

But, yeah, it was very helpful in debugging.

Vern，我很好奇。

Vern, I'm curious.

我们现在看到的这个就是make.com，对吧。

So what we're looking at here is is make.com, as you said.

还有几个类似的，比如Zapier或者n8n。

There's a couple others like Zapier or n eight n.

这些是现成的工具，能帮你搭建自动化流程，当它们……

And these are off the shelf tools that help you build automations when they're...

来连接你的不同系统。

To connect your different systems.

通常这些系统已经内置在Make里了，比如Slack、你的邮箱或者Google Drive。

So usually, these systems are already included in Make, so something like Slack or your email or your Google Drive.

然后还有……

And then there are...

再往上一层，就是自动化或者真正的智能体构建，那就是自己写代码了。

Then there's the sort of the next level of automation or or actual agent building, is just called coding yourself.

当……

When...

人们应该怎么考虑……

How should someone think about...

这是不是就像一个入门级别，然后你再开始构建智能体？

Is is this like an initial, like, almost like beginner tier and then you then you start to to build agents?

或者人们应该怎么考虑什么时候用Make，什么时候用Cloud Code？

Or how should someone think about when when to use Make and when to use Cloud Code?

这个问题问得好。

That's a great question.

这个嘛……

It it...

在我看来，至少以我的学习方式来说，要想充分利用Claude桌面版或者ChatGPT桌面版的功能，我得先这样搭建起来，才能明白背后的原理。

In my opinion, I think, at least in my learning style, for me to be able to take full advantage of what Claude desktop or take full advantage of what, Chad GPT desktop can do, I needed to build it like this first so that I understood what was happening.

你……

You...

现在还需要这样做吗？

Do you have to do this nowadays?

当然不需要。

Definitely not.

你不需要再经历这些了。

You don't have to go through this anymore.

现在的智能体已经非常强大了。

The agents are incredibly capable nowadays.

我最近一直在玩Grok机器人，那是SpaceX AI几周前刚推出的。

I've been playing around with Grok bot recently that just came out from SpaceX AI a few weeks ago.

但如果你真的想深入了解，或者至少对数据的处理过程有个深刻的理解，我建议从这样的东西开始，尤其是那个超级简单的入门项目，就是连接webhook然后发一封邮件。

But if you really wanna get into the weeds or at least have a a deep understanding of what it is that's happening with the data, I would recommend starting with something like this, especially like that super simple one as a starting point of just connecting, listening to the the webhook and sending an email.

就是那种特别基础、特别简单的。

Like, it's so basic, so simple.

但通过亲手搭建这个项目，我觉得你会对它有更深的理解。

But going through the exercise of building that, I think, would build a deeper appreciation.

这样就不会只是想着，这个代理在干什么？

So it's not kind of like, what's the agent doing?

比如，出问题的时候，这里到底发生了什么？

Like, what's going on here when, when something goes wrong?

不过这只是我的学习方式。

But that's just my learning style.

所以Vern，如果我是一名技术人员。

So Vern, if I'm, I I'm, I'm a technician.

好的。

Okay.

我现在在现场，而且我开了自己的公司，所以我现在得站在一个技术人员的角度来说话。

I'm in the field and I started a business and so I have to speak from the mind of a technician right now.

如果我在听这个，我会想，我知道怎么拧扳手。

If I'm listening to this, I'm like, I know how to turn wrenches.

我知道怎么换压缩机。

I know how to change compressors.

但作为一个新老板，或者将来想当老板的人，我完全不知道该怎么搞这些东西。

I have no idea how to implement this kind of stuff as, as a new business owner or a wannabe business owner one day.

这是你们公司提供的服务，对吧？

This is a service that your company provides, right?

你们就是帮人做这些的。

You help with this stuff.

所以并不是说老板自己得琢磨怎么把这些东西都建起来。

So it's not like the business owner themselves is going to have to figure out how to build all this themselves.

你们可以帮忙做这个。

You can help with this.

这是你们业务的一部分吗？

Is that part of what you do in your business?

我觉得我们现在差不多到了那个阶段，每个人至少都……

I think I think we're almost at the point now where everyone is at least...

至少都试过和LLM聊天了，我想。

Has at least tried chatting with with an LLM by by now, I would think.

如果还没到，那也快了。

If we're not there, we're gonna get there pretty soon.

每个人都试过ChatGPT。

Everyone's tried Chad GPT.

每个人都试过Claude，至少为了看看它是什么。

Everyone's tried Claude, at least just to to see what it is.

但要达到能在日常业务运营中稳定获得一致结果的程度。

But to get to the point where you can reliably expect consistent results in your day to day operations in your business.

有两种方法可以处理这个问题。

There's two approaches to it.

比如说，你可以花时间去学习怎么做这些，然后自己搭建，那当然很好。

Like, you could take the time to learn how to do all of this and and build it yourself, and that's that's great.

但是，你得投入多少时间、精力和心血才能做到，相比之下，找像我这样已经经历过学习过程的人来帮你搭建，几天之内就能搞定，马上就能用。

But how much time and effort and energy are you gonna have to put in to do that versus, you know, having someone like me that's that's gone through that learning process come and build that for you, it's gonna be ready to rock and roll in a matter of days.

不过，是的，这确实……

But, yeah, that that's...

这是个很好的问题。

That is a great question.

你说每个人都有机会试试Claude之类的。

You said everybody's had a chance to try Claude or whatever.

我自己大概三天前才第一次用Claude。

I just used Claude for the first time, like three days ago, like myself.

我之前听过别人和它对话的录音，但我是开车的时候测试它的。

I've heard other conversations that other people have had with it, but I was testing it as I drove.

我开了大概一个半小时的车，跟它聊了好多。

Like I had like a good hour and a half drive and I had so many conversations with it.

而且我发现它好像是英国口音。

And I actually found like it was a British accent.

不知道是不是默认的，但就是个男声，英国口音。

I don't know if that's the default, but it was a male British accent.

我听到的就是那个声音。

That's the voice I got.

不知道是不是默认的，也不知道能不能改。

I don't know if that's the default or not, or if you can change it.

我还没去看那些设置，但跟它聊了大概三四个跟我生活相关的话题。

Haven't even looked at that yet, but the conversations I had with it about probably about three or four different topics that pertain to my life.

到最后，我听着它的回答，一直在点头。

I was actually like at the end of the, his response was nodding my head.

它给你的反馈真的挺不错的。

Like, the feedback this guy gives you is actually quite, it's actually quite good.

这不是瞎扯。

It's not BS.

你确实能从Claude那里得到一些非常好的反馈。

You actually get some really good feedback from Claude.

我真的很惊讶，不过总之，我就是想提一下这个。

I was really surprised, but anyway, I just wanted to throw that in there.

那么，Furman，请继续。

So, Furman, go ahead.

你能让我接下来共享屏幕吗？

Can you actually let me share my screen next?

我想在这里调出一些东西。

I want to pull up something here.

我之前整理过这个。

I put this together at one point.

我当时在一个建筑性能……

I was at one of the building performance...

四月份在BPA会议上发言时，我就在思考AI以及AI在三个层面的应用。

The BPA conferences in April speaking and sort of think about AI and the use of AI in three levels.

第一个层面就是我们所说的聊天。

So the first level is what we call chat.

第二个层面是上下文，第三个层面是代码。

The second level is context, and the third level is code.

这是一个简单的理解方式，我相信还有其他的描述方法。

Simple shorthand for understanding, and I'm sure there's different ways to describe it.

第一个层面就像Gary你刚才那样，只是和它聊天，问它问题，有点像用谷歌搜索，再加上一些其他的用途。

Where the first level is, as Gary, you were just just chatting with it, asking it questions, using it a little bit like a Google search, plus some some other use cases.

第二个层面是上下文，你可以使用这些相同的工具，但利用一些内置的，我想Claude称之为连接器的东西，实际上就是内置的连接到你的系统的方式。

The second level is context where you can actually use these same tools, but using some built in, what I think Claude calls connectors, but are effectively built in ways to connect to your system.

比如连接到你的Gmail、谷歌日历、Dropbox，还有你的……

So connect to your your Gmail, to your Google Calendar, to your Dropbox, to your...

无论你有什么数据存储系统，都可以让你开始做一些事情，比如你刚才描述的邮件自动化，Vern，你可以说，嘿。

Whatever systems of storage of data you have allows you to start to do things like even that email, you know, that email kind of automation that you were describing, Vern, where you can say, hey.

帮我建这个，或者就是搜一下我的邮件，找出那些我没回复的客户之类的。

Help help me build this or just, like, search through my email and find find me customers that I haven't responded to or something like that.

你可以开始查询你的系统了。

You can start to query your systems.

然后这第三个层次就是我们说的……

And then this third level is what what we...

我称之为代码，就是真正去构建，通常利用这两个……

What I call code, like, really building often using the two...

AI代理或语言模型本身来写代码，然后帮你自动化某些任务。

The AI agent or the LMs themselves to to write code that then helps you automate certain certain tasks.

我很好奇。

I'm curious.

我觉得这第三个层次就是你真正需要帮助的时候。

And the way that I think about this is this third level is when you really need some help.

通常，大多数……

Usually, most...

除非你是那种有内部资源的大型企业，大多数听这个节目的小企业主说实话都没有时间或者资源去自己搞定这些，所以他们可以雇Vern。

Unless you're a large enterprise company that has internal resources, most small business owners who are listening to this are not gonna have the time, frankly, or the resources to to coat themselves, and that's why they can hire Vern.

在WorkHero，我们其实也做了很多这方面的工作。

At WorkHero, we actually also do a lot of this.

Vern，我很好奇你是否同意这种分类方式，以及从聊天到上下文再到代码的这个递进过程。

Vern, I'm curious if if you agree with this this kind of taxonomy and the progression from chat to context to code.

对，说得非常准确。

I mean, yeah, it's spot on.

而且我……

And I...

就连我们最近看到的Anthropic和OpenAI的一些变化也验证了这一点，因为聊天始终是入口。

Even just some of the changes we've been seeing lately from Anthropic and from OpenAI kind of validate that this is the case because the chat is always the the gateway.

Gary有过一些很棒的对话。

Gary had some great conversations.

这和他之前那些对话中的经验相吻合，然后这自然就引导人们进入下一步。

It lined up with his experience about what he already knows from the the conversations that he had, and that kind of leads people into, okay.

我可以信任它，因为它在某个领域是专家，而那个领域我也是专家。

I I can trust it that it's a subject matter expert in something that I'm a subject matter expert in.

现在让我们把它应用到一些数据、一些信息上。

Now let's actually apply it to some some data, some information.

我最喜欢做的事情之一，就是用Claude或ChatGPT进行上下文提问，一旦你把它连接到你的Housecall账户，就直接问它，嘿，我们上个月做得怎么样？

One of my favorite things to do with a contextual question with Claude or with Chad GPT is once you've connected it to your Housecall account, just ask it, hey, how did we do last month?

就让它自己发挥，让它自由运行。

And just let it cook, just let it rip.

它会查看你运行的所有工作，查看发票，查看所有东西，然后它会非常直白地告诉你，你做得怎么样，哪里需要改进，你错过了什么，你忘了这些未结发票，诸如此类的事情。

And it's gonna look at all the jobs you ran, it's gonna look at the invoicing, it's gonna look at everything, and it will be bluntly honest with you, here's how you did, here's here's where you need to approve, here's what you missed, you forgot about these open invoices, all of that kind of stuff.

对，上下文……

And, yeah, context...

添加上下文是下一个层次，然后从在电脑上下载应用程序的经验来看，Claude桌面版和现在的ChatGPT，都只是一个按钮。

Adding that context is is the next level, and then, you know, just from the experience of downloading the applications on your computer, Claude desktop and now Chai GPT, it's just a a button.

你只需点击一个按钮，就能从ChatGPT的工作环境切换到OpenAI环境中的Codecs，然后从Claude桌面版切换到Claude Code。

You just click a button and you go from Chai GPT work to now Codecs in the OpenAI environment and then Claude desktop to Claude code.

对，就是这样……

And, yeah, that...

这就是自然的下一步发展。

That's the the natural next progression.

所以，没错，百分之百。

So, yeah, a 100%.

这绝对是未来的方向，

This is for sure the way it

没错。

goes.

我喜欢那个直言不讳的部分，Vern，因为如果是人类来汇报这些，他们可能会有情感上的牵扯，比如朋友或家人在工作。

I like the bluntly honest thing, Vern, because, I mean, if it's a human being reporting on this and they're like, there's an emotional attachment, a friend, a family member working.

我不喜欢……

I don't like...

你懂我的意思吧？

Know what I mean?

你会得到一些……

You get some...

也许是一些含糊其辞的回答，或者把事情说得天花乱坠。

Maybe some wishy washy answers or sugar coating of things.

所以，那种直截了当、不带情感的事实信息，我觉得对老板来说真的非常重要，他们需要理解或知道，即使他们可能不喜欢这个答案，因为很多时候，直白的真相往往不讨人喜欢，但我认为，在经营企业时，说出直白的真相非常重要，百分之百。

And so the bluntness and the factual information that is detached from emotions, I think is really, really important for a business owner to understand or to know, even though they may not like the answer because a lot of times blunt truth sometimes isn't favorable to the person, but I think that it's very important that a blunt truth is told when running a business, a 100.

你提到你和Claude的对话带英国口音。

You mentioned your Claude conversation had a British accent.

桌面应用的好处就是，你可以加入一些其他上下文。

That's a great thing about the desktop apps is what some other context you can put in there.

你可以给它上下文，这就是……

You can give it the context, and, you know, this is...

这可能有点过了，但如果你愿意的话，你可以。

This might be a little bit much, but you could if you wanted to.

你可以给它设定一个上下文，让它用你最喜欢的商界人士的声音跟你交流。

You can give it the context that you wanted to communicate with you in the voice of your favorite, you know, business person.

比如说，你想用沃伦·巴菲特的风格来评估上个月的情况，或者想让安迪·弗泽拉来指导你上个月的表现，然后把你狠狠批评一顿。

So let's say, for example, you wanted to evaluate last month in the style of Warren Buffett, or you wanted to coach you on how last month last month went in the style of Andy Furzela and just rip you a new one.

如果那种你认同的声音能激励你，你可以提供这些额外的背景信息，然后它就会用那种沟通风格来回应你。

Like, if it's gonna be the kind of voice that you identify with that's gonna motivate you, you can provide that additional context, and then it will use that in the style of communication that it uses.

我很快地...

I'll quickly actually...

因为这次对话的目标之一，就是让每一位企业主或者想成为企业主的听众，至少能达到第二个层次，也就是上下文层次，不只是使用Claude或ChatGPT，而是真正把它们连接到你的系统里。

Because one of my goals for this conversation for any business owner or wannabe business owner listening is to at least get to this second level, this context level where you're not just using Claude or ChatGPT, but you're actually connecting it to your systems.

我马上来点实际的，给大家看看我的Claude实时账户。

I'll just get super practical for a second and show you guys my live Claude account.

如果你往下到这里...

If you go down here to...

点击你的名字，然后进入设置，再往下到自定义下面，会看到连接器。

And click your name and then go to settings and then go down to under customize, it says connectors.

在这里你会看到所有内置的...

This is where you will see, all the built...

这些是内置的。

These are built in.

所以你只需要在这里添加浏览连接器，这些是我已经连接好的，但假设我想连接Asana。

And so all you have to go in here is add browse connectors, and these are the ones I've already connected, but say that I wanted to connect Asana.

我不使用Asana或其他现成的软件工具，这些工具都有数据存储。

I don't use Asana or any other off the shelf off the shelf software tools that you use that have stores of data.

一旦你连接了这些，你就可以开始查询和提问，做出明智的决策。

Once you connect these, then you can start to query and ask questions that that make smart decisions.

再往上一层，也许有一件事我确实想在我们结束之前提一下，Vern，那就是你实际上怎么……

The next level after that, and maybe one thing that I'd make sure I do wanna cover, Vern, before we go is, well, how do you actually...

对于那些听节目的HVAC老板们，你们怎么用Housecall来构建MCP？

For those HVAC owners listening, how do you how do you build the the MCP with Housecall?

你怎么连接你的……

How do you connect your...

不管是Jet、GVT还是Claude，怎么连接到Housecall Pro，以确保你……

Whether it's Jet, GVT, or Claude to Housecall Pro to make sure that you're...

这样你就能查询那些系统了？

You know, that that you can query those systems?

我知道你录了一些视频讲这个，也教育了不少人。

I know you've you've recorded some videos on this and have educated a lot of people.

有什么建议吗？

Any advice on how to do that?

我讲这个话题的时候，通常会分三个层次来讲。

When I'm talking about this particular topic, I try to present it in three different levels.

第一层是最费劲的方法。

Level one is the hard way.

基本上你可以直接去Housecall Pro的API网站，也就是文档网站，然后像我当初建MCP连接器那样，打开你的云端桌面或者ChatGPT Codex桌面应用。

You can basically just go to the Housecall Pro API website, the documentation website, and do what I did when I built my MCP connector is you you have your either cloud desktop or your ChatGPT codex desktop application open.

你正在做一个编程项目，然后你把每一个API端点都复制粘贴到那个AI里，跟它说，给我建一个MCP。

You're you're working on a coding project here, and you're gonna copy and paste every single API endpoint into that AI and say, build me an MCP.

你可以这么做。

You can do that.

这非常耗时。

It's incredibly time consuming.

我大概花了二十四小时，也就是整整三个工作日，才完成这些，包括构建、测试、验证，确保它正常运行，然后把它设置好。

It probably took me probably twenty four hours, so three straight workdays just to to do that and then build it and test it, validate it, make sure it was working properly, and and set that up.

当然，你可以直接去做。

By all means, go right ahead.

如果你对这个话题非常熟悉，对处理AI也非常熟练，那完全在你的能力范围之内。

If you're super comfortable with this topic, super comfortable dealing with the AI, that's that's well within what what you can do.

下一个层次就是直接使用别人已经创建好的。

The next level to that would be to just use one that someone's already created.

所以我创建了一个，然后发布在……

So I I created one and I posted it on...

它在一个叫Appify的系统上可用。

It's available on a system called Appify.

我和他们没有关联。

I'm not affiliated with them.

那只是一个各种MCP和代理的市场。

It's just a marketplace for different MCP and agents.

你可以去用那个。

You can go and use that one.

还有另外几个在外面流传。

There's a there's a couple other ones that are out there floating around.

据我所知至少有三个，但我……

There's there's at least three that I'm aware of, But I'm...

我做了我自己的，所以我信任我的，我知道它能用。

I made mine, so I I trust mine, and I know it it works.

Vern，我们可以试着链接你的……

Vern, we can try to link to your...

那个Appify MCP，放在这里的节目笔记里。

That Appify MCP in the in the show notes here.

好的。

Cool.

谢谢。

Thank you.

而且市面上有些是只读的，但我自己搭的是完全访问权限。

And some of the ones that are out there are read only, but I I set mine up for full access.

你控制MCP能不能修改Housecall Pro账户信息的方式，就是看你在Housecall Pro账户里用的是哪个API密钥。

The the way that you manage whether or not the MCP can change information in your Housecall Pro account is by which API key you build in your Housecall Pro account.

但是我做的那个……

But the the one...

那个MCP……

The MCP that...

我搭的那套系统是完全开放的。

System that I built is wide open.

所以那算是第二层。

And so that's kind of tier two.

与其自己费劲搭MCP桥，不如直接用别人搭好的桥。

Instead of doing the work of building the MCP bridge, use someone else's bridge.

然后第三层就是，如果你想直接跳过，按一下遥控器上的快进键，直接跳到能用的状态，我们可以聊一聊，我可以帮你设置好，屏幕共享什么的，直接帮你搞定，让你系统里已经有能用的MCP。

And then tier three is if you just wanna skip ahead, hit fast forward on the remote, and just get to the point where it's working, we could chat, and I could set this up for you, screen share, whatever, and just set it up so that you've got the MCP ready to go working in your system.

这就是我讲这个事的时候提到的三个不同层次，因为建立这种连接真的很有价值，这样你和AI的对话会非常非常高效。

And that's kind of the three three different levels that I I present when we're talking about this because there's there's a lot of value in in having that connection set up so that your conversations with your AI are super, super productive.

特别是现在用桌面应用的时候，他们最近大力推的一个功能就是远程访问你的桌面应用。

And especially when you're using the desktop apps now, the feature that they've been really big on pushing lately is remote access to your desktop app.

所以桌面应用功能齐全，性能最强，但你可以通过系统的移动应用远程控制桌面上的应用。

So the desktop app is all singing, all dancing, it's got the most horsepower, but from the mobile app of the system, you can remote control the app on your desktop.

所以我们之前聊的那些，比如你能不能有一个系统，你只要给代理发个语音留言，它就能进入Housecall并帮你完成所有那些事情？

So all of the things that we talked about, like, could you have a a system where you just do a voice note to your agent and it goes into Housecall and does all that stuff for you?

这就是实现那个目标的一条路径。

This is a path to doing that.

嘿。

Hey.

我们从手写工作单的复写纸翻页本，走到现在，真是进步太大了。

We've come so far from the the the carbon copy flipbook for your your work order, your handwritten work orders.

好的。

Cool.

我……

I'm...

我知道我分享了很多，但我……

I know I've been sharing a lot, but I...

我再展示一个东西，然后我会念给大家听，方便没在听的人。

I'm gonna flash up one more thing, and I'll read over this for those not listening.

这是另一个……

This is another...

这是来自……

This is just from...

这是我们在Vern这里认为有效的一些做法，我想听听你的看法，作为总结，那就是……

Some of the things we believe work here in in Vern, I'd love to get your take on this as sort of a wrap up, which is...

我知道我们聊了很多信息，但有哪些基础的事情，是你在尝试构建或自动化之前，作为技工行业和暖通空调的小企业主，必须做对的？

I know we've talked about a lot of information here, but, like, what are just some foundational things that before you even try to build or automate things that as a small business owner in the skilled trades and an HVAC you need to be doing correctly?

其中一些我们相信的东西，也是我最初那次演讲的一部分，就是，比如，直接用个FSM软件。

Some of these that we believe in, again, was a part of this original presentation I did is, like, just use an FSM software.

正确使用它。

Use it correctly.

那是一个数据存储的地方。

That is a store of data.

团队问责是第三点。

Team accountability is this third point.

一定要让你的团队，你的技术人员，负责把照片和笔记录入那个软件。

Make sure to hold your team accountable, your technicians accountable to getting their photos and notes into that software.

在可能的情况下，要有清晰的流程意识，比如，你知道……

Where possible, have, like, a clear sense of process and, like, where you...

哪里出了问题，或者，就像我们常说的，你不能自动化一个已经坏掉的流程。

Where things are broken or or you can't automate a broken process, as we like to say.

然后人员这块就很有意思了，对吧，因为这些事情越来越难做，你可能不会有人专门在内部处理。

And then personnel is the interesting one, right, where increasingly, these things are really difficult to do and you're probably not gonna have someone in house.

所以我们会链接到Vern的信息，去找那些可能能帮到你的人。

And so we're gonna link to Vern's information, look for people out there who might be able to help you.

我不知道你会在这上面加什么，就是在大家……之前。

I don't know what you would add to this in terms of just, like, before anyone...

甚至什么……

Even what...

你的清单上有什么？

What's on your checklist?

在任何人尝试把AI应用到他们的业务之前，你会加什么？

Anything you'd add before anyone tries to implement AI into their business?

我去帮别人处理这类事情时，看到的模式就是数据完整性。

The the pattern that I see when when I go and try to help somebody with this this kind of thing is is data integrity.

这其实就是说，你有没有按系统该用的方式来用，还是我们走了捷径？

And it's just a fancy way of saying like, are you using the system the way that it should be used or are we taking some shortcuts?

我们是不是只用名字字段，而其实可以用名字和姓氏？

Are we only using the first name field when we could be using the first name and the last name?

这听起来好像挺蠢的。

And that seems like such a stupid thing.

就是特别小的一件事。

It's so it's so minor.

感觉好像无关紧要。

It seems so irrelevant.

但当你想要创建一个自动化流程，给房主发条短信提醒一下，比如，嘿。

But when you're going to go and try to create an automation that will send a text message to the the homeowner with a heads up, like, hey.

技师正在路上，或者嘿。

The tech's on their way, or hey.

我们发了这张发票。

We we sent this invoice.

请查收一下。

Please check for it.

数据完整性就超级重要了。

That data integrity is super important.

如果你想非常正式地说，比如‘琼斯先生’，如果姓氏字段是空的，那就会变成‘先生空白’，因为那里什么都没有。

If you wanna be very formal and you wanna say like, mister Jones, if that last name field is blank, it's gonna say mister nothing because there's nothing there.

所以这就像是任何大型项目的第一步：要清理数据需要做什么？

So that's kind of like the the very first step of any kind of large project is what's it gonna take to clean the data?

数据清理永远是起点。

Data data cleaning is always the starting point.

它……

It...

电子邮件也是一样。

The same thing with emails.

你想让你的代理访问一个电子邮件账户。

You wanna give your agent access to an email account.

收件箱里是不是有一万封邮件？

Does the inbox have 10,000 emails in it?

因为那样是行不通的。

Because that's not gonna work.

读一万封邮件会烧掉你一百万个令牌。

You're gonna burn a million tokens reading 10,000 emails.

先把收件箱清空，然后再走这个流程，虽然这像大扫除一样又麻烦又头疼又痛苦。

Getting to inbox zero first and then going through this process as much of a chore and a headache and a pain as it is spring cleaning.

没人喜欢大扫除。

No one likes spring cleaning.

有些人喜欢，但这绝对至关重要。

Some people do, but that's absolutely critical.

我总是这样解释，我学到的关于数据完整性的道理就是——

And the way that I always try to explain this, the way that I was taught about data integrity is what what...

垃圾进，垃圾出。

You know, garbage in, garbage out.

所以，如果你连地址都不统一，Housecall Pro 会在你开始输入时自动填充地址，你只需要从下拉菜单里点选。

So if you're not consistent on even addresses, Housecall Pro will automatically populate addresses for you if you just start typing it, and you just click it from the drop down.

如果你不用那个下拉菜单，你的地址就会一团糟。

If you're not using that drop down, your addresses are a mess.

那绝对会是一场灾难。

They're gonna be an absolute disaster.

所以垃圾进，垃圾出。

So garbage in, you're gonna get garbage out.

你可能会因为没注意哪个邮政编码对应哪个地址字段，导致文本跑到另一个州去。

You're you're gonna have text running to a different state because you weren't careful about which ZIP code got assigned to that address field.

所以我要补充的唯一一点就是，如果你的数据不干净，就要准备好清理数据。

So that that would be the only thing I would add to this is be ready for cleaning up data if your data is not clean.

Vern，我们得请你再来一次，因为要讲的内容太多了，但感谢你来做客。

Vern, we're gonna have to have you back because there's so much more to cover, but thanks for coming on.

这真是个重要的话题，感谢你这么快就讲得这么清楚。

This is a really important topic and appreciate you breaking it down so so quick clearly.

非常感谢。

Thank you so much.

我真的很感激。

I I really appreciate it.

如果你以后真的拿到一些关于哪种声音风格有效的数据，我很想知道，因为从心理学角度来看，这很有意思，因为那就是全部。

If if you do ever get some data on what voice style works, I'd like to know because it's interesting to me from a psychology point of view because that's all it is.

这就是心理学，人类心理学。

It's psychology, human psychology.

你启发了我。

You've inspired me.

我今天就告诉我的经纪人，去弄一些不同的声音来试试。

I'm going go tell my agent to spin up some different voices today.

太好了。

Sweet.

跑一下这个测试。

Run this test.

试试霍克·霍根的声音。

Try Hulk Hogan.

那可能会有效。

That that could work.

谁知道呢。

You never know.

那个会很贵。

That that one would be expensive.

那个……

That...

那就是

That's the

预付款。

pre invoice.

你得额外付钱

You had to pay extra

为了那个。

for that one.

所以你得付一笔钱。

So you have to pay one.

好的。

Okay.

我之前不知道这个。

I didn't know that.

好的。

Okay.

太好了。

Cool.

行。

Alright.

谢谢，Vern。

Thanks, Vern.

我真的很感激。

I really appreciate this.

非常感谢。

Thank you so much.

非常感谢。

Thanks so much.