Build a Shared AI Brain for Your Company (No Code Required)

Most founders are the bottleneck in their own company. Your standards, your voice, your client knowledge. It all lives in your head. And AI mirrors the problem when everyone on your team prompts it differently.

A shared Dropbox folder just became the biggest AI breakthrough Alane Boyd and Micah Johnson have experienced since they started using AI. 

The discovery? Consistent results aren't about better prompting. They're about giving AI access to the same company knowledge every time.

In this episode, Alane and Micah walk through how to build a shared knowledge vault using Obsidian, Dropbox, and Claude Desktop that lets your entire team operate at your standard without you in the room.

You'll learn:

• Why founders become the bottleneck and how AI makes it worse without shared context
• What a CLAUDE.md file is and why it's the single most important file in your AI setup
• What belongs in your knowledge vault, the stuff that currently lives only in your head
• How Skills turn shared knowledge into repeatable workflows anyone on the team can run
• The 30-minute minimum viable setup to start getting out of the bottleneck this week

If you're ready to move from siloed AI experiences to shared company intelligence that delivers consistent results, this episode breaks down the exact system.

Alane mentioned our free 30-minute webinar on Claude Cowork during the episode. If you'd like to learn more, join us on March 18 at 2:30 pm CST. Register here.

To hear our first episode on the power of Claude Cowork, listen to The Future of Work: Claude Cowork next.

Tools/Platforms Mentioned

• Claude Desktop - AI assistant with file access capabilities
• Dropbox - Cloud file sharing and synchronization
• Obsidian - Note-taking app for markdown files and knowledge management
• Wispr - Voice-to-text dictation software for Mac

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:

• Free Claude Webinar
• 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

Connect with Us:

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

Transcript

Micah Johnson (00:01)
What if I told you that the biggest unlock with AI that we have found since starting to use AI is not better prompting, it's just a shared folder on Dropbox. Seriously, this one folder has completely changed how we work with Claude and more specifically Claude Desktop.

Alane Boyd (00:20)
And if you listened to last week's episode, we ended it on a cliffhanger on how we're using Dropbox with Claude Desktop for sharing results between our teams securely.

Micah Johnson (00:31)
So what we see ourselves, well, used to be ourselves, Alane, but what we see with a lot of the clients and the companies that we talk to when we're presenting and speaking is that a lot of times everybody's stuck in this loop, especially from a business perspective. We have to re-explain to the AI everything about our company or give it so much context and...

Alane Boyd (00:36)
Mm-hmm.

Micah Johnson (00:54)
There's some workarounds right now that work okay, but they're time consuming and they're only as good as the people that are actually doing it. So you get a ton of inconsistent outputs. And if you have people on your team that are new or don't have all the context or don't make good judgment calls, then their outputs are going to be worse than other people's outputs. So if you're interested in getting past all of that,

and never having to regurgitate all this information to AI over and over again, this episode is for you.

Alane Boyd (01:29)
One of the problems that I see a lot, Micah, when we're talking about AI with other companies is this idea of how do I get each person using it the same way? How do I make things consistent? Because a lot of it is a siloed experience when you're just chatting with AI. And what I'm seeing with what we're doing now between the Dropbox and Claude Desktop is that now we're having a shared usage and like a core

company information that you and I can use to work with, whatever AI experience we're needing at that time, but that is still going back to our core company knowledge that we set, that we wanted it to have access to.

Micah Johnson (02:09)
Yeah. So I've got a question for you, Alane, cause I haven't asked you this before, but we've implemented this. I've been using it for a little bit to work out all the kinks. You and I have been working with this together with Claude Desktop. We have our own Claude desktops installed on our own machines, but we're sharing a data set for the Claude Desktops that we have installed, which we'll cover here in a little bit on how we have it set up. I would love to hear in like.

one word or one sentence like the difference that you're seeing or experiencing.

Alane Boyd (02:44)
Gosh, how do I get this into one word? It's exhilarating. Like we do cool shit, Micah, and this is the coolest thing that, and I think what we heard from feedback in our other episode is like how excited we were that they could feel our energy. And what we're doing, since we've implemented this, is exhilarating. Like it is the most exciting thing. It removes so many roadblocks. Like I was telling you yesterday, it's like,

Micah Johnson (02:59)
You

Alane Boyd (03:10)
There are things that have been sitting on my list of things that I could not put the time into to overcome that I did in five minutes yesterday.

Micah Johnson (03:19)
Yeah.

Alane Boyd (03:21)
with these resources, which is this Dropbox folder. We're gonna talk about what Obsidian is and then the Claude Desktop. Complete like it's wow.

Micah Johnson (03:31)
Yeah. And I mean, I couldn't have come up with a better word, Alane. I knew if I asked you that on the cuff, like you would have such a good answer. You did not disappoint. That was fantastic. And it is, it is re-energizing how fun it is to work, which sounds really ridiculous saying, but I'm having so much fun, just doing work.

Alane Boyd (03:36)
Who knows what'll come out of my mouth?

Mm-hmm.

Micah Johnson (03:58)
I don't know how else to explain it. We get up in the morning and I'm thinking about like, man, how could I improve this? And when I start chatting with Claude and Claude Desktop, it's able to pick back up on what we talked about the day before, or it's able to get the context that it needs when I want to explore a concept. For example, if I,

came up with an idea over the night and woke up with this idea I wanted to explore. I can chat with Claude Desktop and it will go and look at our personas for our client and our audience. It will go and look at our offerings. It'll look at our price points. It'll pull all this context. So then when we're chatting, it's speaking to exactly what we do and who we are, not generic training data.

Alane Boyd (04:31)
Thank

Micah Johnson (04:47)
And this is no RAG database. We don't have to set up any of that. It's a different setup. Context is still a concern, but the way that Claude Desktop works and how this integrates is phenomenal.

Alane Boyd (04:47)
Mm-hmm.

So Micah, let's talk what our setup looks like. Cause we've been throwing out, you how does Claude Desktop even work with Dropbox? we doing this securely?

Micah Johnson (05:10)
Yeah. So we have Dropbox, which holds a set of files and folders. And then you and I have access to those file and folders that Dropbox does the syncing. So when my Claude Desktop runs, I run it in that synced folder from Dropbox. And when you run Claude Desktop, you run it from that synced folder in Dropbox, which means if I make additions, let me rephrase that.

If I ask Claude to make some improvements or some additions or edits to any of the information contained within these files and folders, then when you're working with Claude, it's going to have those updated bits or documents and information or knowledges. it knowledges isn't a word, but we'll keep that in here. So for example, let's say theoretically we change a price point on one of our programs.

Alane Boyd (05:47)
Mm-hmm.

Mm-hmm.

Micah Johnson (06:06)
I would

have Claude update our program documentation that's in this vault. And as soon as you start working with Claude the next time, it's going to have that updated information.

Alane Boyd (06:16)
same information.

So I also want to say, Micah, I'm going to cut you off for just a second, is that we have a Dropbox company account for our team. And so we have set up this specific folder that Micah is referencing in this example. Him and I are the only two people in our company that have access to this folder so that we're keeping this information secure. And Micah, like one of the things that I'm thinking about and where we're talking about sharing data is creating that Stripe dashboard in about

Micah Johnson (06:24)
Yes.

Alane Boyd (06:46)
30 minutes. You created the plan using Desktop, it synced with Dropbox, and then I was able to go in and review the that way I could ask any questions, see if there were any things that I thought was missing from it before we built. And it was all using Claude Desktop, with syncing with our Dropbox folder and then viewing it in Obsidian. So I'll let you then...

piece together the rest of this Micah, which is where Obsidian fits.

Micah Johnson (07:11)
Yeah. So Obsidian, it is and started as a note taking application. That's all based on the simple structure of folders and markdown files. Now, coincidentally or not, LLMs like Claude and ChatGPT and Gemini, they love taking in markdown format and outputting in markdown format. So this is like a match made in heaven. And this is why Obsidian makes so much sense.

It's a free application that you can download on any OS. You download it and it's like a reader or a way to view these files. But these are just normal text files. They're markdown files. So don't save passwords and API keys in them. Warning. But because they're just folders and files in markdown format, AI loves it. So Obsidian gives us a nice front-end interface to view all this stuff.

And it's perfectly suited for a tool like Claude Desktop to be able to access and roam and pull out what it needs and have linking between files and read and write and edit and delete as you see fit, which means you can run this entire knowledge base by just prompting with Claude and then Dropbox syncs it to the cloud. So anybody else that has access gets that information almost instantly.

Alane Boyd (08:31)
Yeah, so, Micah, the way I think about this, and tell me if this is right, I see Obsidian as like this middleman communicator between our Dropbox and Claude. It's syncing with our Dropbox, so we're not losing any of this information in Obsidian, but it's translating the information really easily between, you know, what do we have in Dropbox? It's creating these readable files for AI and then vice versa. It comes in from AI and then syncs with our Dropbox.

Micah Johnson (08:58)
Yeah, honestly, technically you could do all of this without Obsidian at all because it's just folders and markdown files and Claude, in our case, is creating all of that. But what Obsidian really is is the ability to see these files in formatted views and see the linkages between these files and view the knowledge graph that's being created as you're

Alane Boyd (09:05)
Mm-hmm.

Micah Johnson (09:25)
adding to and editing this whole set of files.

Alane Boyd (09:30)
Yeah, and this is how, you fed it, you know, our packages, our services, our messaging, things about just us as a company, thinking about rebuilding our website. And so this is how, you know, we could put the information in, and then we did use Pencil, which we talked about in the last episode for the design of the website. But this is how we had our core company messaging.

Micah Johnson (09:41)
Yes.

Alane Boyd (09:54)
information for it to even know, hey, what do I need to do here? What are the brain colors? What are the logos? You know, whatever we needed could be saved here, communicated with Claude Desktop, and then created from there using one other software. But this is where that core information sat.

Micah Johnson (10:10)
Yeah, absolutely. mean, and to be clear, I haven't written a single line of text in any of these files. I chat with Claude and the first thing that I did was had Claude write its own instructions on how to manage this vault. And so every time I work with it, it knows the formatting, it knows the structure, it knows everything, how I want this to operate. And so it created a standard for itself.

And so I know myself well enough that I know the standards that I want, but I sure as hell know that if I went and started editing this vault, I would not follow my own standards because I would be rushed or not thinking about it. But Claude follows the standards every time. So it's easier for me and less time consuming to ask Claude, Hey, don't forget, I need to add a note about our new programs and here they are, or

here's the file that we were working on internally, et cetera. I give it the context, and it beautifully writes the notes. It will connect different notes together, based on content. And every time we have a session together, Claude and I, this is getting really weird, I say, all right, let's end this session. And it has instructions that says, end, actually create

an addition to the weekly session log. And so everything that we do, everything that we add, everything that we take away, everything that we accomplish is automatically getting logged to the session log. And then I have a recurring task. I know this is getting totally crazy, Alane. I have a recurring task at the end of the week that reviews all the updates, including the session log, including our current focus, where everything's at, what happened in this vault,

Alane Boyd (11:50)
I know.

Micah Johnson (12:01)
and sends me and you a beautifully designed email of the state of the company, the progress that we made, what happened, what got done, all of that stuff.

Alane Boyd (12:03)
Mm-hmm.

roadblocks.

What Alane didn't get done. part that Micah is talking about is a little bit of a tangent. He added this piece and actually at first I was like, okay, that's okay. But it is kind of cool because we do forget how many things we got done during the week. And if nothing else for me, it was like, wow, we actually did some shit.

Micah Johnson (12:16)
What Alane didn't get done.

Yes.

Yes.

Alane Boyd (12:37)
And then it clearly just kind of packages up like, hey, you didn't spend time on this other thing that you said was one of the goals. And it's like, yeah, okay, thanks Claude.

Micah Johnson (12:48)
Yeah. Here's another example. I was super busy a couple of days ago and I had started a couple things and fortunately Claude took some notes on all of that. So I was able to go back to Claude and be like, Hey, check out the vault and see if I left anything hanging. And so it would check out the session log and it would check out where things are at and look at the related notes that were linked and then go, here's three things you might want to double check on. And sure enough, there were one or two things in there where I'm like, oh yeah.

Alane Boyd (13:04)
Mmm.

Micah Johnson (13:17)
Oh, that's right. I was starting that and it got me back on track. So for any of those with the ADHD brain that jumps all over, this is a lifesaver. All right. So what's that?

Alane Boyd (13:26)
Or a bad reminder. Or

a bad reminder.

Micah Johnson (13:32)
Yeah, or a bad reminder.

So one of the things that we struggled with as we got this set up, Alane, was identifying what should go in the vault and what should not go in the vault. So let's already assume you've got Claude Desktop, you set up a tool like Dropbox, and you've set up Obsidian. So, and I guess I missed the part where to set up Obsidian, you install it, and you start your Claude Cowork session in...

the folder of the vault that you created in Obsidian. And all you have to do is tell Claude, hey, I'm going to create a new Obsidian. I just created a new Obsidian vault in this folder. Help me initialize this or help me set this up as a knowledge base for my company. It'll ask you a couple of questions. It'll make some suggestions. You give it the feedback and it will create the entire folder structure. It'll create the starter files. And then as you're working with it, you just...

ask it to make changes. Like, I don't think we really need an index in every folder. Oh, okay, cool. I'll delete those. And then you give it permission to delete those files and you just kind of work through it from there. But there's the bigger question of, do I just dump everything in the vault and I have this 2,000 note file vault, or do I whittle it way down and keep the minimal?

The answer is I'd start with the minimal. The more information, the more chances that it has to get confused. So some of these things are going to be like archive notes you want to put in there. And some of them are going to be core company things like offerings or personas of your client audiences. Things like that are great, great attributes and great notes to have in there. But,

if you just like every idea you've ever had in your life and you just keep stocking it full and stocking it full, Claude's not gonna know what's real and what's not. So I would be very, very cautious and be careful about what you're adding and spend a little time tidying it up every once in a while.

Alane Boyd (15:32)
I mean, it goes back to context and we talk about this often in presentations and when we're talking with people is that the AI only has so much context. We can't overload it with so much information or it's just kind of running in circles. There's too many places for it to go. Just like a human would feel if you're overloading it with way too much information, which is why if we go to a three hour presentation, we're not going to remember every single word that came out of the presentation. We're only going to remember pieces of it. So,

we just really have to think about like, is this something core to the business that we want to keep it coming back to? And then is other stuff like you said, Micah, notes that, you we don't have to keep necessarily in the vault. And there's a way in Claude Desktop to say, do you want this to be in a folder or are you just gonna work, you know, on your own here with Cowork.

Micah Johnson (16:22)
Yeah, exactly. So for our vault in particular, here's some examples that we found super, super helpful. We have a programs folder that contains all the information about our cohorts, our AI Leadership workshops, our groups, et cetera. This includes schedules, includes And that way, anytime we're referencing or need to talk about or.

write anything about any of these programs, Claude can get access to that and know exactly what the hell we're talking about. We're experimenting with client folders inside of this vault where maybe we do some status updates or we do some development plans. That's a work in progress. I'm not fully decided that I want all that in the vault. You can give that a try and see if you like it. This one for sure, brand voice, writing style rules. Oh my God.

So good, like stop using those f'in em dashes, Claude.

Alane Boyd (17:18)
Micah, one thing that I wanted to say on this is that you can create different vaults that are about different things. So that maybe you have one about, you know, core company branding and messaging and packages and programs. And then like I have a different vault for my speaking brand. And so I work with it in a different vault that has that core information. So, you know, you can have these different vaults depending on,

Micah Johnson (17:23)
Yes.

Alane Boyd (17:45)
department or company or whatever it is.

Micah Johnson (17:49)
Yeah, absolutely. So that gives us the context. Now, another feature of Claude Desktop is skills. And maybe we do need a whole episode on this as I'm not going to end it right here on a cliffhanger for you. Skills combined with the vault and the context with Claude Desktop is incredible. So really recent example, we're planning out this podcast episode, Alane.

And we have a certain way that we want this structured for us so that you and I can work together on what the hell are we going to talk about? And so we built, we built, we had Claude build a skill for itself to grab context from the vault. And since we're talking about the vault, it could actually look at what we're doing and help us write pieces.

Alane Boyd (18:24)
Mm-hmm.

Micah Johnson (18:42)
Of the outline of this episode and put it in a specific format. Now the next podcast episode that we want to get a bulleted list of a specific topic and then Alane and I can go through and do it. Claude has that skill, we just have to say, Hey, here's a few ideas. Can you format this into a sheet for Alane and I to work through while we're talking about our podcast planning and it will load up that skill and reproduce.

this exact same format that we need every single time.

Alane Boyd (19:15)
Yeah, now that that skill is created, Micah, that can be shared with me and then I can go in and use that skill whenever I'm using Claude Desktop, which, you know, we're really looking at how do we break these silos? How do we share information? And this is creating that dynamic for us so that we're able to share information, these skills that keeps consistency between what we're doing too. We want to sound organized when we're on our podcast,

as much as we can. And so even just like us using this as an outline and how we like to do it and that keeps us more organized, we don't have to keep telling Claude that every single time. Now it's a skill that we can reuse.

Micah Johnson (19:55)
Yeah. Or even hope that it gives us a format in a similar way, which is what we were kind of dealing with before. Now there's a big caveat to all of this, Alane. And I think we kind of alluded to it earlier in this episode, but there's a reason that I just started with it and now it's just me and you. And we haven't gone, Hey team, here's Claude Desktop and here's this vault. And here's all this company information have at it, because that would be the biggest shit show.

Alane Boyd (20:02)
Mm-hmm.

Micah Johnson (20:25)
we could ever imagine.

Alane Boyd (20:28)
Yeah, I really are, because this is new technology. This is new. We're implementing it ourselves. We are big believers in training and giving people the knowledge to execute using these tools because then we get a better performance out of them. If we keep just throwing people to the wolves on this kind of stuff, then we're gonna get mixed results, just like with anything. And so, Micah, you started out with testing it, then...

Micah Johnson (20:50)
Yes.

Alane Boyd (20:54)
you brought me into the mix and we're loving it. We're figuring out some, hey, like this, this is going to be a roadblock with the team. Let's document how to do this. What is the company way of doing it? And then now we're bringing on two more members using this. We're starting to talk about it with our team saying, Hey, this is what we're experimenting with. We're using it in these use cases that are applicable to their role so that they can start getting excited about us testing this out, how it could relate to them.

I mean, we've got one woman on our team and she is just stoked about it from just hearing us talk about it. And it's very specific to how she can use it.

Micah Johnson (21:34)
Yes. And I think it still comes down to the fundamentals. We still have to create standards. We still have to train people how to do it. We still have to support them in making good judgment calls because what some of this is doing is taking away the busy work, which is what we want. But what that leaves is, well, we have to decide if the output is good or not. So we couldn't just say, Hey, Claude, here's a topic. Here's the vault.

Alane Boyd (21:37)
Mm-hmm.

Micah Johnson (22:02)
Write the podcast for us and then Alane, you and I record this podcast because even though it has context It's not us. It's not our voice. It's as close as it can get, there's still judgment that you and I spend time on before every episode thinking through. All right. How do we really make this happen? How do we want to talk about this? What is the Does this wording?

Alane Boyd (22:10)
Listen.

I feel it.

Micah Johnson (22:27)
Completely resonate and yes, it's saving us so much more time than what we would have to do otherwise But there's still that judgment call and each day and each piece We're like building a little bit more of the blocks to build up this whole Claude Castle, I don't know that that'll work for my analogy today ⁓ But we didn't go. All right, and here's the castle and you know

Alane Boyd (22:48)
Yeah.

Micah Johnson (22:53)
that whole castle would just fall apart instantly. And we do see that happen a lot of times. And a lot of the reports that come out say that when you just launch AI tools and when you just launch these new capabilities and when you just launch new software across your team without defining the standards, without defining the training, without understanding how to use these, what are the pros and the cons, that's where it all fails.

Alane Boyd (22:56)
Yeah.

I mean, over and over again, it's not just with AI, like you're saying, it's this is a big piece of, why we're being thoughtful about how we're using it, what we want to use it for. Everything still depends on a human to look at the result. You know, we're getting to be better at what we do, Micah, and implementing our vision and strategy by using these tools. So we're not just sitting there doing the tedious work that, you know,

Micah Johnson (23:22)
Yes.

Alane Boyd (23:46)
moving data between things or writing out things. I mean, you don't even hardly type anymore.

Micah Johnson (23:51)
I mean, this is to the point and this is probably now let's get to the cliffhanger of maybe the next episode or an episode will record in a little bit, which is how do you connect Claude Cowork to your other platforms? So I have something called Wispr on my Mac and it allows me to press a button on my keyboard start talking and it will dictate everything that I'm saying into text into any text field that I'm

that I'm clicked into, which is perfect for Claude Desktop, because then I can just verbally say what I want it to do and hit enter and I'm done. We don't even have to type for that. And then it goes out to the Obsidian vault gets the context that it needs. Now I've connected it to my Gmail. I've connected it to my calendar. I've connected it to Pipedrive. And this is my new favorite one as we're working deals. And as we have stuff.

the bane of my existence and I think so many people's existence is putting data into a CRM. That sucks. So maybe this is the cliffhanger for the upcoming episode is how do you get Claude to connect to a tool like Pipedrive and you just say, hey Claude, check out the latest emails in my inbox. Can you create deals and Pipedrive for me on those? Let's do separate deals for each opportunity we talked about in that thread. And it works.

Alane Boyd (24:51)
This one's yours.

All right, Micah, you can't give it all away.

Micah Johnson (25:19)
Well, did I give too much?

Alane Boyd (25:21)
Maybe, I don't know, but maybe just enough to get people interested.

Micah Johnson (25:24)
We'll

edit out some bleeps in there so they have no idea what I'm saying.

Alane Boyd (25:30)
Okay, so that one is going to be a great next episode or future episode where we talk about these connectors and plugins and different things on making these workflows happen. But if you want to learn more about setting up your Claude Desktop and what we've been talking about in the last couple of episodes, Micah and I are doing a webinar on March 18th. We'll link in the show notes the link to register, but we are going to be talking about setting up your Claude Desktop.

So come and join us. And if you are listening to this after March 18th, we will be posting the recording on YouTube so you can still access the information. And also, Micah, this has been a hot topic in our AI mastermind. So if you're interested, reach out to us. We meet every other week and we talk about the latest ways to use AI outside of just writing books.