How AI Creates SOPs in Minutes (ft. Hayden Miyamoto)
By Tiago Forte
Summary
## Key takeaways - **SOPs Unlock Team Scalability**: Without SOPs, knowledge lives in people's heads as tribal knowledge, preventing team scaling. SOPs act as onboarding tools so anyone can take over if you're sick or on vacation. [00:29], [00:39] - **Three Levels of SOPs**: Level one explains the full process from start to finish over months. Level two covers one department's piece, and level three is a tactical guide for one person. [00:59], [01:20] - **Pre-AI: 6 Months for Diagram**: Creating the core process diagram with whiteboard, post-its, and manual Excaladraw entry took a full day plus 6 months of refinement. It shows sequential steps from onboarding to LOI with on-market and off-market paths. [02:08], [03:28] - **AI Recreates Diagram in Minutes**: By screenshotting the old diagram into multimodal AI with a master prompt defining business context, it understood and recreated the full core process diagram accurately, including parallel tracks and KPIs, in minutes without prior sharing. [03:43], [07:47] - **AI Master Prompt Drives SOPs**: Master prompt injects style preferences, business identity, core values, and process frameworks like RACI, prompting AI to ask clarifying questions, answer from its knowledge, and confirm before proceeding. This created detailed guides like onboarding call and investment thesis in 3-4 minutes. [04:10], [09:55] - **RACI Clarifies Process Ownership**: RACI stands for Responsible (does the work), Accountable (one boss tracking KPIs), Consulted (two-way input), Informed (one-way updates). It's added to each task for clear ownership in the AI-generated diagram. [06:19], [07:12]
Topics Covered
- SOPs Unlock Team Scalability
- AI Compresses Process Mapping
- RACI Clarifies Process Ownership
- Master Prompts Teach AI Frameworks
- AI Copilots Reshape All Business
Full Transcript
[music] Hey Aiden, awesome job on the video.
People [music] loved it. In fact, I was wondering, do you want to make another one?
>> Yeah, for sure. Um, I'd be happy to show people a little bit more on the process side. That's something that's been uh
side. That's something that's been uh super necessary for for anyone who owns a business. What we're going to cover
a business. What we're going to cover today is not just an SOP, right, which is just a single standard operating procedure, but a whole sort of systemat system systematized framework for
creating SOPs. Without SOPs, knowledge
creating SOPs. Without SOPs, knowledge really just lives in people's heads. You
know, you call it tribal knowledge, and your team really can't scale. With SOPs,
it acts as an awesome sort of onboarding and training tool. If you get sick, you know, you're away for a week or you take a vacation for a week, the SOPs exists.
Anyone else can really take over. So I
would say you should make SOPs for anything that is a repeated process and especially anything that will be done by more than one person. I think of SOP as
several levels. Level one explains
several levels. Level one explains everything that happens from start to finish and it might be over the course in our case of months. The level two explains kind of one piece of it that
might involve one department right maybe a few people. And then the level three, which is a guide, is going to be, you know, a very specific part that's for one person typically, and it's the most
tactical and the least strategic. I have
a company, Acquir. It helps people buy businesses, and we have a product um called Accelerator Plus, that does a lot of the work um for actually identifying
um businesses for sale. Core process
here might be, as an example, um your delivery to success, the actual fulfillment of whatever product or service you're doing. And so each product or service would have a
different delivery to success core process. And in this video, I'm going to
process. And in this video, I'm going to highlight a delivery to success core process. I thought it might be
process. I thought it might be interesting to actually show what it looked like preAI um and now what it looks like post AI. And this is what it kind of looks like preAI. So this is
Excaladraw. It's just, you know, kind of
Excaladraw. It's just, you know, kind of like Miro but cheaper. Originally it was a whiteboard. I did this with a director
a whiteboard. I did this with a director in Toronto and we had a big whiteboard and we put post-it notes on it and we just identified the process from start to finish sequentially and that probably took a full day and then once we really
figured it out we then put it in manually into Excaladraw and we played with it in such a way that it would be a little bit easier to identify. I'll go
through very quickly. Basically, you
know, we receive money, we onboard, um we have a call with the AE, the customer, the acquisition entrepreneur.
We have an internal search uh kickoff call. And then we onboard into two
call. And then we onboard into two different paths. One path is on market,
different paths. One path is on market, meaning these are this like a marketplace of brokered listings that you can peruse. And then we have offmarket at the same time, which is
basically businesses that are not listed for sale and you're going to reach out to and see if they want to sell.
throughout this all that we don't need to go into, you know, too much detail, but basically there's like this path that involves a couple different roles in the company, right? Another path that involves a couple different roles.
There's some um overlap here and each of these things each of them has a link and that link goes to the actual SOP. The
beauty of this, you can almost think of this as like the core process in the sense that at any point you could take I could explain to you who doesn't really know anything about the business exactly
sequentially what we do on behalf of the customer. This probably took I'd say 6
customer. This probably took I'd say 6 months like a good 6 months to do and then it was still quite a bit of improvement that happened over time.
That's the old way to do it. when AI
first had the ability to become multimodal and allow you to screenshot.
I was playing with that ability and trying to understand, you know, what AI was capable of. And I actually screenshotted that that diagram that I showed, the Excaladraw diagram. I
basically just said, "What is this?" And
I was blown away that AI had a pretty decent understanding of what it was.
That to me was like was the Eureka moment. I just spend a couple hours at
moment. I just spend a couple hours at least every day kind of tinkering, testing AI's capabilities. Perfect
example of how do I how do I work with AI? This was basically I was preparing
AI? This was basically I was preparing for this video. I have my master prompt in. So, just so everyone knows, this is
in. So, just so everyone knows, this is a a pretty simple master prompt for Qua.
It's got, you know, not too many files in it in the in a project. So, it's
taking my style preferences and it's putting that it's s it's injecting that into every single prompt I say. It's
also taking everything within here, right? And so this is sort of like the
right? And so this is sort of like the business identity and this was a lot of this stuff was covered in the master prompt doc. Um so this understands like
prompt doc. Um so this understands like my unique value proposition, my brand voice. Um I have sort of the company
voice. Um I have sort of the company mission, our core values. All I did is I this was the prompt, right? So create a visual diagram for the core process of sale to delivery and to delivery success
for the accelerator plus deal search product. It knows what Accelerate Plus
product. It knows what Accelerate Plus is because of the project files. I'm
just saying it begins after the sale has been made and an invoice has been sent.
It ends when an LOI is signed and we've been paid for phase two. So I just say clear clear beginning like scope beginning and end. because of my master prompt, it always basically asks me
questions one at a time um and and tries to answer it based on its own knowledge.
It's answering its own question. It's
talking like when the handoffs happen, what the activities are, what the success criteria is, etc. And it asks me if that's right. And so I basically add something in here, right? I said there's
non-market team. It asks a couple more
non-market team. It asks a couple more questions. It has a very interesting
questions. It has a very interesting question. It's interesting that it gets
question. It's interesting that it gets this so fast. So his question here is like what happens when you find multiple deals um that could match multiple AES how do you prioritize assignment right
it identified that within a few seconds which is interesting because it's it's something that we spent a lot of time thinking over and basically I answered it here it created it right here
and what I ended up doing just just so we look at this you know the the final version is I I then asked it to add REI
um so Reie is is kind of it's process speak. It's it's R A CI, right? And it
speak. It's it's R A CI, right? And it
stands for who's responsible, who's accountable, who's consulted, and who's informed. And it's very useful for
informed. And it's very useful for process because it it shows very clear ownership. So responsible would be the
ownership. So responsible would be the party that's actually doing the work. It
could be multiple people. Responsible if
if people are working together or in parallel, most of the time it's one person. Best practice would be one
person. Best practice would be one person. accountable is usually that
person. accountable is usually that person's boss and it literally means they're the one that like counts the KPI for success and so that's accountable.
It should always be only one person. If
there's two people accountable, nobody's accountable. Consulted is who does the
accountable. Consulted is who does the responsible party typically need to have a conversation with to get this done?
And so that means like it's a two-way street, right? I'm going to talk to them
street, right? I'm going to talk to them and I need their input as well. And
informed is who just needs to be informed. one-way conversation who needs
informed. one-way conversation who needs to be informed that this happened.
That's it. So for each task, we have a responsible and accountable consulted and informed. So this would be again the
and informed. So this would be again the the core process starts with an onboarding call. It knows who's
onboarding call. It knows who's responsible. This is all correct. The
responsible. This is all correct. The
KPIs are correct. I didn't have to correct it. We it talks about the
correct it. We it talks about the investment thesis development. So if we again if we compare this, you know, we have this and then we have the same investment thesis which actually part of
this process here. It then activates a parallel track which is these two things. Again, I did not share at all
things. Again, I did not share at all this image or excal with AI. It just
knows on the master prompt and common sense and it basically says okay we have two tracks on market and off market. It
walks through each step A1, A2, right?
A3, A4, a decision point again, racing and KPIs for each and then shows where the tracks converge and what happens for each. Um, and then it came with this
each. Um, and then it came with this sort of key performance metrics, key roles and responsibilities at the end.
And this is just an HTML artifact and it basically represents this entire thing.
The next part after this, I basically said, "Okay, create a guide for the onboarding call, right?" So that's basically this one right here in a separate artifact. Again, normally in an
separate artifact. Again, normally in an onboarding call or an SOP, you just I would interview the person who's already doing it and I would just basically have them walk me through it. We could screen
record a call and it could uh based on the recording I might ask it to I might ask Claude to make an SOP and and that would be a version of it that would take two people maybe 30 minutes. The problem
with that you're going to get what you're currently doing and you may not get what you should be doing. So this I I try to do both right. So one is what are we currently doing and what is you
know what what should we do that basically created this and again all I said here is create a guide for the onboarding call in a separate artifact make it appropriate for a company of acquire size it knows who owns it
basically knows what tools we need to use does the step-by-step process the email so I would then review this with Nathan who who is the one the
responsible party here uh and we would look at you know is there anything in here that we're not currently doing, which I'm sure there is. Is this the SOP we want to use or do we want to go with
something simpler? And we can just keep
something simpler? And we can just keep on going. I can show you how long it
on going. I can show you how long it actually takes. So, if we go back to
actually takes. So, if we go back to this process, so that was the first SOP.
And if we say create a guide for investment thesis here's I'm just doing this live now. Uh
create a guide for investment thesis development a separate artifact. Okay,
so this is done. took about maybe three or four minutes. I think it created a probably too much detail. Uh we'll we'll take a look at and so I'll show you what happens when when that usually happens.
And by the way, I have prompt in here that tells AI how I like to see process made. And this is something that we can
made. And this is something that we can share. So it's actually part of it's
share. So it's actually part of it's part of my style preferences. I have
this one. So we can share this with people. they can get this and put this
people. they can get this and put this into their master prompt as a like a protocol for creating process and teaching AI like the language of process. My my first thing is I'd like
process. My my first thing is I'd like you to be concise with your wording and to always ask me clarifying questions when I ask for something and to answer your own question based on your
knowledge then wait for my confirmation before proceeding. This is so important.
before proceeding. This is so important.
At first I was just asking people ask asking it to ask me questions one at a time, right? Because if it asks you
time, right? Because if it asks you eight questions and overwhelms you with texts, then you end up missing them. And
then I realized what if AI already knows the answers. And so when I asked to then
the answers. And so when I asked to then do this to answer it based on your knowledge and wait for my confirmation, I realized that 95% of the time it got it completely right. Uh, and then the
other benefit of that is you see where there are holes or gaps in the AI's knowledge and you can then clarify and add to your master prompt or edit your your master prompt to address those
gaps. One thing that you could add to
gaps. One thing that you could add to this and actually I have that as my style preferences in chat GBT is to also score your own answer the answer you have based on your knowledge based on
your confidence um out of 10. And that
way you can have an understanding of like when Chad when Chach GPT might be hallucinating or making sort of vague assumptions because it'll say like you know my confidence is five out of 10.
When I ask for advice for you to be a thought partner I want you to be skeptical and conservative in your estimates. Not every idea is a good one.
estimates. Not every idea is a good one.
I want you to challenge when it makes sense. So this is sort of the you know
sense. So this is sort of the you know anti-copant statement. In my business
anti-copant statement. In my business businesses I'm always looking to become a category of one. I'm always looking to sell high ticket products and services the wealthy in the early stages of a business. I always want to have an 8%
business. I always want to have an 8% gross margin. When comparing two or more
gross margin. When comparing two or more options, list the pros and cons of each.
Then ask me if I want you to create a hybrid approach, taking advantage of the pros and mitigating the cons. All of
these I encounter, you know, multiple times a day. These are like my top four like little snippets to just making AI a better thought partner. There's another
couple of things here and we'll share this. So things like KPIs and process
this. So things like KPIs and process have different definitions and different understandings based on different frameworks. It's important to tell
frameworks. It's important to tell claude or chatbt or whichever LLM it's important to tell it what language we want to use and to speak the same language. In my case, I have sort of my
language. In my case, I have sort of my own framework for process which is that that process there's this you know mutually exclusive cumulatively
exhaustive way of looking at process in which we have sort of three different types of pro process. We have sort of the strategic planning or business identity. We have the core processes of
identity. We have the core processes of a business which is you know that loop we've talked about in previous videos and then we have enabling processes. And
so I just explain what these are in to AI so that it understands the way I look at process. So it knows the core
at process. So it knows the core processes are market to lead, lead to sale, sale to delivery, delivery to success, success to lead, success to market and it also understands some of
the pieces within them. Um it also understands what the enabling processes are and it understands what a lot of the strategic planning processes are. So
that's one piece and then the other piece that I keep in here for that's process related is this one. So
functionally understanding SOP levels.
So I basically define what is a core process and I also define what is a blueprint right which is a full breakdown of a of a repeatable business function and then I also talk about and
like what when you create an SOP what should be in the SOP of a blueprint right objective scope triggers etc. Um and then what is a guide and so all of this will include but then you can put
it in your project um or in your system preferences and now you know it will make SOPs the right way. So this is the investment investment thesis development guide again it's following the
instructions that are in those system preferences. So what's the trigger?
preferences. So what's the trigger?
What's the prerequis prerequisite? What
does it apply to? What are the required tools? What's the step-by-step process?
tools? What's the step-by-step process?
Quality standards. So this is very long this piece of it. So it's using investment thesis in a much more exhaustive way than we actually do within our business. So for example, it's trying to understand like the
holding period and the exit strategy for the AE, which is probably unnecessary.
Um, so I I could go back and and tell it or I would go back and say, you know, we don't need section F. We can remove that. I'd say we don't need section F.
that. I'd say we don't need section F.
One little hack which I can we can go over a little bit is you can create a little custom app to house everything.
What I originally did was I actually just had Claude help me make like a notion template, but I think it's even faster if you just have it create like a
React app. React is is basically just uh
React app. React is is basically just uh it's a it's Claude's prefer preferred way of creating little web apps. So,
it's just a it's a programming stacked.
You could just make it in in HTML, which would be a little bit less efficient.
What I imagine it will create is it'll create it should or what I'm looking for it to create is for it to create
basically this page except I can click here and when I click here I then get, you know, this SOP. And so what I can do from there
is I can take and copy that code and I can put it into another platform like Vzero um or like Manis and I can just basically say recreate this app and host
it for me and Claude doesn't really do that well for you right now. It's not
really appropriate to do it in Claude.
And from there I can now I now have basically an app I can share with my team. Again, this is now becoming a lot
team. Again, this is now becoming a lot more sophisticated, but it's a it's a way to share the project in a way that's much more user friendly um and exists outside of Claude. Whoever's watching
knows how to code, you know, they can just host it on like Netlefi or something and that's fine, but I'm just trying to be mindful of people who may not know how to code. I'm seeing right here nothing's happening here, right?
Which sometimes you see. I'm going to screenshot it and say doesn't work.
Yeah. So, it knows what's wrong. it
sees, okay, you just see a plank a blank purple gradient screen. If you don't screenshot it and you say it doesn't work, you're not really giving it any context. And you have to understand that
context. And you have to understand that it doesn't see, you know, your screen.
So, this is the first draft to give you an idea of how it could look. And so, you could imagine in this
look. And so, you could imagine in this chat, I might go and say, "Hey, create it for active listing monitoring."
Right? And then once we get it, we could copy and add it to here. Again, this was just I was just prototyping it here. You
don't necessarily do that here. I I
could bring it to v other tools, replet, vzero, lovable, manis, and I could literally just give it the code, right?
Put it in and say, you know, I want this to be hosted. I want this to have a database, all that other stuff. Um, and
you could turn it into a real app. You
know, I I would say I'm I'm probably top 1% in like AI experimentation and I am still amazed constantly at how much
business is going to change. The more I look at this, the more I see, you know, business really is just going to be a co-pilot to do everything. And there's
going to be a certain amount of work that it takes to set that co-pilot up. I
think businesses can do it today. And I
think all businesses will be doing it in 5 years. there's a window of time and
5 years. there's a window of time and opportunity um to I would say extract a lot of value getting that done quicker
Loading video analysis...