Writing procedures has always been the part of business systemisation that nobody wants to do. It's time-consuming, it's dry, and most of the time what you end up with is a document that's too vague for anyone to actually follow.

Hi, I'm Mike Fox, host of this podcast, "Lone Wolf Unleashed."

In this episode, I walk you through the Claude SOP skill I've built inside my Obsidian workspace — a five-stage framework that takes raw inputs (a description, a transcript, some rough notes) and produces a complete, high-fidelity standard operating procedure that a delegate can follow without asking you a single question.

I cover the five workflow stages: scoping the task, gathering raw materials, building the procedure, running quality checks, and presenting the document for review. I also walk through a real example — the webinar setup procedure I built for my wife, who's just joined the business — and share how the same approach produced a 148-page procedure for a client that was 85% accurate straight out of the gate.

If you're ready to start handing work off to other people, this is where you start.

Visit lonewolfunleashed.com/resources to find more useful help.

Mentioned in this episode:

You might also like...

Check out the "Websites Made Simple" podcast with Holly Christie at https://websitesmadesimple.co.uk/

This podcast is part of the Podknows Podcasting ICN Network

Good day. My name's Mike and you are listening to Lone Wolf Unleashed.

And today I'm gonna show you that it's not so difficult to create procedures that people can follow.

We've talked a lot about procedures on this podcast before,

and yes, it is one of those really boring, laborious things that has to be done when you're systemising your business,

but not so much anymore.

And I'm gonna explain to you in this episode about how I'm using Claude to produce run sheets for people in my team.

And that will help me to set up things more regularly, such as webinars,

and how that can save me time in the long run because I'm not the one doing it.

And it provides the person doing the job certainty about how they can get the job done — the right way, the first time.

And this is something that — I was at a business lunch yesterday

and a lot of people were talking about one of the pain points they have this year,

particularly around people and process,

and that, you know, their systems just aren't up to scratch for the size that their businesses are.

And this is a very common thing.

And yes, building up systems does take skill, it does take specialists to be able to do that,

but it doesn't have to take that much time to do anymore.

So I'm gonna show you now one of the skills that I've built into Claude

that helps me build procedures — basically produces that structure.

I believe that this is already on my website, so you can go and check that out.

You can find that on my website: lonewolfunleashed.com/four- p-procedure.

If you can't find it there, you'll be able to find it on my resources page on the website,

along with all the other resources that are there.

So I'm gonna open up my Obsidian workspace.

If you haven't seen Obsidian, it's basically where I have Claude interfacing with all the markdown files.

And Obsidian is just a place where I manage and review that,

and then I interact with Claude through VS Code to make changes and updates and all that sort of stuff.

Or I can just update them myself if they're really small within Obsidian itself.

This is where I'm managing a lot of my ecosystem now,

and it is an absolute game changer in terms of my efficiency in the things that I'm doing.

But here we have a standard operating procedure.

And the skill basically outlines how the procedure comes together.

And we're inputting a task description, or a recording with the transcript, or some process notes,

or an existing rough procedure.

And the idea here is that these things are gonna have enough context in them for Claude to be able to interact with this

and produce something that's of high fidelity.

And what I mean by high fidelity — it is of a high quality that people can use.

And what's the output? It's a single markdown document containing a complete SOP —

which is a standard operating procedure — with all sections filled in,

delivered as text for review before saving to file.

And then I have a section for accepted arguments,

where the user may provide arguments describing the task to document.

this is a client onboarding process for a VA,

or this is the weekly invoicing process, et cetera.

And then there's the workflow.

So work through these stages in order.

The quality of the procedure depends on getting the inputs right before writing begins.

So what we're trying to do here is set up the agent or the skill

to be able to act in a specific way that's gonna maximise the quality of the output that it gives us.

We don't just wanna put in with no context,

"Hey, write me a procedure about how to use Xero." That's not good enough.

It's not gonna give you what you need,

particularly if you're going to want to hand this document to someone else to do.

If you already have context around it, you know how to use the system.

We want to assume that for the most part,

the person that you're gonna be handing to will not know a whole lot about how to do this.

The whole idea about building systems — and McDonald's have mastered this —

is we wanna be able to give the task to the person with the least amount of capability,

but who can still get the job done.

Why is that? It's because they are a cheaper resource.

So you're paying less, but they're still producing the output that is appropriate.

Keep that in mind as we go along.

So then we wanted to clarify the scope. This is Stage Zero.

Before generating anything, we need to establish: what are we documenting, what is the task?

Can we name it in three to five words?

We've talked about naming convention before for tasks — typically verb, noun, because we're doing something.

who will do it? Who are we handing it to?

Is it a VA, is it a contractor, is it a staff member? What's their role?

Is it your future self?

Does this task require judgment or just knowledge?

We've talked about this in a prior episode — there are tasks with high judgment

where there's a lot of decision logic that might go into making certain decisions as a task is happening.

And we wanna make sure that we provide enough context

that if a decision does need to be made while doing a task,

the person is equipped to do that as well, with those different conditions.

Then we have Stage One, which is gathering the raw material.

What are those inputs that we're gonna use?

Is it a prior procedure, or a verbal description?

Is it a video of you going through the task that's been transcribed?

We want to be able to grab those.

What systems are we using? What systems are involved in doing this task?

What are the typical steps, the URLs that we might need,

the menu paths, the commands used?

The techniques — not just actions.

Not just "click save", but: "click save, then wait for confirmation of the confirmation banner before proceeding."

It's that type of instruction that we're getting down to there.

Then we're gonna have a definition of done.

So how do we know that the task is being completed?

And who are we handing over to, if anyone?

I covered this a couple of episodes ago — communication is performance, right?

If we don't tell the next person that it's their turn to do something,

you didn't actually do the task.

Because they didn't know it's their turn to do the thing.

And they need to do that for a process to be able to operate really well.

Then there's any known edge cases.

When might things go wrong, when might there be an exception that happens?

We can think through that.

Then we have Stage Two, which is building out the procedure.

And in here there's a template structure about how the document's laid out.

We've got the task name, we have some of the fields that are in the properties —

which include the version number, the owner, last updated, et cetera.

And then we have the trigger: when does this task kick off?

The inputs, prerequisites, required information —

what steps we go through, what decision tables there might be so we can make those decisions,

what the definition of done is, and then what the handover chain might look like.

Any frequently asked questions, and then some revision history.

And this is updateable as well,

so if there's a frequently asked question that's asked regularly, you can come back and add it to the document.

Then we have Stage Three, which is review and sharpen.

And I put a review and sharpen stage in a lot of my Claude skills.

And that's because — you know, what first document did you ever write that was right first time?

Yeah. Exactly. Probably none.

And that's just because you write it out, you review it, and then you make tweaks.

We're getting Claude to do this the same in this sense.

So before presenting the document, run these quality checks.

There's the delegation test.

Could someone follow this procedure tomorrow without asking you a single question?

Read each step as if you've never done this task.

Where would you hesitate? Where would you need to ask?

There's the observable outcome check.

Every definition of done item must have something a delegate can see, count, or verify.

Replace anything subjective — "data is clean" —

with something observable: "no cells in column B are empty."

There's the action verbs check.

click, open, navigate to, check, verify, enter.

Not "the system will..." — we wanna tell the delegate what to do, not what happens.

Then there's decision coverage.

If there's a decision table, check that it includes an escalation owner row

for situations that are outside of the delegate's authority.

And then there might be missing steps.

Count the micro decisions. A simple task often has ten to fifteen or more steps.

If your procedure has fewer than five steps for anything beyond a trivial task,

you've likely skipped over the obvious ones that trip up delegates.

Then Stage Four is present for review.

We want to produce it as a markdown file so that I can review it — or in this case, you can review it.

And then after the document, add a brief editorial note covering what gaps are flagged —

any extra to-dos, the assumptions that have been made, any judgment calls, and any recommended next steps.

And it was really funny because at the moment, AI is being sort of criticised about

always wanting to continue the conversation, right?

"Would you like me to do this? Would you like me to do that? Would you like to explore this further?"

And that can get annoying.

But in work like this, there's always gonna be updates and adjustments to make.

So I'm not really so much concerned about that.

What I wanna make sure is that when it does that,

the next thing it's asking about is really going to be helpful to me.

It's not just a question for the sake of continuing the conversation —

it's a question that's going to spark additional value.

So that's what we want.

Then there's Stage Five, which is save.

So it's going to save into the dedicated folder.

And then there's what this skill does not do —

it does not execute the procedure or automate the task. It's just a document.

I'm not getting the agent to do the thing.

It's not what I want it to do.

If it's a procedure about how to use Asana,

what I don't want is for Claude to then interact through the MCP server to create tasks.

I don't want that. This is literally just a document creation tool.

We're not creating extra training materials, checklists, or process maps — anything like that.

It's just a procedure documentation tool only.

Do not fabricate steps, tools, or systems.

Do not produce HTML slide decks.

I have separate skills to do those specific things.

And do not replace the testing step —

the procedure must still be tested with a real delegate.

So that, in essence, is my procedure skill.

Now, I use this — this week my wife has joined the business, which is very exciting.

And I thought, oh, I've really had a lot of trouble having enough time to be able to set up some of my events.

March — I've just completely skipped the event that I had planned

because I've just not had the time spare.

I've been really, really busy on project work, which is great.

But yeah, I really do want to be able to run more of these events,

and so setting up that is really part of this thing.

So what I did is I said, "Look, I want to be able to set up these webinars in GoHighLevel"

and for me, GoHighLevel is done through another company called Growth Monster.

And so I've basically said, "Look, I wanna be able to produce these webinars."

"Walk me through what you need to be able to build this procedure."

And my interaction with Claude was a very interesting one

because I discovered that the tools that I currently have

don't meet the requirements for managing webinars properly.

So it meant that I had to go out and find a tool.

Thanks to my producer Neal — he had recommended Boomcaster to other people on LinkedIn.

And so I'm now on a trial for Boomcaster, which is great.

And so I basically fed all of this information into Claude and said,

"Can you please produce a procedure on how to set up webinars in GoHighLevel using these tools?"

And what it's done is it has produced an entire procedure —

an entire run sheet that my wife can now go through and use to set this up.

So when we're in testing now, we're gonna see if she can do this based on this.

But I have very high confidence that this is gonna be right

because I've done this for another customer of mine

where I did a very, very detailed setup procedure

on how to set certain workflows up in Power Automate within a SharePoint ecosystem.

And that was about 148 pages, if I remember correctly.

And it was 85% correct.

Which is — I mean, that's astounding to have that much content.

And there were just a few nuances there where the information that was missing was actually really easy to find.

So it didn't take us too much longer to find that extra information

to be able to get on and get the job done — which is amazing.

And following a procedure like that in a novel setup for a company,

it would've probably taken us about three weeks to do.

And instead we were able to set up the run sheet and deliver that part of the project within a few days —

which is a really, really great outcome.

Particularly great for the client, because they get to reap those benefits earlier.

So back to this — we've got the procedure.

How this works: this webinar setup uses three platforms —

Boomcaster, YouTube Live, and Growth Monster which is GoHighLevel —

the marketing layer which has the registration page, email, SMS reminders, CRM tagging, et cetera.

Then we have the trigger: the trigger is me requesting a new live webinar to be set up.

And the prerequisites are: we have login access to GoHighLevel, Boomcaster, and YouTube,

and we've tested the login to all of those to make sure the credentials work.

Required information from me is the webinar title, date and time, expected duration, speaker names and bios.

A webinar description — there's a whole list here

that the person needs to make sure they ask me for before they continue,

because these are the required inputs.

when we go to do a process map, every task on there is a separate procedure.

And what a process is managing ultimately is inputs and outputs.

And so what this is saying here is: these are the inputs I'm gonna need to be able to complete this task.

If you don't have the inputs, you can't do this properly.

If you don't have the Boomcaster link, we can't broadcast.

And if we can't broadcast properly, we're not really hosting a webinar, are we?

So all those prerequisites are there.

If I have not already defined that, then they know upfront that they need to ask me for that information.

So then I have the steps.

schedule the YouTube Live event.

So you go log into YouTube Studio, create a schedule for later, fill in the stream details,

set the schedule, click create stream,

copy the YouTube Live URL into a notepad — probably Asana in this case —

copy the stream key, do not close YouTube Studio.

Then we have Part B, which is create the registration form in GoHighLevel.

So we'll navigate to forms, create a form, name it, add the form fields —

first name, last name, et cetera.

Set up the form, click save.

Part C, et cetera, to create the webinar funnel. Part D: customise the pages. Part E: build the automation workflow.

A lot of these you might say, well this is a lot, Mike. This is actually a process in and of itself.

And in a lot of cases I would agree with you.

But for the most part, I would want someone to sit down and do all of this in one go.

I don't want someone going out to YouTube to set up the stream

and then just parking it for the day and coming back tomorrow to do the rest.

No — because I have to keep it open anyway.

What I want is for the webinar and this entire thing to be set up in one go.

Now, it might take a couple of hours or a few hours to do all of this.

And you know what? I'm okay with that.

What I really want — the outcome of this — is that the webinar is set up right.

What it's done here is — my words on here are four and a half thousand words.

Four and a half thousand word procedure — which is a lot.

And I've criticised customers before about the size of their procedures being over three thousand words.

Sometimes it's appropriate.

Would I pull some of this out? Maybe.

Maybe testing can happen at a separate time. Maybe there's build webinar and there's test webinar.

So you might be able to separate out into different ones.

But ultimately what I've got here — I can do that easily.

I can rip that content out now that it already exists, and pull it out into a separate procedure.

So what do we have? We have a procedure skill that has defined the structure of a procedure.

The things that are needed to create a specific procedure.

And then what's happened is we've given it that content

and it's produced a very high fidelity procedure

that someone can follow through, have what they need to do the job well,

know how to escalate and what to escalate for,

and what decisions they need to make, and what steps they need to take.

And it's all there.

This is how we create an ecosystem where people are set up to succeed.

If you're gonna have people in your business, then procedures are gonna be necessary.

And using AI, using these Claude skills to be able to do this,

is gonna help you drastically set up an ecosystem very quickly

to be able to hand off work to other people.

All you need to be able to do is articulate it in a way that Claude can pick up,

and then format it all for you

and investigate how to do those other extra steps that you may have missed.

So that's it. I'm gonna leave that with you today.

Head over to my website: lonewolfunleashed.com/resources.

You can find a whole bunch of stuff in there about how to get started with AI and some other stuff.

How not to be fooled by the AI bros.

And I wanna thank you so much for joining me today.

You could have been doing so many other things.

You're listening to a million other podcasts,

but you decided to hang out with me and learn how you can use Claude

to start to delegate out your work to other people

in a fast and easy manner that is gonna get the outcome that you need.

Thank you so much for listening, and I'll see you next week.