Introduction: From Non-Technical Background to AI Workflows

So I'm not a technical person by any means whatsoever. I have never coded a day in my life. It's just not my skill set.

My background, I started my career as a TV producer, so making TV shows out in Africa, and I ended up transitioning into marketing roles. So a lot of my work is in helping translate complex technical things to the general market so that they can adopt it.

What Mindstone Built: The Rebel Platform

right and that transitions nicely into the work that i do with mindstone which is you know not just helping people um learn and understand how to use the tools in their daily jobs but we also have

this very amazing platform that has been built by our technical team in collaboration with all of our trainers so taking just like what they've been seeing in the real world that everyday non -technical people struggle with in using AI and how do we translate

that into a platform that any company can kind of take and give to all of their staff to be able to create agentic workflows right so what I'm going to be

showing you is how I use this platform that we call rebel to do some of my own marketing work right and build my own agentic workflow so I can just like come into the office and just say hey rebel like do this thing that i we need to get done so i can

accomplish my kpi right so um just to kind of start right one of the things i'm going to show you is how i actually like prepared for this demo right so there's a lot of things that we're doing

to figure out okay how do we use ai in our work right there's a principle principle that we work with internally called building your own factory right so it's everybody's responsibility to figure out like how do i really leverage ai to take what i'm doing to the next level especially as a small team right i'm a solo marketer in this team how do i make sure that i'm using ai effectively to

What You Need for Agentic Workflows (Models, Connectors, and Setup)

do my work so um just for a little bit of um a nerdy background right so what this platform does is first it makes the the setup of everything that you need to do to create these agentic flows easier right so what do you need to create an agentic flow

you kind of need your AI models right whether whatever those models are open -source chat GPT we personally use multiple models via this company called open router right so it's also set up so that like you can easily kind of like like Connect, and if you're a company, your tech team can kind of do the backend work for everybody else.

You need what we call connectors, essentially tools that your businesses use that AI can kind of leverage to be able to do tasks for people, right?

So somebody mentioned CRM, monday .com. For me, we use HubSpot a lot, right? Or, you know, I need to use Google Docs. It needs to be connected to Google Analytics. Whatever you need to be connected to

to kind of help the AI help everybody accomplish their goals the the sorry the application is kind of set up that it's easy to connect right so I can even ask it to help me connect to cause something custom and it will just create it right so how do you take away those barriers that prevent people from just kind of coming in and doing their work with AI how do you take this away that's what

Preparing the Demo: Using Voice and Iteration to Build a Deck

kind of like the platform does so to prepare for today's presentation um i had done a version of this before um for one of our internal um internal demos so i kind of just came in this morning and i was just like okay i'm not super prepared but i know i've done something before so and i had like

this deck that i had created so i came in we tend to use um rebel voice first right because we see that like when you want to transition people from using ai as an answer engine to treating it like a partner the best ways to kind of talk to it right so i came in this morning and i said um hey rebel can you just do a duplicate of this deck that i had done before right and i i gave it some

instructions of things that i want to like give it this name blah blah blah right and then it created a first version right i'm going to do something live so i'm just kind of pulling you through it created a first version and that's where kind of the human expertise comes in right like your domain expertise is in your taste um your sense of like what good looks like and i'm just like

not really a fan give it some more instructions of like hey i want it to be more brand aligned Like you didn't do it right the first time. And eventually we arrived at this version of a deck, right? That is essentially taking you through how we would build agentic workflows, right?

A Simple Framework for Non-Technical AI Use

So as a non -technical person, there are three ways that I think about when I'm thinking about, OK, what do I need to help AI kind of get me through to kind of do this task, right? Right.

One of the things that I have to do a lot in my work is produce a lot of content. Right. Like we are a business. We have to be able to tell other businesses about our stories.

And case studies are one of the ways that we do that. And one of the things that we're also now doing is that how do we like also show up in the market in a higher quality way? Right.

So how can I send a document to someone that is different from somebody else? right how can i show that like i really understand what your pain point is and can give you something that is personalized to you but how can you do that at scale with like 42 people in a room or 100 and like scale that up right so these are some of the ways that i'm using rebel in my own work

Workflow Example: Creating a Customer Case Study

so the first one is um to help me create case studies so one of the really great clients that that we've worked with recently is the Home Depot, right? American company, a lot of employees, and we had the opportunity to sit with their CEO. I did the interview, so all of that data is recorded.

But there's also these little bits and pieces of really interesting things that, they're not captured in an interview, right?

It's because our customer success team, they were sitting with some of their team members and there's people are saying things about like oh how the program experience was that I wasn't there for they might not even remember that I did that right but because we have a platform that has

almost like a collective data collective intelligence of what's going on I can leverage this platform to kind of gather the stories right kind of figure out what is the best angle to tell the story that i want to tell and produce it right so i'm just going to say oh i'm connected

to my headphones yeah that's what happens when you talk a lot so i'm going to disconnect okay

Prompting the System to Gather Context Across Tools

so we are on my mac so hey rebel so i need to create a case study we just finished work with the home depot um i have done the interview with you know the ceo as well as their chief of staff but there's also a lot of information that is probably on slack it's probably in google drive or somewhere else that i don't know there's probably like ways to make the story richer

that i kind of just don't have context of so i would like you to kind of go through everything everything, and run the case study skill, and create a draft of the case study deck for me. 1So as you can see, in the land of prompting, that sucks.

But what you're able to do by the time you start having a system like this is that it starts to learn you. It learns your preferences.

It learns what is Kovo's writing style, what does Kovo mean when she's like, check whatever document right like who are the people in Kova's team that have the relevant information that she might need right so in the back end it's able to kind of

pull all that context for you to be able to get you to your goals faster because at the end of the day I'm not always gonna be able to wanting to connect all the all the things that I need to do to to get this done so I'm just gonna send it and it does take a while because it's doing multiple things and the back end

right so just for a bit of context of some of the things that are going under under the hood it's connected to hop spot notion slack we have an internal platform right so internal platform data that tells it like who was a part of this program etc um it can also do web search so sometimes it will do web search to try to what is going on with home depot right now what is the larger context of

you know them training their ai staff that can help me write the story angle and um we also use gamma sometimes to create these decks um i think you mentioned gamma earlier so it's like a ai presentation app so sometimes you use gamma i personally prefer ai using html just because like you get nicer presentations but gamma is also great and um essentially i i mentioned

Skills: Encoding How Your Team Works

I asked it to use a skill earlier to do this, right? So skills are, if you've used Claude, you might be familiar with skills. They're essentially documents that encode how you like to work, right?

Everybody has a preference for how they like their Dexta and how they like email, and sometimes it's company -specific as well. Skills are how you kind of encode that for the AI to use.

And the great thing about Rebel is that because it is almost like a shared space for everyone I can also take a skill from someone else and modify it or use it as is right so for example our product designer has created a skill for everybody to be able to have

beautiful designs I can tell my my agent to kind of go and use that or even like combine it with other skills to be able to produce something right so it has an interview prep skill that it's if somebody else on the team hasn't done an interview before and they actually need to produce this case study they can go ahead and do that it has a case study

creation skill where it's going to figure out the story angles etc and actually create the deck and then it will create the publication ready output right so those are all the things going on in the backend and we can check on what's going on and you can kind of see right it's it's taking some time because it's it's not just going ahead to try to give me an answer right it's planning

it's thinking through okay what is this skill that i need to use right i need to go read about that skill right that's been encoded um and it's super interesting because the first version of this i had to kind of write the skill which is just like every time i have to write a skill but as kind of the platform has developed now i can just like do work and say like oh with all the feedback that i'm giving you go and create a skill and and let's use that moving forward right

so it's checked what the skill is it's gone ahead to read my case study writing guide because you got to write how i like um and then see it's going to read all the existing well maybe this is kind of cheating because we've kind of done the home depot case study before but But, and then it's confirmed, you know, the scope, like, okay, what is the scope of this project? What do I need to deliver?

Under the Hood: Planning and Delegating to Multiple Agents

It set a mission. And what is also interesting about this is that what Rebel is also able to do is that it's sending out different agents to do this. So it's not one agent trying to do everything at the same time, right?

It's sending out different little AI agents that are specialized, that are gonna go do the research, that are going to go, you know, okay, be like, okay, we're the ones editing, etc. And now it's, it's delegating to those different knowledge workers across. Yeah, we're gonna wait for that.

But what I can also do is go ahead and show a version of this that was done if it's taking a little bit of time okay you know what let's

Parallel Workflows: Email Triage and Slack Updates While the Deck Builds

let's do something fun I'm gonna do another case just to see what we can get out of it right so I'm also going to ask rebel and this is kind of similar to how you end up working when you have this agentic kind of workflows is that you can have something right in the background and then go to do something

else right so I can say could you check my last three emails and give me an idea of the most important ones that I haven't responded to that I need to respond to and draft a response in my inbox all right and the great thing is

that i can also move on to something else all right i can say okay send a message to everybody on slack about the amazing time that we're having at the newcastle event let them know how many people are registered and in the room and um what we have talked about today we'll see if they can figure that out because it wasn't on for some of the stuff but so now i have it running three

different things in parallel so we can kind of check what's going on with the first thing still working again this is a live demo so you can see it working in the background but i can also

Review, Comment, and Regenerate: Human Taste in the Loop

show you some past conversations that I've also had with this. Let's just say case study. So this is a similar example of that conversation.

So in this one I just kind of asked it use it to create a Home Depot case study. It kind of went ahead and drafted an angle for me based on all those skills, right?

How 40 Home Home Depot execs on lock, blah, blah, blah. It gave me the at a glance. And at this stage, it's basically asking me

for me to review it, right? So here's a draft of the full case study based on all those skills that I mentioned, right? It's staged for my approval.

And one of the cool things that I can do on Rebel is that I can just make specific comments, right? I can say, this sucks. this sucks make it sound more exciting right i add that and i can send it

to a current conversation let's just send it to a current conversation and it's going to figure out how to make that change right so it's also created i mentioned before that we have a lot of our brand guidelines skills etc it created this brand aligned HTML which it's not finding but this is an old conversation

so maybe we can check oh there we go this is the data so it's actually created this is a new conversation it's created a draft for me here and this MD file I'm gonna say link the MD file but it has something that I can review right

it's told me how story with why support it's given me key additional context that it pulled by itself because again I mentioned connectors it's connected to different things and it can pull that context it pulls something from a a meeting that we had.

I mentioned Hofstra earlier, I pulled from the CRM, it pulled from Google, etc. And this is that MD file.

Again, bugs, these other things we live with. Okay, so this is the MD file.

So as you can see, it's kind of similar to kind of what we produced before, right. So it's given me what the cover what the slides will look like.

And then I can just tell it like if i like it i can just say go ahead let me just say okay rebel this is a good draft you can go ahead and create um the html slides from it and we can go ahead

okay what else did we okay and sometimes rebel will ask you follow -up questions because so that you can do a better job right what should i use for the room count sometimes i would be like that's a silly question you should know that actually go look for that information so I can say go find the info from the admin back end or sometimes you can be

nice you know so now I actually forgot I didn't tell it what channel to post it in so he's asking me like what slack channel should I post it in I can say let's do general any must include highlights any that we should put maybe great energy exactly we'll tell it to say great energy and off it goes now

Directing the Day: Staying in Flow While AI Executes

we'll go to the email triage and this is literally what my workday looks like right just like telling AI to do stuff and then go in between conversation I'm essentially almost like taking on a director role and just like managing different workflows essentially priority ranking okay helps by telling me to

check my 2fa I haven't checked my emails is that bright okay and then it's drafted something in Gmail okay I'm just to clarify is this in my Gmail drafts already and if it's not put it in again you can see that like avert I would not be able to just

like say stuff like this like maybe like two years ago the system itself needed a lot more prompting and stuff but the models have gotten better right like we're able to connect to a lot more things so we don't have to do some of that anymore I can just kind of be in a flow state of of work, if you will.

How much time do I have left? Okay.

Let's go on to check our Home Depot case study. Okay, it's taking some time. It's drafted the markdown file.

It's checking some of the approved decks that we've done before, again, so that it's meeting my guidelines.

Email triage, let's see. Oh, it said it's in my inbox, okay. You guys keep my email information secret, if you will. So we're going to check my inbox, is it in my draft?

And it is.

Why It Works: Tools, Context, and Team Knowledge

So that's kind of the power of these kind of agentic workflows is that it kind of just lets you come and be in flow of the work that you want to do. And it's taking very simple concepts of AI needs tools to be connected so that it can not just have context, but actually use those tools to deliver work for you.

it needs to kind of know you not just as an individual but as a team right like you can see it asked me like what channel I even forgot of like I had to give it like what channel to send it to right but you already knows that so it can ask you to be able to help you do that work we're not going to send this

email just okay Newcastle admin shows 120 registered okay yeah we actually haven't told it doesn't have it in the back end how many people actually showed up but we'll just tell the team 120 people how about that okay it said it's done the deck let's see what it did and there we go it's created a 19 slide deck

and like I'm of it I would still tell it no but for our first draft you know like Like it's created something that, you know, if I was having pressure from my boss to say like, girl, we need something right now. I can be like, here you go. It's our first draft.

We can perfect it later. Just check the information. Right.

So this is like there's no way that I was working like this, like two years ago. I remember just being like, oh, I have to do a deck. This is like horrible. but now I can just kind of all I have to do is like try to make sure that I get

the elements of the story for rebel to be able to help me do this work all right so I'm gonna do a time check just to see if we can okay I'm making a time okay like the fact that it created the deck in time I think we're doing good

okay so checking slack but in just this presentation I don't know how many many minutes I've taken, but we've done an email triage and actually drafted an email to, I don't know, someone random from my inbox.

We've created a full HTML deck from Rebel, just kind of gathering all those stories and putting this together.

And maybe it'll be able to give me a shout out so I can look good to my team on Slack. We'll see. But if we don't have enough time, it's going to do

Governance: Permissions, Privacy, and Safe Sharing

that in the background but i think what's also important for me to highlight is the importance of permissions right and safety when you work in an organization some of the things i talked earlier about it having context some of the context that let's say some of the execs have i shouldn't have right because some of that information is sensitive the great thing about

this kind of system is that you can start to give people spaces hold on let me go to the library right so I mentioned skills earlier so you can see that there's skills that are in my own private space because like okay tricks of my trade I'm not I'm kidding but some of it is not relevant to everybody on the team and it organizes it by itself right some of it is in our general space you can see customer success community and those things are the things that you you can start to build collective intelligence in an organization and share those skills across the board.

Memory, there are things that like will be safe to my own kind of personal memory or space that are not in the general space. And those kinds of things are important when you start thinking about security and stuff.

And then just in terms of, excuse me, permissions. Everybody has kind of, every company is different, different, every person is different. There are settings that where I can choose how I want Rebel to ask me for permissions to do stuff, right?

So, in the beginning, I was a little bit like, don't just send a Slack message, ask me first before you send that Slack message, right? So, 1there are different kind of permission settings that give you the control on how you want the different AI agents and Rebel itself to kind of act

Granular Controls: When Rebel Should Ask Before Acting

as it goes on let's check that and see it's okay let's see did it send it to general let's see general lots of lots of slack channels and there we go it's in our message to slack