← School Business Insider
Loading summary
Transcript113 lines
- [00:02]
A
You're listening to asbo international's school business insider. I'm your host, john brucato. Each week on School Business Insider, I sit down with school business officials and industry experts from around the world to share their stories and explore the topics that matter most to you. Find out what it means to be a school business official and get your insider pass on all things school business. Hello everyone and welcome back to School Business Insider. Artificial intelligence continues to move quickly from curiosity to real world tool in schools and the public sector, but adoption hasn't been simple, uniform or without friction. In today's episode, I'm joined by Aziz Agaev and Jamal Amanova of Flow List to talk about what AI adoption actually looks like on the ground in education and public sector organizations. We'll explore what districts are really using AI for where the challenges and concerns still exist, what's working well, and how frameworks like TADA can help leaders move from experimentation to confident, responsible use. We'll also dig into beginner and advanced use cases, quality control, privacy and security considerations, and recent updates around AI tools in K12, including managed workspaces. Aziz, Jamal, welcome back to the podcast. Happy to have you both.
- [01:32]
B
Hello. Hello.
- [01:33]
A
Hey Aziz, you're no stranger here, so I know everybody. You're like a household name in School Business Insider. But Jamal, before we jump into to everything, why don't you give everyone a little bit of your background and what brings you here today?
- [01:48]
C
Sure. Well, thank you so much John for having us here. I'm super excited to be part of this conversation. Well, you know how organizations and institutions come together to solve real problems and serve people through products, solutions and offerings they have. So I partner with them to create those solutions and turn them into meaningful experiences for people and make sure that we bring positive change to this world. And that's why I am here joining my forces with Aziz. I have three reasons for that. One is technology. I love what I do and so far I partnered with Fortune 50 companies like General Motors, JP Morgan Chase, Cigna, Walmart, and currently serving as a designer within the IBM Red Hat to be more specific. And I learned a lot, I tested a lot and I have some great, great best practices that I wanted to share. The second is education. I do have background in education. I used to teach art in high school AP courses and so on and so forth. I do know how challenging it can be and how educators sacrifice a lot of their time and passion and patience that serve for the next shaping next generation. So I want to bring technology and help them to Adapt them in the best possible way so they can continue their passion in a way that they can amplify their skills, serve and shape people around them. Last but not least, I'm here because Aziz is very annoying. He has been messaging and guiding me and I know, I know. And I was like, okay, for years he has been pinging me, messaging me. So I'm going to not, I'm going to stop dodging this guy and see what he has so far. So last year I attended his workshop. It was very impressive and I saw firsthand how people were excited to learn more about AI and especially within education. So I decided to join his company.
- [03:50]
B
Awesome.
- [03:50]
A
Well, wonderful. I'm so glad to have you on. So let's zoom out a little bit. Let's talk about really what the big picture is with AI. Can either of you just really, how would you describe the current state of AI adoption and education in the public sector right now? I mean, Aziz, you know, you've been doing workshops all over the country. You and I have partnered up on a few workshops and you know, I've seen some varying use cases from people who are really comfortable with AI and the platform use to those that just kind of know it is out there and have trying to experiment with it. So what are you seeing throughout all the trainings and the workshops and everything that you've been a part of? What's the big picture right now?
- [04:29]
B
So I'm thinking that it's getting better and better. When I go to workshops, I do have, I do see a lot of skills, skepticism and people are saying how AI hallucinates and they cannot trust AI, they cannot use AI and because of that, they just quit. And I see some people would say, well, I used it about two years ago and it didn't do the job for me. And since then, I'm not using it. However, in. Not in two years, in two days, things can change drastically and suddenly you have a new model that is a lot smarter, that is a lot better responsing model that you can take it and use it. I compare it to, I did code writing, you know, the vibe code that everyone is talking about. And six months ago was completely different. Six months after, it's doing a lot better job at code writing, you know, taking your ideas and turning it into proof of concept. And you, you did that too, right? You did a few examples of wipe coding. Back then it was pretty all over the place. And now it's still impressive though.
- [05:52]
A
I mean, you know, not somebody who has done coding in a Long time, it was still pretty impressive. So the fact that it got even better than that is. Is pretty, you know, exponential in terms of progress, right?
- [06:04]
B
Exactly. And when you look at the math calculation before, like I would say a year ago, it would just go crazy with math. It wouldn't calculate correctly. It would have a lot of problems, a lot of issues. That has changed, and it's changing every single day. Every single model that is introduced is. Is drastically different than previous model. Take Gemini. Gemini completely changed when Gemini 3 was introduced. It is extremely fast. And it's accurate. It's fast and accurate because usually if AI is fast, it's not accurate or it hallucinates more. If it's slower, then it gives you a better answer. But Gemini 3 changed that conception, and now it's changing the world.
- [06:55]
A
Well, it's funny, I was actually just speaking to a friend and colleague of mine earlier today, and we were talking about, you know, just the spectrum of people's experience, school, business, officials, experience using AI. And I was saying how I use it really on a daily basis now. From anything from payroll analysis to helping drafting memos and emails, really, it's just kind of been woven into the fabric of my everyday work life. But I was telling him how it wasn't always that way. And when ChatGPT kind of hit the streets in 2022, I remember signing up for an account after hearing about it for weeks and logging in and using. I'm like, oh, it's just kind of like a cool chatbot. And, you know, it was pretty advanced back then, but when you think back just to 2022 now in January of 26, it's almost like a completely different experience. I think the way you can leverage it now versus when it originally came out, and it's not such a cool feature novelty thing, it really has been kind of transformed into an assistant to where you can leverage it and save so much time in your workflow. Are you seeing kind of that sentiment creep in from those that have maybe been early adopters and skeptics to where they are now?
- [08:05]
B
It depends if they continued, if they didn't continue, and if they saw what you saw and said, it's just a toy that I don't want to play with. But there are people who actually continued, you know, using on a daily basis or on a weekly basis as it improves. Yes, you can see those business officials that would use it on a daily basis. They would do analysis, they would do presentation, screening. You know, I would go to the board presentation, and I want to get your feedback as a parent and see tell me if anything is wrong with my presentation kind of approach. However, there are also a group of people who think they know, but they don't. And based on their knowledge they say AI will not do good for me. And that group is really hard to convince because my mission is when I go to trainings, I try to explain that the response quality will be based on the question quality. The better you ask, the better it gets. And people don't accept it. People sometimes they just say, oh, I asked this question and it was irrelevant. It gave me a wrong information. But you didn't specifically said go and search for the current guideline or current law or current information. You, you just ask the question. A very generic, very vague question. So it depends on the question you ask, the response you get. And that's what I'm seeing right now.
- [09:46]
C
And I want to add something onto that as is so so far when you look at industry, healthcare, finance, automotive industry, they have been adapting AI for decades. They have been using it for claim adjudication, for speeding up the processes, bringing efficiencies and so on. And so for now they're moving to agentic AI level four, level five. Now you may not even noticing that you actually using AI while you're driving or making purchasing decisions, or when you look at the reviews and saying oh this is the good thing, or when you Google something and choose to go with one source over another. They are influencing a lot. They are influencing at scale actually. And today my biggest concern is how social media is influencing a next generation. They are adopting AI and based on research, 80% of teenagers already using AI and they are using for many reasons. They sometimes cheat, of course, but sometimes they do. They create very meaningful and impressive media. So I believe it's time for education to not only adapt but also bring literacy to next generation. Because today we are, as Aziz mentioned, facing some difficulties in convincing leaders with an education to adopt AI because they are skeptical about certain things. But if they are not using it or adapting it, or role modeling how and when to use it, next gen will lead the way. And without having that safe guardrails in place within education where they actually start exploring, we may face some consequences.
- [11:41]
A
One of the biggest concerns we had in my school district when AI really started gaining popularity in education and with with students was surrounding our ELA and our English department in terms of our kids just going to now put prompts into AI and spit out essays and things like and I, I believe I've talked About this on the podcast prior, I think it was, maybe even it was with you, Aziz. But one of the biggest shocks to me was that our, our ELA department was all in on using AI and they weren't really concerned about the, the cheating and processing all of that because they deal with that every single day. Anyways. What really I didn't come to understand until I sat down and spoke with them was that, well, we know our students voice, we know what they write like, we know what they sound like. So when we get essays and assignments back that are using vocabulary words that are kind of close but not typically used all the time, it's a, it's kind of throws up a yellow flag. And so when you take it a step further and then you kind of probe the student and ask them, okay, what does this really mean? What were you talking about? And they kind of have this blank look on their face because all they did was have AI spit out a two page paper. That still is very similar to your traditional copying notes and cheating and things like that. At least is how, how we're approaching. Are you seeing that educational institutions and schools are looking at AI as a completely different force to be reckoned with in terms of integrity and work or are they approaching it kind of similar to where I just explained? It's just kind of a different way of having to check kids work.
- [13:19]
C
To me, I think we need to adapt. There's no way to escape this tool, right? And as long as they solve problems and they bring their authentic voices, let them use it, let them explore it, but teach how to use them, how to draft the initial essays, for instance, prompt in a way that it brings their voices and make sure that they can explain and articulate their thoughts. After all, I think we do need to adapt. And today I don't see that they are just unfortunately judging kids for using AI instead of helping them to adapt in a secure manner, in a creative manner that would help them to improve their critical thinking, creative approach in creating anything. I think we do need to kind of step back and think about new generation, new tools and how we want to adapt and level up in this game. Because education, to me it's still traditional. Same classrooms, same curriculums and same teachers with same lesson plan. So how about that, right?
- [14:34]
A
Well, you know, I will say there is a part of me that is concerned that the original, a part of the original voice could be lost. And let's step out of the classroom a little bit. Just in everyday business operations, I'm getting pretty good at Noticing when people are just copy and pasting AI and sending me an email. And part of me wonders, are communications over email just going to be AI talking to AI? Because if someone's sending me an AI response and I respond with one back, who's actually doing the communicating here? So, you know, I appreciate what you're saying, Jamal, and I think there really needs to be some education or guardrails on how to effectively use it. Because I don't want to be in a world where I'm just copying, pasting from an AI prompt to communicate. I think there still needs to be some preservation of original thought and original voice.
- [15:20]
C
And that's where writing comes along. Right. But for email. Do you read your emails, John, to be honest?
- [15:26]
A
Well, I actually have AI somewhere.
- [15:29]
C
Here we go, Here we go. So why, why, why are we discussing. Of course emails are going to be generated, right?
- [15:35]
A
I mean, if it's something very detailed, when you're working with like student issues and things like that, of course. But if it's just kind of boilerplate transmission of data, then yes, rather than having it me sit there and read an entire page worth of something, if I just need key bullet points, that's where I think I can really leverage AI. But if it's something specific to, you know, that's really detail oriented, of course I'm going to read the entirety of the course.
- [16:01]
C
So you're in control, you see, you know when and how to use it. And this is what we're looking for.
- [16:07]
A
Right, Right. So let's talk about some of the common challenges and concerns you're hearing kind of out in the school and public sector from leaders and those that could potentially be using AI, but maybe they're slow to adopt. You know, Aziz, like I said, you're kind of all over the place in terms of trainings and you've seen all walks of adoption. What are kind of the common threads that you're seeing in terms of people being a little bit slow to adopt? I mean, I think AI has only really been out there for three years. Ish. In terms of people really using it, but it has changed exponentially. So it's not kind of a one for one. It's evolving kind of how we do. It's a completely different product than it was just a couple months ago. What are those common threads in terms of the slow to adopt and challenges you're seeing?
- [16:56]
B
So one of the biggest challenge, I would say is the security concern. People are very concerned to share data with AI and are afraid of the PII leaks and giving information that they shouldn't give to AI. However, that actually brings up the really good conversation about ChatGPT for K12 or Gemini's account for education. These platforms, these subscription models, they have guardrails in place. ChatGPT K12 will create a workspace and all of the data that is shared with that ChatGPT will stay in that workspace and will not leave the workspace to train the models. And sometimes what I see, educational institutions, they're not aware of this. What I tell every education educational organization that it's to our benefit to create that workspace and invite all of our teachers to use that. Because if we don't, everyone has ChatGPT account on a side and they use the ChatGPT account on a side for school purposes. And now we cannot control if they use their personal accounts. We cannot control what they share and what they don't share with ChatGPT. So our responsibility is to create that safe environment where our leaders and teachers and our educators can use AI models safely. Gemini is the same thing for Google districts. If you have Gemini, then it will be safe to use Gemini because it's educational account and they are aware of it and they do not use any of the data data to train their models. So I think one of the biggest issues, biggest concerns people have would be security. The other one I see is hallucination. People say, I don't trust AI. I don't know if they if the information that it provides is correct or not. And that's where the human in the loop comes in, as Jamal usually mentions, that human in the loop being very, very important. If we take ourselves out of the picture, then as you said, it's a copy and paste. And if you didn't read it, that doesn't mean anything to me. So kind of using AI to shape our ideas, instead of using AI to give us ideas. I look at it as an assistant. If you use it as an assistant, then it will do a lot more for you. But if you treat it as your director or supervisor, then you will lose track and you will go in wrong directions because we need to direct it right now, at least at this point, it cannot direct us. Maybe in the future, yes it will.
- [19:54]
A
But for now, we'll have an episode on artificial general intelligence and artificial superintelligence, but for a later conversation.
- [20:02]
B
Exactly. Yes.
- [20:05]
A
So I have a question for you, and I wonder if you're seeing this crop up. This is one of my frustrations. You're talking about one of my biggest Concerns was privacy and PI. I saw how much I could leverage AI, but I did not want to expose sensitive data to, you know, AI models being trained and just kind of the general public. We are a Microsoft district, so we use copilot. I really appreciate the hooks with SharePoint in teams and all of that, but I just, I don't know if I'm just more used to the CHAT GPT interface, but I really feel like Copilot is just kind of lagging Behind Gemini and ChatGPT. Are you experiencing that when you're out speaking to folks and maybe they are Microsoft district and you show them everything they can do with ChatGPT? Because I feel like it can do most of it. Maybe it's just a familiarity issue, but I really think that Copilot is really kind of lagging behind and I'm wondering if maybe that is dissuading adoption a little bit too, because Microsoft is prevalent in public sector in schools as well.
- [21:07]
B
It is. I personally don't use Copilot that much. I don't like it that much. And it's because of inconsistency. I would say sometimes. Exactly. Sometimes it would respond pretty well, but sometimes it would respond weird and you don't understand why. It gave you a response that it gave you. And then you take the same prompt, you use it in Gemini. It's a great response. You take the same prompt you use in ChatGPT. It's a great response. I usually, when I do trainings, I usually run deep research as a. As a demo. And when I do deep research, I do in all three models and compare those to see what it came up with. You can also do the same thing with all the prompts, like you're creating an Excel formula. Just put that prompt into all three and see what it comes up with. So from my experience, I don't use it that much that I can speak on it. I don't have. My workspace at our company is Google, so we use Google Workspace. We use Gemini, we use Google Drive. I don't use365 to interact with. Sometimes I would use Copilot within the Microsoft products, within PowerPoint, within Microsoft Word, sometimes within Excel. But I don't use it outside like a Copilot. As an. As an AI model.
- [22:39]
A
Yeah, I don't enjoy using it. I guess I'm getting more tolerant of it, but it just feels like It's Gemini and ChatGPT 18 months ago Whenever I use it, you know, so it's not quite there. It's getting the job done, but it's taking a little bit more effort on my part and maybe I'm just spoiled by using the other two platforms. I'm used to just doing what I ask it.
- [23:00]
B
Right.
- [23:00]
C
And we do need to keep in mind that those companies, they have different business model, they have different priorities. Right? So OpenAI, they came with different objectives and mission. That's why they ahead of the game when it comes to gen AI. But Microsoft has been here for a while. They had different business model and different mission to establish, so that's why it sounds like they're catching up. But actually they have different priorities and this is something that we also need to keep in mind. But based on what I heard, it sounds like District has specific tools that they offer and you guys adapt and basically provide your feedback saying, hey, this is working well and this is not.
- [23:37]
A
I believe for us, if there's anything that isn't using sensitive data, I'll just use my ChatGPT account because it just seems to produce results that are saving me time rather than me getting frustrated with Copilot. But you know, I think, you know, Aziz, one of the demos we did together at that workshop with ASBO New York was I use it for payroll analysis. And even though a lot of that information is publicly accessible, I still feel more comfortable using it in a closed environment. So I've been able to train Copilot enough to where now it's. I'm getting the results I'm looking for, but it did take a lot more effort to get to that point.
- [24:16]
B
Point.
- [24:16]
A
So, you know, to. To Jamal, your point, maybe over time their as their business model is likely going to evolve as they're getting eked out with the AI race, maybe it'll change, but we'll. We'll see. But for now, I will continue using Copilot begrudgingly when I have to.
- [24:31]
B
So one tip that we can give regarding Copilot is agents, as you said, creating agents that are specifically designed to do some tasks. However, when you create those agents, create the instructions with ChatGPT or Gemini for that agent.
- [24:53]
A
That's exactly what I did. And it worked great.
- [24:55]
B
Yeah, exactly right. So we take the instructions and we take ChatGPT or Gemini. We upload a good prompt engineering book or prompting book that Google has white paper on it. OpenAI has white papers on it. They both have good documentation on their website. Taking that information and putting it inside a custom GPT to create prompts and then ask that to create instructions for Copilot agents that can change the interaction, that can change a lot of things.
- [25:36]
A
And kudos to you Aziz, because what you just explained, I, you know, I got the nomenclature from you, but just the prompt engineer. One of the biggest takeaways that I've gotten from you in everything I've learned in AI through you and your company was building that prompt engineer and had completely changed my approach in terms of building agent specific tasks. And you know, I in turn have now trained some of my colleagues on doing so. And just to be able to demonstrate how easy it is to get a fully fleshed out prompt that would maybe take you 30 minutes to really try and think of everything and knowing you'll probably miss something. But using and leveraging a white paper from the people who created AI to effectively build a prompt is just a no brainer and that has really changed it. So thank you Aziz. That's really helped me with my prompting skills.
- [26:26]
C
And this is where creative thinking. Yes, right as is. And this is where creative thinking comes along. Right. You basically solve problems for people, you look at all available tools and you basically experiment and bring solutions.
- [26:41]
A
Yeah, absolutely. So let's talk about what's going well and what's working. So in your experiences, what do you feel are real world AI use cases that are really delivering value?
- [26:55]
C
Depends as is.
- [26:57]
B
I can take, you can wanna take that.
- [27:00]
C
I know, of course, of course. I mean we definitely shape products, test them and iterate upon them faster than we used to. Now I see teachers also creating websites easily, they creating presentations so easily, they're moving really fast. And when it comes to concerns or things that are not going well, those are basically gen AI implications where you need to be very careful when it comes to sensitive data. And this is where as is, you probably have some use cases to share.
- [27:37]
B
Okay, interesting use cases that I can think of. It is important to have use cases that relates to everyone, especially to our audience, business officials who are dealing with Excel spreadsheets every, every single day. I would say one of the best use cases is to create formulas with AI and it does pretty good job. If it fails, that's okay, just go with it, iterate through it, respond, you know, change your prompt, ask it again, it hallucinates, it couldn't do it, ask another chat, ask it again and it will do pretty good job. I'll give you an example. If you're trying to create an Excel formula that you have no idea how to, you can just open dictation tool, you can switch to your Excel and you can start explaining what your Excel Is explain every column, explain the structure, explain what you're trying to do. And after you get the formula that works within one cell, another trick is to say, great, I like the formula, it works. Can you now change it so that it generates the whole column from one cell? And it will do that too. It sometimes writes, I don't know, 10, 15 lines of formula that you can just paste and it naturally works. Another one would be if you come up with a really complicated formula, put it in AI and ask to simplify it, or ask it to restructure it, or ask it to give you a different suggestion or a different approach. So for our audience, for business officials, it is extremely important to know that if you don't want to upload your data, the second way of doing analysis is to ask it to give you formulas to and analyze your data. So that would be one use case that relates to our day to day. The other one would be Deep Research. I figured that not many people use or know of a Deep Research. It's an incredible tool that can produce draft documents for you. I'll give you an example. Most of us right now are concerned about AI policy. AI use policy. And we're saying, oh, we need to create an AI use policy for our school district. How do you do that? You go to ChatGPT or Gemini, you activate Deep Research and how you activate it, you hit the plus button or in Gemini there is this tools button that you press. Then you select Deep Research. Then simple prompt act as an experienced policy writer. I need an AI use case use policy for our school district. And this is what our school district is. This is who we are. Put your website in and then say go and search for AI use policies in the surrounding districts. If you can't find it in our area, go to whole state. If you can't find it in state, go to all United States and get me good policies. Then take those policies, analyze the best one, then write me up to 30 pages policy that is ready for production. By the way, before you produce it, ask me any questions before you respond so that I don't miss giving any information. I don't want you to assume anything and it will go out sometimes. Search 30, 40 minutes, read 150 pages of articles, analyze many, many other documents and then summarize the information out of that and will have at least 85, 90% ready. Then you just go read it. If you're fine with it, then it's your policy. I think I gave two use cases. That's, that's Amazing.
- [31:56]
A
And I will say, speaking from experience, deep research has changed the way I approach some really complex problems. And if it's something that you can kind of just give a pretty thorough prompt, let it go in the background. The product you get when you come back 20, 30 minutes later is absolutely astounding. Hey, everyone. I just wanted to take a moment to thank Today's episode sponsor, UnitedHealthcare. Every school district is unique, and your health plan should be too. At UnitedHealthcare, our group health plans are built for you and your employees. Whether you're managing a small district or overseeing a large network of schools, we offer solutions designed to fit your needs, goals, and budget. Explore your options today@uhc.com Insurance coverage provided by or through UnitedHealthcare Insurance Company or its affiliates. Administrative services provided by UnitedHealthcare Services Incorporated or their affiliates. Health plan coverage provided by or through a UnitedHealthcare company. Again, thanks to UnitedHealthcare Today for sponsoring the episode. Let's get back to our conversation.
- [32:58]
C
Exactly. And in one of our workshops, actually, I heard one person sharing his thoughts when isis you demonstrated how to generate deep research, he shared that he did experience the, you know, that's kind of boost in his confidence because of the results that he got. And he was able to leverage those results, articulate the thoughts, and use it in his salary negotiation. And he got the raise.
- [33:29]
A
That's amazing.
- [33:30]
C
Yeah.
- [33:31]
A
I want to ask you both a question that's kind of off topic, but I don't want to forget it. You both are in the AI space so heavily. Is there anything that impresses you anymore? I mean, whether it's agentic, AI, large language models, what recently has really made you stop and go, wow, this is pretty cool.
- [33:51]
C
Yes. Okay. Personality. They are tailoring it in a way they actually building it in a way that it's kind of personalized to you. And each tool has its own personality. That's fantastic. Did you notice that ChatGPT would go and search and try to please you and make sure that you get the answer? Saying such a good question. I love it. Thank you.
- [34:16]
A
I told her to stop doing that. I'm like, you have to stop gaslighting me. It's driving me crazy sometimes.
- [34:21]
C
Yes, you got to do that.
- [34:22]
B
Right.
- [34:22]
C
But Jim and I would argue with you saying you can't ask these many questions. Break it down into X, Y and Z.
- [34:30]
B
Right.
- [34:30]
C
That's a personality. Come back to you and say, you know, let's do this instead of this. Same thing with other gen AI tools So that's something that impresses me. They bring that person of a human touch, you know, that kind of listens to you and talks to you, but when it comes to kind of adaptation, I guess, as is mentioned, and when he talks, you probably notice how confident he is, right? The advanced user versus beginner user. You can tell by the way they talk and use cases they offer. This is where you can see how practice comes along. It's not about intelligence, how much you know, but it's about how much you practice with these tools. So this is where you going to see how people share their thoughts and use cases in a way that they are very. They show the passion, the experience, that. The confidence. And that's why I, to me, not only being impressed by how those AI tools are improving our confidence kind of brings excitement, but also like seeing how people share those knowledges and making sure that they are kind of sharing it in a way that everyone can see and everyone can hear. I see a lot of comments on OpenAI, for instance, or like Gemini, where users would come and say, I love this, but I don't like this. That's also impressive to me. They are raising their voices and they are saying what is actually working, what is not.
- [36:02]
A
Aziz, what about you?
- [36:04]
C
What about you, Aziz?
- [36:05]
B
Okay, I have two that I will talk about. One I heard from my friend, and this is probably about eight months ago, I heard this prompt from him, and I don't know the sole source who came up with that prompt, but it's amazing. It's beautiful prompt. So I recommend all of the listeners to follow this prompt based on, you know, based on. Based on what you know about me so far. Roast me and don't hold back. Oh, my God.
- [36:44]
A
It's amazing.
- [36:46]
B
It comes up with so many jokes, so many roasts. You just. I was impressed how it comes up with those little things that, yes, make sense. That's to kind of second what Jamal said. Personality touch, right? Imagine someone is roasting you, and we need that to kind of bring our ego down, right? I do it from time to time. And one of the things that it said, it said, you did a few trainings and now you think you're in Elon Musk. And I said, really?
- [37:31]
A
Gotcha.
- [37:34]
B
But in all seriousness, the latest thing that I was impressed with, and it just amazes me how Gemini's canvas mode is extremely fast to come up with a small application. If you didn't try it, try it at home. It is safe as long as you keep it safe. If you ask Gemini first you need to turn on the Canvas mode. You go to Gemini, you do Open Tools, then you enable Canvas mode. Then the prompt you want to say is this. Act as an experienced software engineer. I want you to develop a dashboard for me based on the Excel spreadsheet I am uploading. The spreadsheet I'm uploading is expense data. I want you to use the spreadsheet to structure the software, but I don't want you to use any of the data out of the spreadsheet. I want you to give me one button so that I can click and upload my spreadsheet. And after I upload it, you will show me the dashboard. Add one pie chart, one bar chart, and add some filters for account structure. Every state has a different account structure. Take one account structure and say filter by fund, filter by object, filter by program, and so on. And I want you to give me a dropdown right above those graphs to summarize it by any element I want and make it colorful. Have a modern design and give it one minute. In one minute, it creates the software, it runs the software within the chat, and then you can upload your spreadsheet and it will give you great analysis. If you want more, you say, can you add a button that says AI insight that I can ask questions to? It adds a button and then you can chat with AI within the chat to your data and ask questions to your data. Now this changes the perspective, this changes everything because all of our business officials, all of the superintendents, all of the principals, whoever works with the data, they're all concerned that the software they use does not satisfy their needs, right? This is.
- [40:27]
A
So this is a highly customized software. In a few minutes, exactly, it's less than a minute.
- [40:33]
B
It writes 600 lines of code in less than one minute and it works.
- [40:41]
A
That's incredible.
- [40:42]
C
And as is right now, very impassioned because he loves what he does. But the thing is like. And he sounds smart, right? But all he does here is like he's using the framework tada.
- [40:53]
B
That's what she meant.
- [40:54]
A
I know, I was waiting for that. I was like, where's the other one he's using?
- [40:57]
C
Tada. Framework T stands for act as A stands for aim for the specific. For instance, as he's assigned a task, right? So he's assigning a task. Then D stands for define constraints, do this, don't do this or go search this and come back with this. And then inviting questions. That's it. If you follow that, you are going to get the. You're Going to get incredible results. And that TADA framework was developed based on research and analysis and courses that we took from great institutions that out there, you know, to simplify and make it easy for you guys.
- [41:35]
A
You both gave really sophisticated, introspective, amazing examples of what impresses you about AI. And I was going to share mine, and now it feels stupid because it's really not applicable to school.
- [41:48]
C
Please, please, please, go ahead.
- [41:50]
B
Nothing is stupid. You are a very power user. I know you. So go ahead.
- [41:55]
A
It has nothing to do with large language models. Bear with. All right, bear with me for a second. So you know how when you're driving in your car at night, a lot of cars have the automatic high beams that come on and when a car is coming towards you, they turn off? Not very sophisticated. Yeah, I. I figured out how to unlock this feature on my car where I'll be driving with my high beams on, and when a car comes around the bend, it will blank out and follow that car and lower the high beams just where that oncoming car is, and the rest are still illuminated in the high beams. I'm not doing it any justice explaining it on a podcast. It is the coolest thing I have ever seen automotively in my entire life. The fact that the headlights are sophisticated enough to follow multiple vehicles coming toward you and blank out, so they're not getting dazzled by your high beams, but yet keep everything else on is one of the coolest things I've ever seen. So not really practical to school business.
- [43:01]
C
No, it's not really cool. Exactly. Exactly. A lot of sensors and a lot of smart people working behind the scenes to make it happen.
- [43:08]
A
Thank you.
- [43:08]
C
So same thing. Yeah.
- [43:10]
B
And what I want to contribute there. First, it tells us that you drive through really dark roads. I don't know where you drive.
- [43:18]
A
Second is the back roads of Connecticut have no lights.
- [43:22]
B
Exactly. Second would be. Not every car is like that. It's your car. Right. But the technology becomes very. I completely agree with you. The technology is getting more reachable, more available for a lot of people and becoming a Standard. Before, like 10 years ago, we didn't have adaptive cruise control, and now adaptive cruise control is in every car except for the luxury cars. They want you to pay for it, but it becomes standard in Toyota, Honda, like all of those cars, it becomes standard because it's a safety feature. So as you said. Exactly. And those are AI usages. It's not just by itself popped in. It's not just a code saying, if you see this, do this kind of
- [44:17]
A
thing yeah, and I mean from a safety standpoint too. And I'll stop talking about headlights in a second, but because it keeps the high beams on the sides of the road where I live, there's a lot of animals and you don't, like you said, the roads are really dark. So to have that illuminated just is another safety feature and another layer in safe driving. So although it is could be perceived as kind of gimmicky, which I guess it kind of is impractical use cases. I'm still impressed. But yet there is a safety component to it as well.
- [44:49]
C
Yeah. And you understand that because you kind of process it and you're interested, you're curious. And this is where mindset comes along. Right. So if you want to actually become unstoppable, you need to understand and switch. You need to understand how it works and switch your mindset saying, this is designed by human for humans to serve us and amplify our skills. And you start with that and you basically understand that this is kind of tool that learns from data. You switch on and off, you provide some instructions and it learns from you and adapts and then responds to you. So this is something that I think it's critical to have in mind saying like, it's just a tool.
- [45:33]
B
Right.
- [45:34]
A
So we're coming up on time, but I don't want to let you guys go without talking a little bit more about the TADA framework. Jamal, you had talked a little bit about it. Can you explain it exactly? Kind of the process again? And really how does using this framework help users move from kind of just basic experimentation to more intentional high quality use of of AI.
- [45:58]
C
So again, this is a framework for those who are uncomfortable using AI just yet. So they have probably hit the road. They were confused, they were struggling and they didn't know how to kind of proceed. It's just something that we kind of observed and we created a TADA to help them get those prompt engineering in place. Because AI is a tool, Right. It's designed for us and you need to prompt and you need to talk to it. Like humans, they also appreciate storytelling. You need to tell your story, you need to tell what kind of problems you have, what goal you have in place. And this is where TADA comes along. Tada. Again, Aziz and I worked a lot on it. We went through Harvard Business School courses. We will look at MIT, we looked at, we even asked ChatGPT. So at the end we came up with simple framework saying, ta da. Make sure that you assign a role title to your prompt and Then to your agent, if you will. And then make sure that you highlight the overall aim. You assign a task and then you make sure that the defined constraints are there. And then you invite, you ask more questions, you allow your agent to ask you to many questions. This is the starting point. More you practice, the better you become and you master it. You notice you're going to see Aziz completely destroying TADA and starting with, oh, I have this question. And now you act like this and then assign this. So because he is very comfortable, like he mastered that TADA framework, right? And this is happens to me as well. This is all over the place. Like at the beginning you're very hesitant talking to AI, making kind of requests. And then when you master it, you become so proficient that you are just talking to and saying, give me PDF and it will understand you because it already has that database. It's like it got so much information from you that it's easily adapted, adapting to your needs. And Aziz, if you want to add anything else to that, please feel free to go from here.
- [48:17]
B
For those of you who are interested in remembering Tada, we have two avatars. One is mine, one is Jamal's. People choose Jamal's over mine. That's very disturbing to me because I don't know why they like her avatar better. Because, because for some reason she has a lightsaber.
- [48:38]
C
I got a laser.
- [48:38]
B
Yeah, yeah, she has a lightsaber in her hand and I don't. Now as Jamal said, it becomes natural after a while until you master it. Create a prompt engineer custom GPT and apply that TADA in there. And whenever you want to ask a question, just go there, ask a question and see what result goes comes out of it. For example, you say tell me about giraffes. Then that prompt engineer will spit out and say, your role is animal. An experienced animal scientist who has deep knowledge on giraffes. I want you to go and search about giraffes and give me some information about it. And these task. And your output needs to be a comparison between giraffe and other animals. Comparison. So it defines output and it says ask me any questions before you respond to me. So doesn't have to be hard. Again, TADA is if you wanted to remember it when you show up to people's home without invitation and you say, ta da, it's me. And that's kind of a framework that we developed. So always give title because giving title gives perspective to AI model. It knows how to look at your question. If you say act as an Attorney. It will give you attorney's perspective. If you say, act as a parent and tell me if you don't understand anything from my presentation, then it will read as a parent who doesn't have financial knowledge. So act as is extremely important. Assigning task. Just speak as much as you can.
- [50:30]
A
Yes, John, I was just going to say, I mean, assigning it to or telling it to act as the importance of that is really narrowing the scope. Because if you think about these large language models, they are learning from every corner of the Internet, pieces of literature. So you want it, you don't want it to do an amicus brief. And as Shakespeare, you wanted to do it as an attorney. Right. So you're narrowing its focus and getting a better or yielding a better result by doing so.
- [50:58]
B
Exactly. Because if you said, look at my presentation and tell me what's wrong, it will just look at your presentation. And so it's seamless, it's perfect. But if you say, look at the presentation as a parent and tell me if you understand it or not and you don't have a financial background, then it will say, oh, you just use this jargon. You used P and L or you used balance sheet and people don't know what balance sheet is. People don't know what P and L is. So can you define it more or can you use a better language to explain it to people who are not financially savvy? Now, assigning the task. Talk as much as you can. Put all the context you have. Talk about yourself, talk about the task, give examples, give success rubrics. You say, if you are successful, your response is successful. If your response has a table, your response is successful if so and so. And then give all your constraints, put your dos don'ts. I want you to give me a table. I want you to give me two paragraphs, five paragraphs. Whatever constraints you have for the output, give all that. And then inviting it to ask you questions. The other day we were working on a proposal and we asked it to ask us questions. We said, this is the proposal we're working on. Can you give us questions before you respond? And this was for the prompt Engineer. We were using Prompt Engineer to create a prompt. We said, ask us questions before you create the prompt. It asked us 54 questions. We took about 45 minutes to an hour as a team to answer every single question in detail. And it gave us 85 page contract. Wow. After the deep research was done. So it's important to kind of make it your habit and it doesn't have to be long, you can just say act as a teacher and tell me if it makes sense. That's it. Right. So act as, tell me if it makes sense and then whatever that it is, you explain that. So it doesn't have to be complex. It can be three sentences, but those three sentences can still follow the framework. As you use it on a daily basis, as you use it in your daily interactions, you master it and you become better.
- [53:28]
A
So tada, it's an acronym, right?
- [53:31]
B
So yes, it's an acronym.
- [53:32]
A
T, right, Right. So T, you're assigning it title to act as Title A is to assignment. Assignment D is the do's and don'ts and then the, the other, the last A is ask any questions.
- [53:46]
C
Ask any questions if you have. Yes, a follow up. Exactly. We carefully chose those four components because you know, you, we, we've seen so many people, you know, asking questions and getting confused and sometimes giving up on it. Right. And based on best practices, what we see in industries right now we have one person having four agents working together for that specific expert. One is creating emails, another one responsible for marketing, another one is responsible for answering questions, another one is basically taking and going through analysis and so on and so forth. See how powerful you can be right. After understanding how it works for you. That's why after multiple iterations with this other four critical components that you need to keep in mind in order to be efficient, make sure that you understand the role, make sure that you assign the task and make sure that you define constraints, the format, don'ts and do's and don'ts and then from there invite for asking questions. You can be very flexible, you can skip some of them if you do really need to move fast. But again, those four components were very critical. Once you master it, you're going to create so many like agents and working tools next to yourself so you can be, you can spend more time on things that require more strategic thinking, creative thinking and so on and so forth. You're saving about two to four hours. I personally say two to four hours per day.
- [55:10]
A
Well, when I showed my wife the headlight thing on my car, I think I did say tada. So it's similar but different.
- [55:19]
C
Well, now you, now you know, now you know. And for those who want to, to adapt, for those who are very, you know, at this, you know, big like first phase of adoption and don't know how to start, they need to understand their role first. Are they teachers, are they leaders or business managers and so on and so forth. Right. Define your own role, you know who you are and then where you want to go within a year, within the five years, and so on. And then make sure that you have critical three objectives or goals that you want to achieve. Is it going to be save some time or bring engagement to your classroom and make sure that you appreciate your teachers, whatever it is. Right. Make sure that you have those three goals in mind and then have three to five tools to experiment and have a roadmap to adapt them incremental. Start with I'm going to use this tool once per week to create my presentations. I'm going to use this tool to adapt and move this one per week, for instance. Right. And then from there see how far you go and I'm sure you're going to enjoy the process you're going to go through. Like, hey, this is cool to oh my gosh, this is not going to work. And then, oh my goodness, I love it. I'm going to share it and I'm going to bring more people to this.
- [56:38]
A
Well, it's like any Jamal. It's like anything else that you would practice. It becomes muscle memory after a while. So when I think back to when I first started using AI, it was, I was very clunky with it because I didn't, I wasn't familiar with the interface, I didn't know how to prompt and now it's just kind of instinctual in how I use it and have woven it into kind of my everyday work life and personal life. So to your point, the more you use it, the more you get familiar with it. You use the TADA framework, it just kind of levels you up even, even further. So it's like with anything else, you have to keep practicing and becoming familiar with with the interface and how it responds to you.
- [57:12]
C
Yes. And be creative.
- [57:13]
B
If you haven't read Compound Effect by Darren Hardy, Compound Effect is real. And if you this is exact compound effect example, if you use it on a daily basis, it's compound effect. After six months you become natural. That's one the second recommendation I would say compare yourself only to yourself. Do not compare yourself to those highly skilled programmers who use AI to generate agentic agents, who sends emails and do create businesses. Do not compare yourself to anyone. Just compare yourself to yesterday's yourself. Compare yourself to last week's yourself, because that's when you improve and that's when you can concentrate on improvement rather than chasing the dream.
- [58:05]
A
And that. And for those listening, that's why I don't go, hey Aziz, check out what I did, because I know he's probably already done it like two years ago,
- [58:12]
C
but still, I think it's interesting. You never know. Sharing is caring, right? Keep sharing. Absolutely. I personally learned a lot. There are so many ways of doing the same thing and getting completely different results and personalizing it. So yeah, keep sharing.
- [58:28]
B
Aziz.
- [58:29]
A
Jamal has such a positive anchoring spin back to everything that we're talking about. I go from joking about Headlights, she's like, no dummy. We're talking about how to leverage AI. I appreciate it. Jamal, we need more of you on the podcast. Stay focused, stay focused.
- [58:45]
C
But thank you so much for having us here, John.
- [58:47]
A
Yeah, yeah. So before I let you go, I want to just end on this and ask you, for those listening, what is the best way for everyone to stay connected, informed and inadaptable as AI inevitably is going to continue and evolve? I mean, we talked at the top of the episode just how quickly something like ChatGPT went from kind of a goofy chatbot to something now we can use in an agentic sense. So what, what, what can listeners do to really just kind of stay on top of how these things are evolving?
- [59:18]
B
I would say sharing is the most important thing. If you had a success with something, go on LinkedIn and post a message. Doesn't have to be perfect, it doesn't have to be great post, it doesn't have to be anything special. If you had a success with expense analysis and it gave you something, then just go on LinkedIn and share and tag your group, tag ASBO. Tag ASBO International. Tag, tag whoever your group is, your audience is and share this information. Because if you share, people will have that confidence too. They will start using what you use to. You don't know if something that you used, because sometimes this is what happens. We do something and then we think, this is so simple, this is so easy. Everyone knows that. No, they don't. Just share what you had success in and then read other people's success. That is, I think, one of the most important things today to share because we hear from engineers, we hear from those high tech companies, we hear from those self made millionaires who only talk about let me teach you how to use AI to create agents. But we never hear from the business managers who says, you know, this is how I actually automated my workflow and it helps me on a daily basis. We need to hear from you so that other people can get encouraged, can get inspired by you. And that's, I think, one of the most important things today.
- [60:54]
C
And I just want to add to that, like have fun with this. Just enjoy listen music while you're doing it. If you have that power, create a tiger team. Have a goal, experiment, share again within the team, within the group, within the platforms that you have. And if you interested in joining any community that is out there for leaders, they probably have a lot of communities out there where they share their knowledge or have invites, please go ahead and join them. I usually do that and this is the, I think one of the great ways of connecting and building your network. And the more you talk about AI, the better. Your brand is going to kind of change eventually and and you're going to be that asset within the school, within the district that all people are going to come to you and ask questions even though you're thinking, I'm just experimenting, I'm just sharing.
- [61:49]
B
No. And that reminds me, Jamal just reminded me to tell everyone, please mess with your friends. Just generate images of, she mentioned tiger. You know, make images of them and tiger or tiger fighting them or whatever. You think that will be funny because if you apply to your personal life, then your professional life becomes extremely efficient.
- [62:14]
C
Yes, indeed.
- [62:15]
B
That's great.
- [62:16]
A
Well, Aziz and Jamal, thank you so much for joining me today in School Business Insider. It was a great conversation.
- [62:21]
B
Thank you for having us.
- [62:23]
C
Thank you for having us. Yes. Enjoy. Bye.
- [62:26]
A
Thank you for tuning in to School Business Insider. Make sure to check back each week for your favorite topics on school business.
- [62:33]
B
SA.