# Optimizing Educational Workflows with A.I. - A Process Design

- Channel: [Los Angeles Regional Consortium](https://www.youtube.com/channel/UCZ1ctndrPotopXgFJ10az9Q)
- Published: 2026-01-26
- Source video: [Watch on YouTube](https://youtube.com/watch?v=8fGRhseOBOQ)
- Read on ReadTube: https://www.read.tube/p/videos/8fGRhseOBOQ
- Language: Original (source language)

## Summary

**AI Workflows Streamline Education**

The session shows educators how to design repeatable AI workflows to save time, improve consistency, and support tasks like rubrics, syllabus updates, accessibility, scheduling, and email triage. Presenters stress that the most important part is human oversight, especially for grading, student interaction, bias checks, and academic integrity.

The workshop explains that a workflow is more than a single prompt: it is a repeatable, multi-step process that AI can help complete from input to output. The presenters demonstrate how these workflows can reduce repetitive work in education and administration, especially when tools like Gemini, Microsoft 365 Copilot, and other cloud-based systems now combine many functions in one place. They emphasize that workflows are most useful when tasks are repetitive, data-heavy, or structured, such as creating rubrics, updating syllabi, building Canvas materials, drafting grant content, generating study guides, or organizing inboxes and meeting notes.

A major theme is that workflow design should always include human review. The presenters repeatedly stress that educators must stay involved in tasks tied to relationships, judgment, or accreditation, such as student discussions, feedback, recommendation letters, and academic integrity decisions. They also highlight responsible AI practices: check citations, verify sources, protect sensitive data, and make outputs accessible with alt text, correct headings, and readable formatting. Accessibility and privacy are presented as core requirements, not optional extras.

The session includes several live demonstrations and examples, such as building reusable prompts for music creation, class matching, discussion rubrics, weather tools, and accessibility checklists. It also compares different tools and use cases, noting that Gemini and Microsoft are especially useful for integrated workflows, while tools like Midjourney, Canva, Claude, Suno, Runway, and Windsurf can support more specialized creative or coding tasks. The closing message is that AI workflows can make work faster and more flexible, but people still need to design, test, refine, document, and oversee them.

## Article

### Consortium welcome and session overview

for the Los Angeles Regional Consortium, also known as Lark, which represents 19 community colleges across the Los Angeles region. I also have the pleasure of co-leading the Lark AI task force with chair and assistant vice president Dr. Nar McKean. Today's AI professional development session, **Optimizing Educational Workflows with AI: A Process Design**, is made possible through our partnership with the Los Angeles County Office of Education, also known as LEO, and the coordination and vision of Lark's AI task force. Leading today's session are Educators Technical Director Dr. Raina Walzinger and Managing Director Marcy Drummond, who are ready to help you identify inefficiencies, implement AI-driven solutions, save time, improve outcomes, and so much more with AI-supported process design. You'll find today's professional development session and more in the Lark AI learning library, and we'll be sure to drop that link in the chat. And with that, I turn it over to Educators to get today's professional development started.

> **Judy:** Awesome. Thanks so much, Judy. It's great to see you, Dr. McKeen. Thank you so much, and Kevin, for all your support.

I'm Dr. Raina Walzinger. I'm here with Marcy Drummond, and we're really excited to be here today. This is one of our favorite subjects that's changing quickly in workflows. If you've been following AI news in 2026, if you even just Google workflows, you'll see tons of apps coming out with workflows. Where I teach, we already have Google Edu on board, and I think that's coming to all of us working at our community colleges, and we'll be even showing some of that today. So with that, let me share my slides. Let's see. Yep, I can do that.

Okay. Along the way, this is going to be interactive. You're going to have a lot of chance to do your own workflows during the session. And if you have certain workflows that you want to try or you're interested in certain things, just put them in the chat so we can include them. We've been doing this with workflows and coding in different areas for the last, well, I've been doing it for the last about five years, so excited to show you guys that. And with that, let me let Marcy introduce herself.

> **Marcy:** Yeah, good morning everybody. Happy to be joining you today and appreciate Lark providing this opportunity. And the only thing I would add to the opening is that not only are these new tools coming out, but all the AI tools we've been using extensively, like ChatGPT and Microsoft Office and Google, used to require you to go from tool to tool to accomplish a task. Now they have embedded multiple tools like image generation and video generation, so you don't have to leave. So workflows are becoming much more simple, and you can finish an entire workflow using just one tool now. So when we first envisioned this workshop, we actually had: now use this tool for this, and now move that file over here and use this tool to create the image. And now you no longer have to do that. So it's really about instructing AI how to complete an entire task from A to B that has multiple steps. So it's much easier now, and we're really, really excited to show you what they can do.

Perfect. All right. Quick there. Okay. So what is a workflow? It's different than just our regular prompt. It's actually something you can chart out, like your process, like you can do in any leadership structure or process for enrollment management or scheduling classes. It's a repeatable process that you can do over and over, and you set up the AI to do this for you. While we're doing this introduction part before we get to the exercises, think about things that you do that are repetitive and take several steps. I know one for me when I was a chair was a big one: classroom management, who was teaching what, where, and when, and then also coordinating that with the college. Or even getting new classes together in Canvas or creating new processes for healthcare for part-time faculty. I mean almost anything.

So in education, we have a lot of these that we spend time on, kind of a lot of time on the computer, and that's kind of how you know, like are you sitting there on Excel and Word, kind of doing things over and over, where there might be something that could be more automated. So workflows are basically your own program set up to do it. Now, this is an interesting concept because we've relied on edtech until now to give us these, and we still do. But some of the stuff that we do is personal to the way we like to work it, like it could even be the way emails go out or anything. And so you can set up your own.

Without workflows, so we have one-off, like sitting at usually Microsoft because that's what we're using, things. You have changes in results. You might not include the equity pieces, and accountability could be changing. And then with your workflows, because you're setting up all the processes, it's going to be repeatable. That's kind of the big thing. But you have to have the most important element of human oversight over what's going into this data and your output. Everything that I've read says that's the key element: somebody has to be checking these things. And this is an important one even in the development of AI. The companies themselves are using AI workflows to update their own AI models, but then when they release some of them to the public, you've seen OpenAI do it. Sometimes they backtrack and they're like, you know, they didn't have enough human oversight to catch problems with them, and they're doing malicious things like Grock does, which is why we don't use that too much in education.

Creativity is a system. This is the exciting thing. I just did a talk with K-12 teachers, and it was interesting bringing arts teachers together with computer science. Usually they don't do workshops together, but now it's becoming kind of one thing. So the computer science people are teaching coding and arts people are teaching visual outputs, and then now they go together. And how do you get people to think on both ends of these things and not in silos? So you have all your input, which is more technical and structured; iteration, which is really creative; and then your output, which is kind of like a creative technical look. Are there problems with it? Is the output looking good? And that skill on the output is new for workers and students because now you have to be on

### Identifying repetitive tasks for AI workflows

the technical and the visual side of things. So, **AI-ready tasks**: what things take longest in your life and feel repetitive? Again, if you’re keystroking a lot of things all the time, those are the ones that you want to try to create a workflow for. Where do you start from a blank page? I don’t start from blank pages anymore because we have now AI partners, and so you can always get ideas. One of the things that is kind of going away, you don’t hear the word as much, is roadblocks. So when you have a roadblock, you can just ask, “Hey, I’ve got a roadblock on starting this thing. Give me some ideas.” And so that’s one of the things that’s changing.

Pulling together information from different places is another big part of this. Every single workflow or language learning model has a paper clip or a plus sign. So you can pull data and information from almost anywhere: OneDrive, Google Drive, your desktop, almost anything. That’s really important in the workflows, as is your file organization. It’s a big one, and you’ll see why when we show the workflows, and then what would benefit from an alternatives perspective. Anything on that, Marcy? I can keep going.

> No. I’m just putting in the chat about the how also some tools have features that you can load multiple files into and work with them at the same time and synthesize them, integrate them, summarize them. So, I’m putting that in the chat.

Okay. Because I want to make this interactive after this, I’m going to show a quick example and then I’ll move on.

#### Creative literacy and workflow examples

The creative literacy, and this is an area that I’ve been working with in my entire career. So as some of you know, I’ve kind of sit between industry and teaching my whole entire career. So to use, and I’m a biomedical engineer, so creative iterative design, ethical judgment, and visual thinking—that’s how I was actually trained in college, because creating medical devices to be not harmful, and that was actually very challenging to do.

So in that, I’m just going to, because I want this to be interactive, switch my screen for a minute and just show you a couple what a workflow kind of looks like, just so you can see the back end of workflows. So this is one I created called **Sununo prompt maestro**. So what it is: I create a lot of albums, and so I realized using AI I wanted my voice and my instruments, but my prompts weren’t that great. So I created a workflow using this is a piece of Gemini that you can create a workflow in. But this is the back end. So this is what we’re talking about. What are all the inputs from the user? So it’s going to ask the user these questions that I identified when I programmed this in my workflow. Then Sunno is going to give certain outputs, and it’s showing what all the steps are here. And then here’s what it does. And so this is the backend editor: user input, generation, output, assets.

So it can get data from any of the things—uploading files from your desktop, like I said, or any other place, YouTubes, anything like that. The front end of workflows usually look like something like this. So once I do this, I’ll be like, describe your style. And we’ll do this together. Maybe not in Opal, but so you can see each step of the workflow. So now it’s asking me this in just an interface, but I already told it, you know, made this little workflow. So every time I do this thing, it’s going to be repeatable. So, instruments to use. So let me do that really fast and I’ll show you how awesome this is. Key any tempo. So then it’s going to go through and go through that workflow based on the user inputs. And this is what you would do. And then it’s going to give me an output that’s going to be a prompt. So I’m going to get repeatable creative prompts because this is what I built this tool to do, to use in another application where I’m creating music. Takes about, it should be about done.

Okay. So, there we go. So, I’ve got inside of my music app, I’ve got now instrumentation, and it gives me moods, a better prompt that I could probably write, that I could at least decide if I want to use it or change it. And it gives me alternatives as well. So, I can use this tool forever to create prompts for my Sunno app where I create music. So, that’s a really good illustration of what a workflow looks like. It works. Looks like every workflow I ever did in graduate school, but now it’s back here that they automatically put it together for you based on what you’re trying to accomplish.

Let me stop there for a second. See if there’s any questions because that’s a big concept on these workflows, different than what we’re used to with language learning models. Any questions or ideas on what you might want to be doing workflows for? Silence. I got an idea. If you have some, you can put them in the chat.

So, I’m going to put one for myself: **Canvas modules**. So I can create a workflow to say, my modules are these six things every week, create those for me in a canvas format that I can put import in there for 18 weeks and then I can fill in my pages. I could put one in there that says, “This is my college syllabus. Update it with the new dates every semester. I’m going to put in the dates and you fix it and then send me my word document and my PDF.” So those are a couple ones that I like. So, think about that.

All right, and it’s a little bit slow. There we go. And it didn’t go where it was, but that’s okay. Okay. So, we’re here on this part. So, this is I wanted to show you that in the software what it looks like on the back end and Google, Opal, Google. I think a good example. Zapier is another one, but Google’s kind of automating that for us. And it works in, I think, the pro accounts, but we can do something similar in Gemini. So this is what you need to kind of think about: the core stages of the workflow. So this slide is probably the most important. Defining your task. What are the inputs that you need from the user, which could be yourself, could be people on your team, could be your department team. For example, if you’re doing class scheduling and you have a department meeting and everybody wants to give you their times when they’re on campus or whatever, great time or the types of equipment they need in their classroom. And then from there, do your prompt and model use and your output that you want to do.

### Human review and communication ethics

Then you need a piece in there where it's a human decision so you can iterate on that and then run it through again. So it's not going to be like this is a perfect output every time. You have to look at it and go, okay, this is okay, but I want to change X, Y, and Z. But this slide's pretty important.

> Marina can I ask a question?

> Sure.

So, I recently was talking to a professor who is kind of deep in this world of AI. And he was saying that some of his students were exclaiming, they're like, "Oh, professor so and so, you're so awesome. You're always communicating with me. You're always reaching out." But he said it's automated. He set up a whole workflow so that he plugged all that communication in, in touch points with his students as reminders and this, that, and the other. Would that be an example of a workflow process such as we're talking about this morning?

> It is. I have an opinion about that. You guys know me, I'm a touchy feely, emotional type of professor. So for me, I choose to do my announcements and then fix them with these workflows and then put them in there. What I don't do, so that's just fixing, updating my better explanations of things that are announcements. But students on discussions, like for professor, that's where the RSI is important, like we're engaging with the students, understanding what they're learning, what their questions are. So on the discussions, that's all me. So I think announcements and those kind of things are good, but for RSI, for accreditation, the faculty are, it's our responsibility to interact with the students and that's what that's all about. So in my opinion that should be, and this is where those gray areas are, my opinion is that should be done by you and it's not hard because you can do Canvas and discussions all even on your phone. So you can even dictate them in there like, hey that was a cool discussion and thing, da da, what did you think about? So yes, you can automate it, but those are those ethical areas that we need to keep talking about as a system. So that's my opinion. It might differ from other people's, but I hope it doesn't. And that's actually the whole entire fun part of our job on the faculty side, is getting to know your students and helping them progress, right? That's the whole heart of the job. So if that part was automated, then for me that would be a much less fulfilling career. Any other comments on that? I think that's a really great point, Judy, about whether those are automated things or not and whether they

### Meeting students and adding safeguards

#### Meeting students personally

Comment that yes, we need to agree with we need to meet our students, because I get a lot of students who ask me for recommendation letters. If we don't get in touch with them, if we don't know their personality, how are we going to know that we're going to write the right recommendation letters for them? I usually even ask them for their activities. What do you like? Because it's important for them to know, and even also for them to know, let's say, if they're going to get a job, I wanted to know what their skills so that I can give them a very good recommendation of how they can start building experience.

I always tell the students do not wait until you graduate to start looking for a job, because then it's going to be very disappointing. So I always push them to start looking earlier so that they know that they can have some experience before they get a job. And yes, the one way is by knowing them personally, even though we're online teachers. I feel that I have a lot of connections with my students by reaching out one by one.

Beautiful. I feel the same way. I have so much fun doing it. I mean, I don't know why I would not want to be communicate with them. And sometimes students in this age of technology have a lot more questions than they used to. I don't know if you noticed that. And so we're here to help them individually with those as well. So that's what our online office hours or whatever opportunities they have to communicate with us are about.

#### Ethical and non-ethical uses

So, Judy, yes, you can, but then there's that definitely ethical piece on your for some of those non relationship repetitive kind of

Yeah, definitely, like announcements, Zoom meetings, reminders, things like that, for sure. And actually even moving content from your old paid books into Canvas and things like that, where you're doing workflows of, you know, in modules that we do now, and actually even just factual quizzes, not essay type stuff, but there's a lot of things, opportunities where you can do workflows for that. Like if you're always making a 10 question end of the module review with multiple choice on the thing, you can feed all your stuff into there and just have it do it 18 for 18 different things for a semester. Those are if they're factual. If they're more than that, then we do them manually.

But that's a great point. I just wanted to take the opportunity to bring that up because it's a good topic. So, put a little few perspectives on here. So, for there's different focuses depending on your work group. And so for educators who are worried about kind of the way that we teach and and a lot of, and I want to bring something up on here, I have not this even this last week, I've seen more complaining from even in the K12 world parents on academic integrity whose kids were called out for cheating and then they're saying they weren't. But it's getting to be a major major problem for teachers, students, and parents. So, we're in this world right now where we haven't figured out how to fix this and where the truth lies and what how much AI people can use and all that. Just let's bring it up. It's a big issue right now.

Institutionally focused, you know, policy alignments, risk management, equity and access, of course, and institutional considerations. So those are depending on your work group what type of workflows that you want to do.

#### Types of workflows

All right. So we're going to get into types of workflows. And then we're going to do some exercises with you guys.

##### Conversational cognitive workflow

So conversational cognitive workflow, put a thing here. So it's a step by step guided through conversation. Supports lesson designs and assignments. This is what I was talking about. It follows clear phases and AI guides each stage. So, this would be good for creating assignments or even module elements for teaching. It would also be good for creating workflows for repeatable things like classrooms, assignments and like technology and classrooms, things like that.

##### Repeatable prompt-based workflows

Repeatable prompt-based workflows. These are reusable prompts with consistent results like I showed in that exercise in the beginning that I did. So that you get like consistent results every time. And these are really good for nontechnical users. So once you have your workflow, you just answer the questions and it gives you consistent answers.

##### File and document workflows

These are file and document based workflows. These are great for things like in HR could be different accessibility documentation, even for your college rubrics throughout. So anything where you need to input stuff. So if you're inputting like your policies and procedures, board policies, you know, things where you need a lot of input and for teachers it would of course be your course materials.

And they're safe in the two things we have that are safe now: Microsoft 365, so our OneDrive stays within our environments and it doesn't go training on that or sending it to other people, so that we can protect our curriculum, what we're creating, or our policies within the institutions. And coming in now, rolling out, is Google edu. Those are also our personal environments. Google doesn't train on it and it doesn't share it. So we'll have a second environment where we have a safe place to store documents and share them within that environment.

So both Google edu and Google edu is FERPA compliant. I don't know if I got a letter from one of the districts and it's all FERPA compliant. It does because it's FERPA compliant, it changes some of the things from Google Pro because it doesn't let you share in and out. So it behaves a little bit more like Microsoft, and I had an IT professional look at my Google edu and it turns out that it is programmed in conjunction with Microsoft on the back end. And even that workflow you saw is like Microsoft. So they're definitely sharing technology behind the scenes that are similar between them now.

#### Human and AI decision points

Do you want to talk about this one, Marcy?

Yes. So in later in the workshop we're going to talk about automating workflows, but there are tools now where you can start to automate these light, it's called lightweight automation, where it'll start to work through it and you can identify in your workflow when it should stop and ask for input. So that's the identify human versus AI decision points. So you can have points where it stops. It allows human decision points and then regardless of what tool you use, you have to always review what it

### Turning repeatable tasks into standard procedures

output is. So it may be correct, it may not, you may need to iterate and that sort of thing. And then the last thing, and I do this all the time, is once you have a workflow that you want to use over and over again, create a standard operating procedure or a playbook so that because it'll likely be a month before you use it again. You'll be like, "Okay, where is that? I can't remember where I saved that. What what were my steps?" So, I always stop at that point and create some sort of a playbook. Also, it's great that you can share that with others. Especially like for faculty, if you're creating some workflows that you think other faculty can benefit from, then all you have to do is create your playbook and share that with them.

### Choosing workflow tools and accessible platforms

Gemini, because they just came out with Opal with workflows, and it has a seems to work really well, but you can use almost any of them to do this type of work. Microsoft 365 has workflows now in the enterprise version, but I don't think the EDU version; that'll probably roll into it. But you can even make all workflows is that they're coming out with is things that you can do in what's called canvas. You can program these things.

So if I went to Gemini and did this by hand, go into my Gemini and do a new chat. I wasn't going to do this, but I'm going to do this anyways. So in canvas I can make a custom workflow from here and I can just code it like I can say, you know, based on my document I uploaded for a new program or some or whatever, like maybe a program review or a new thing like Judy was just saying, like a sweep grant or something like that. Here, you know, look at that and then make a tool where I can fill that out and have it optimized and make sure it's accessible and then so then the tool will be like Judy's super cool AI tool for sweet projects and then it'll ask the questions from the form and then you fill those out. So it'll make it consistent and then it'll create like you can say create five versions of it based on these criteria. So that's what canvas is is going to do.

So if I create a quick one like create a tool to just do something really quick to pick between two classrooms based on equipment for CTE course, trying to do something short. So it'll create this workflow and then I can use this over and over again. And so it'll be like what's the classroom, what's the class called, what's the equipment in there, and then it'll give me like priorities. And so that's how you create these. And what's nice in Google that's different I think in all of them, it has the easiest sharing protocol because it shares the same way it's as Google class, Google Drive. You can just share and then you can share it to see here it is find and I can change the colors and brand it and all that.

So, you can share just like Google Drive. You can share it and you can have the link to anybody or you can restrict a link to people on your email or people in your workspace. So, it made a little tool. I can't see it all because I'm on Zoom, but oh, I see. No. And then I can decide whether I want to iterate on it. It's pretty great. Like 30 seconds. Now, this is a workflow in the background that I'm creating using Canvas in Google Gemini. And then I can share that. I can share it with, I have a link I can share it in the .edu version. It gives you the option to share it in your work group or in your public work group but not to the world at large.

So, and then you can so then from there you're like oh I left out like some kinds of equipment cuz now I'm doing welding or nursing so I need to add a bunch of stuff. So you go and iterate on that and then you make great tools. You can put all the equipment and your whole all your CTE programs in there on a spreadsheet and say look at this spreadsheet based on the program and the equipment and make this tool. So that is an and now I've got a repeatable workflow for CTE classroom matchmaker. So I just tried tried to do a short one so it wouldn't take forever but you can see how amazing a workflow tool can be. Questions on that. So, you can do this on any of them. I think they all have the easiest ones I think are Gemini is probably the easiest for me to share, but you can do it in Microsoft 365, Chat GBT, Claude, almost anything. But Chat GBT doesn't share and it's also not private within our institution. So I would say for us we should be using Gemini edu when we get it and Microsoft 365. Okay. Any questions on all of that? Let me go back to where I was then. Okay. Here's a second example put together for AI and accessibility checklist. I did this in Gemini. So these are some tools that you could do. So citation tracking is one that I love. So verify AI generated sources. Alt text generation. So in fact Marcy I should have you talk about this because you do this a lot.

> Yeah. Um so I I just put in the chat that typically if your institution has provided you with a license like to Gemini or Microsoft 365 copilot they've already done the enterprise data protection for you in Microsoft actually when you open it there's a green badge that you can see in the upper corner and that tells you that data protection is on. But it's really important if you're down whenever you're downloading a file and uploading it into a tool that is not provided by your institution, chances are that data is not protected. So, always just like with anything, don't use student ID numbers or any information that shouldn't be public.

> Citation tracking. I usually in my prompt, I always ask for citations. Some tools are now giving you citations automatically like Perplexity does, but I always ask for the citations. And then one of my favorite trick tricks is I'll pull it into another tool and ask it to fact check the citations. So, you know, so that's really important that you ask for not only the sources, but specifically ask for the links because it could give you the source name, but not the link to it. And then, alt text generation. It's even beyond that. I use AI tools all the time to make something accessible. So, you can upload that into either Gemini is perfect for this. If you're using images, just upload the image and say, can you create an alt text for this? Or upload an entire file and say, can you give me a another document that's made accessible, including alt text, correct headings for tables, and things like that. So, now they know. So, they've been trained on accessibility. So, you don't even have to specify make the tables. Just say, "Make this document, make an accessible version of this document, and it'll do that for you." So, that's one of my favorite workflows. I do that all the time. Even, you know, now, if you're using Word, it has accessibility built in, but it didn't years ago. So, if I'm using an older file, I can upload that and ask it to recreate it and make an accessible one.

> I know you made an accessibility workflow before for

> I did. Yeah, I had to I made an actual assess.

### Using reference prompts and accessibility notes

This is the great thing about these AI tools: six months ago they did not know. When you use the word *accessibility* with AI, it thinks people giving people, it is user friendly, right? Or available to the public. That is how it defines the word *accessible*. But now if you say, *I’m an educator and make this accessible,* it now knows what we mean when we say accessibility, that it has to be formatted a certain way. So the whole workflow I built a long time ago isn’t even necessary anymore because these AI tools are evolving and now they know what that means.

> It reminds me, in all of my settings for my different things I use, I always tell it who I am and that I’m an educator and what I’m interested in and how I want it to respond. And then that also ensures, and it does when it responds to me. It said, *as an educator and creator blah blah blah blah blah.* So it gives me those responses in that format.

Okay. Uh so here’s a reference guide that is in part of this PowerPoint, and we have it for different workgroups, but some example prompts for syllabus refresh and study guide, a lecture to a podcast, doing deep research, and visual aids. So uh you can try all these out in any of the tools, and the tools that we use are on the left, and those are the ones that our institutions are all using.

And then just as a side note, Nano Banana is part of Gemini. It’s the image generator. It’s still the best one in the world, and other apps are mostly using Nano Banana as their image generator. Um and Gemini. So Gemini is part of Google. Nano Banana is part of Google. Notebook LM is part of Google as well. So that whole world that we have here is mostly Google. Now Claude is made by Anthropic, and they do really nice coding, and so you see the words *Claude code* all over the internet because it’s doing amazing things with code. Then there’s chatb.

So any questions about this? Thought maybe you could enjoy some of these. Um these, and you can see like these would require you to put something in your connected drive. So whether it’s OneDrive or Google Drive or whatever you’re using, it should be kind of one of those two unless it’s coming off your desktop, as far as doing these prompts, because those are the main tools we’re using in education.

Here’s another, a bunch of prompts that we have for administrators or professional staff. Um, and so these are fun. So, inbox triage, that’s a great one. I know. Um, like for all the deans I work for, it’s like impossible for them to get to our emails. Meeting minutes, cleaning data, it’s a big one.

### Automating routine analysis and office tasks

Survey analysis, workflow automation. So you can see there are workflows you can set them up like the workflow automation: every Friday at 3, scan my invoices, extract the vendor name and total, log it in my department budget spreadsheet. So that's a workflow that would take you several steps to do, and you can just have it do it automatically.

> “So, Arena, this is the part I’m having a little bit of a disconnect. So we can design these workflows, but integrating them into the apps—what is that step for, then, the workflow doing what you’ve asked it to do?”

> “So let’s do one like that. Marcy, should we do it in Microsoft 365 since we have it all connected to our email?”

> “Sure, if you want to show them a different tool other than Gemini.”

“Yeah. Okay. So I’m going to go into my Microsoft. Actually, they have a new thing, but it’s not—it's still rolling out—called **Workflows** that does it as well. But I could go into Copilot here, and it’s connected to Outlook and OneDrive and Calendar and everything else. So I can create a consistent thing here and save it as well. And then I can also even do visual things and create a workflow to create flyers. We did this with another cloud, about their football team creating flyers in there. So both of those. So this is the disconnect, and that’s why they’re making this Workflows one, where you can tell it to do those things, like here’s some, and it’ll just create that workflow for you. And that’s everybody’s problem. So you’re asking the exact right question.

“Right now the way they have it is an AI, I would say, it’s an agent. So if I was going to do this, I could build it here and we all have agents, and then it’ll be like, okay, if I want to maybe create weekly email help, then I have to pick one of these. They have all these different ones. Oh, by the way, they have ones that everybody already created and to do things over and over again like that. So that’s one way to do it.

“Another way to do it is just like this. Let’s do a weekly project update. So every Monday morning, review everything in SharePoint and make a project update. And then it’ll connect to my data and it’ll do this planning and it’ll make this app. So it’s in the background. It’s just going to make something that you can use every week over and over.

“So this—see, I don’t know how long that’s going to take. See, it’s setting up a workflow plan. It’s triggering events. It’s doing the same thing that Gemini was doing. And Marcia, without workflows, I know you’ve been doing these manually one at a time in just Microsoft 365. Do you have a good example, or do you want me to come up with one?”

> “Yeah. In Microsoft or in any of them.”

> “One of my favorite things is that because I lose track of my emails, I’ll create a workflow that’s every Friday that says— I ask it to do a couple things. And basically what we’re showing you, I put in the chat window, is you’re just giving AI an instruction as to what you want it to do in that chat box. That’s what a workflow is. So what I do, my workflow for every Friday, as I said, look at my inbox.”

“Now, there’s one thing I want to show you about Microsoft while she’s in there. See at the top where—”

> “Go back.”

> “Okay.”

> “That way, we can save time.”

> “Yeah.”

“See at the top where it says ‘Welcome, how I can help you,’ and right above that, it says **Work** or **Web**. When you’re clicked on Work, it’s accessing all your work applications and files. So you don’t need to tell it where to look, like in Outlook or where to look in OneDrive. If you’ve saved files in OneDrive, it knows it’s going to look everywhere it should look. So I tell it on Fridays, look at my emails. Are there any pressing ones where people ask me to respond that I haven’t responded to? And it’ll—or I can say, maybe I lost track of Reena. Reena emails me all the time. Maybe I could say, gosh, can you find all the emails from Reena, find a pressing one that I haven’t responded to, that I need to respond to within—you know—and it’ll find them.

“The other thing I do on Fridays, another workflow, is I say, create my to-do list for next week, and look at my emails, look at my—because we keep things in OneNote—and it’ll start to create a to-do list for things it knows I haven’t accomplished this week. So that’s the beautiful thing about using Microsoft Office 365 to do workflows: it accesses all your files and all the applications.

“And once you do this, it’ll save that chat and you can keep doing it over and over. So whatever that is, and then it will—I’m pretty sure, because this is in beta so we don’t see it yet in the edu accounts—but the Workflows is going to be, it changes it into an app format so that you can just open that app and answer the questions like it’ll already say, I’m your weekly email reminder tool, and you just answer the questions in there and it’ll create your output. So that’s the 2026 thing, Workflows that’s rolling out.

“So, yeah, if we did this—that’s why Google and Gemini edu and Microsoft 365 for us in education are great, because it does stay within one of those environments. A recent example is we were trying to—we were having all these discussions both in our meetings and via email as to how we were going to update our ChatGPT classes, and I lost track of where we were because we had so many emails going back and forth and we had so many discussions. So I just asked in this box, I said, can you tell me where we landed? Have we finished this, or are there any outstanding decisions? What’s next? Who’s supposed to do what? And it created this beautiful, you know, here’s where you are. This is what’s still to be decided, and then here’s the tasks that are outstanding by person. It was wonderful. And it looked in multiple places to create that for me.

“I’m just curious if I just ask it to create an agent if it’ll just kind of do it because…”

> “Yeah.”

> “So you can actually just tell it to make an agent and it’ll—let’s see.”

### Making workflows repeatable in practice

Yeah. So then you can summarize, categorize. So this is why, because it gives it to you in that kind of that answer response, like it's you're asking questions, it's giving you it's asking questions and then you have to talk to and get answers. So that's why so many people do this. But since it to make it repeatable, the workflow thing that's coming out will make it into a tool instead of just a conversation.

So right now, this is how you would do it: you'd have your question and then you'd have to be like so, like a summarize unread emails and let's let it flow through the OneDrive and do all that and Copilot. So there it's going to go summarizing a bunch of emails and telling me all the dates and whatever all. So, so you still get the same response, but you're kind of manually what I'd call a manual workflow at the moment, and that's fine. So, it'll go through and do everything that I need it to do. Did that answer your question?

We're going to do some more of these.

> yes, thank you.

Okay. But this is super cool. This is like way cooler because it makes that little app for you. So that's what we're get. I think all of the LLMs will have workflow functions sometime probably by the middle of this year I imagine. And sorry I keep going back to the beginning. It's how Microsoft works.

Okay. So we left off here. So the other examples we have are for leadership. So some prompts for policy for grant drafting and for speech writing. So the grant one is very popular now because as you know grants are a lot of work both in industry and in education. And so you can create a type of like a folder for each grant with the right inputs and then use your grant drafting workflow to give you the responses that you can paste in there. So it'll give you instead of, or you know the other way to manually do it of course is asking the questions one by one or putting the entire grant into your drive that you're connected, either your OneDrive or your Google Drive depending on the tool that you're using, and then you'd have to have another folder of inputs that you want it to refer to and then let it run. So it's making grant writing easier. I don't know if you guys are experiencing that, but that's what we see in industry and education for sure.

Okay. So responsible AI checklist, you have to make sure that you're iterating with all of these workflows that it's not doing a bias. So, you have to kind of test it out. As Marcy said, fact-check citations and then make sure

### Building accessible outputs and custom agents

things are accessible. All text for images, depending on the type of workflow you're doing. So these are three things that you should always be doing with every workflow depending on what you're outputting.

So, if you're outputting, one of my workflows I have is a weather generator. So it'll be, in fact I have it here. I think it is. Yep, there it is. So I made a weather agent in Gemini. And so I had it, I just like, here's what I did. So right in Gemini, let's make an agent to give me daily weather where I live. And then it said it was going to generate my HTML. And it did all that. And then I had a problem. It didn't work. And so it fixed the JavaScript. So again, I'm not doing any coding. I'm just talking to it in English. I said I want the map. Awesome. So now it has the map. So now I have that was my conversation. So now I've got this weather tool that I could share with anybody and it goes through and I have the map, all the map, and I can zoom in on there and do stuff and then get the weather. And then I've got this is my preview tool in my code.

So by the way, in Google, it talks about an API key. So when you're using something, a tool that you don't have, like I don't have Google Maps but Google has Google Maps, it asks you for an API key, which is a key that cost basically nothing unless millions of people are using it, but it's a key that ties to that other tool. It's just a piece of code saying, "Okay, I'm going to access Google Maps and put it in your little thing." And so now I can share this with anybody. My daily weather, whoops, my daily me weather map thing. And I lost my little thing there. I had to regenerate what I was doing. So that's, and all you have to do to share it again is you get a link and you can share it. And then the ED this looks different again. It's like within your work group. All right. Awesome. Awesome. One day I'll figure out why this goes back to the beginning and doesn't end up where I want it to be. Okay, we were here somewhere. All right, great. Okay, so the rest of this is exercises for us to try together, you to try on your own, and we'll have a little bit of time to do all these. So, Marcy, do you want to talk about this one before we go through the prompts? Um, sure. And here's a great example of the word accessibility again.

### Cloud-based tools and everyday integrations

This time it means you have access to it. All of these are cloud-based, so you do not have to save anything on your desktops. That is one of the wonderful things about them: all you have to do is click or save them as a tab in your web browser, and they are instantly available to you.

What we are showing is, especially with Microsoft Office and Gemini, how they are integrated with the everyday software you use. We just showed you how it is integrated with Outlook and all the tools in Microsoft. Again, they are prompt driven, so you can subscribe to tools that automatically create rubrics for you, or some of the things we are showing you. Yes, they are out there, but the beautiful thing about this is that all you need to do is enter a prompt, which is giving it instructions or asking the tool a question, and it will create all the rubrics and everything for you. So it puts the power into your hands by simply typing in a simple prompt.

You get to put it in your own words. Your prompt is your own words. Do not worry about the term prompt engineering or the perfect prompt. There is no such thing. Just tell it what you are trying to accomplish. Put it in your own words, and the AI tools will figure it out for you. So they are very fast and flexible and intuitive that way.

And then they are, more and more, designed to be for creativity. They are giving you things that are now more visual and generative. At first they were text based, so what you got was always a text description of something. Now these tools can generate interactive applications for you, or images, or PowerPoint decks, and so they are becoming more and more designed for creativity besides just being text-based, task-based.

> I love the podcast and the videos out of your information because you do not even have to do anything. You just say make a podcast or make a video off of it.

Yeah. So, you know, we used to see, a year ago, everybody would say, “Here are all the AI tools you need in your toolbox, right? You need this and you need this and you need this.” Now all you really need is Gemini or Microsoft Office, or the ones we keep showing you, because they now do everything.

#### Creative tool columns

I want to mention the other two columns because those are what I use. The hardest one, if you are doing creative work, is video generation, because it takes the most computing power. And I have said this before: the one that I use for a lot of images and a lot of videos is MidJourney. For $8 a month, it lets me do as many as I want. I think the limit is like a thousand or something. So that is the most generous, lowest-priced thing out there that works really well, and you do not have to learn very much. There is just a way to create an image, and then there is a button that says **animate**. So if you have seen my LinkedIn post, that is where I am doing that work from MidJourney.

Canva is what all the K12 teachers and students use. So in all of our dual enrollment classes, that is where they are expecting to see video editing, social media posts, YouTube thumbnails, and all that. So they come in already knowing that tool, and it is very AI-infused. It is also compliant with all of the things that we need in education.

The other two are HeyGen, which is for creating avatars. You see those a lot in ads now and also on LinkedIn: talking avatars doing things. Then Runway is the other advanced video editor available for audio and music. Suno kind of took over at the moment as the one that you can use to create your own music and then master it in there. For us musicians who use that professionally, that is how a lot of us are mastering music with instruments or maybe musicians that we did not have in our studio to add into our own music. It is fantastic. And that is why I created that prompt workflow for Suno.

Then ElevenLabs, it is a whole suite of things, anything related to text to audio. So if you have a whole lecture and then you wanted it to be spoken for you in 10 languages, you can generate that in ElevenLabs. But the main ones that we are using are Google and Microsoft cloud, ChatGPT, for most things that we are doing in education.

#### Everyday workflows

Here are some of the workflows that I do all the time, and these are pretty much change the way that I work, and I want to mention these. I do not know if anybody here does music, but Logic Pro and Pro Tools is what we teach in the colleges. Then mastering and Suno is a new skill that we teach students and we do ourselves professionally.

For presentations, some people still use PowerPoint. Some people just type in one of these other ones, create your presentations, and you can go through. It will make a PDF or even a PowerPoint output.

For games, this is one I am going to show you. I love Claude and Gemini for creating games, but there is another one that I use called Windsurf, and I will explain that in website design.

For enhanced learning, let us say in an automotive class you wanted students to learn, they are working on electric vehicles, and you want to know the different areas of how far you have to be from an electrical fire to be safe. You can actually make things to click and have it interact with what is going to happen in each of those areas instead of just reading about it. It is really easy to create in Gemini and Claude.

For website design, let me show you that too: Gemini and Windsurf. Windsurf is a great workflow tool because you can add custom graphics and music and sound effects, fonts, anything into your designs in folders.

And then image to video. I have a lot of examples in those, but let us go through the exercises first.

You want to do this one, Marcy?

So, in the beginning we had a slide that showed you the eight steps you take to build a workflow, and it is in the handout, the workshop handout we gave you. That framework applies to all AI tools, except you may need to give it additional information. So, for example here, and by the way, what you see on this screen is also available in the reference guide.

If you are wanting to create an image, you may need to upload a reference image, or if you want it to have certain colors, you can upload a branding guide or decide that the colors...

### Prompting, style guides, and workflow examples

or if you want it to have be a particular style, you may need to upload the style guide. So, basically the steps are the same, but what you upload, how you prompt, um what you look for in the out out the output uh varies a little bit. And then when you start to iterate with the chat GB tool to refine your output to get it the exact way that you want it, it's a little different if you're creating an image than if you're creating a video. Um, and then notebook LM is one of my favorite tools for developing curriculum. And so it works a little bit differently, too. So what what you're inputting, how you're prompting or giving it instructions is slightly different depending on the tool that you're using.

> And this is all in that reference guide. Pardon pardon me, Marina. That's why prompt engineering is tricky because it it's based on these kind of the output and you have to kind of work backwards about it.

One of my favorite tricks to all of this is looking at the galleries because these end up being visual outputs and you can see some beautiful work and then learn how to do prompts from there. So, I I do that a lot. If I have time, I'll show all of them give you a gallery of examples and if you click on that example, you can see the prompt or the instructions that were used to create it.

Yeah, I'm actually I want I do want to show that a second because it's so good and I love it so much. This is my favorite one. Uh Adobe does a good job. Some of us have Adobe, but this is my favorite new thing in MidJourney. They give styles and like I love these because I can go wow like you know. So, I mean, art is personal, but you can go through and you can even type a style. Like, if I type faces, it'll give me things about faces, but it gives me these artist styles. And so, once I if I can use these to start and they're and they have like a style reference number. So, before people were making websites with these numbers in it and trying to show what they were and now I can just try a style like this and then any prompt I make it'll be within this style. So that's um an example of a gallery that's super useful and and also it's super educational because you can see the difference in things that you can generate and it's made uh images a lot lot better. So in my create page like uh I've got like just kind of cool stuff from all these styles. There's me rocking out at a big rock concert. So, um, and students love these as well because it, you know, to be able to prompt and know those styles is it's that's a little lot more difficult. Uh, Adobe does a good job of that, too. But they all even Microsoft has that under their create button, which is a new button by the way. Uh, they have like if you create an image here, there's all the templates and if you hover over them, it gives you actually the prompt. So, it's and there's just hundreds and hundreds of these depending on what you want to create. And you can do your own brand and color. So, we have our brand in there. You can do infographics, videos, stories, like all this stuff. And it's all the all of those references are in there. So, we've come a long way in a few months on styles. I love it. I did it again. So, me and Microsoft All right. So, we left off here. This is this is just a quick example. So, we're going to do some examples, but just in Midjourney, if you take an image again and animate it, you can make really great stuff. That's the workflow I use in there. So, some examples. This is uh Claude mentioned does really nice um UIUX looking things, colors, accessible colors. Um, and a lot of times when I do apps and other things, they come out like in grays and not accessible. I have to go back and tell it to fix it. So, um, I made one that makes a workflow to make a prompt for, uh, to help you edit a screenplay. And so, that was done in Claude. And then I have a app. Then I can use it over and over over again. And it asks me to paste my script. And then um which I could change that to look at my Google Drive called scripts or something like that. And that's a workflow I made for that. Uh this is one that I made in Firebase which is part of Google. Um it's another app to make another place to make workflows in Firebase. Uh you hook it up. I I had it uh make a yoga plan for uh for myself based on restorative yoga. Really easy. Just open it up and tell what you want to do and it makes this and you can share it. Uh I did this one already that uh this is a I did a pseudo AI prompt in Firebase as well to compare it to uh what I did in Gemini. Uh this is an example of why workflows have changed. So, I actually created this. This is a good illustration. So, prompts on the left were like there's just a box and you get a prompt and you get some music and it's usually like not that great, but uh in the new workflow like they're considering what we were doing and what we were doing in music is making tracks and recording your voice and your instruments and all that, mixing it together and making a song. So now the new workflow completely changed um into Suduno Studio which is a workflow where you can talk to the tracks with AI like hey I want the drums to be more this different kind of beat or hey I want the whole thing to be faster or hey fix my voice and retune it. So it now it works like the w in the environment we're used to but with AI assistance and then so that's a huge workflow change in the music industry and that's why the music industry is adopting this as a tool that we accept in that industry visuals. So, uh, this is actually a picture from Machu Picchu that my daughter took last month. And then you can, um, in midjourney pick a style and then just tell it to animate those and it just does it. So easy workflows there. You all your input is just your picture and then you tell it what you want it what you want it to do. So that's really easy to use. Um, wind surf. It's an Does anybody here use Windsurf? It's like one of my favorite things. Um, I'm gonna just go into this for a second. Uh, I have, let's see, I guess the best way to show it I have it actually here. So in Windsurf you program uh you talk to it and you say hey I want to make a new app and what I made a game actually and I want

### Refining custom workflows with extra assets

to use some characters like I can make these any characters you want and then I want to have some sound effects and so what it lets the other apps aren't as good yet as adding custom stuff to your workflow. This one let you adds folders. Um where stop my app.

> So uh that's my Adobe. So it lets you um go here. It lets you make folders of custom things. So my I might make a folder called music, graphics, sound effects, protocols, rules. and I say refer to this and then make add those in there because if you say make a game and then it's not like my music or my sound effects or my characters or my graphics and it's just like whatever game it did.

So, in uh workflow advanced workflows, I'm using an app called Windsurf and educators get it for I think it's like $6 or $7 a month and it's it's another coding thing just like Claude and all that, but it lets you create those extra folders of custom content that it refers to and brings it in. Google Gemini is getting there with adding more custom content into their apps. So, pretty soon we'll just be able to do all that in Google Gemini, but um last year we weren't able to do that. It's coming. It's starting to get better now. Um because that's like it has to access your your desktop or your drive where all that information is and it's been really good at text. Hasn't been as good with graphics and music, but now it's it's getting there.

So, we're going to do a exercise down here. Okay. So, yes, we're going to ask you to choose a task relevant to your role. And here's some examples. And then ask whatever one of those AIS on the left um to create a first draft. And then we're going to have you do the second step. So, um you can choose one of these or something else. And then um you can like I said you can use chatb or Gemini or notebook LM whatever you prefer 365 and then we'll go to the next step. So I'm going to do this on mine as well. Okay, I'll do mine on my phone since uh got two screens. Okay, hopefully let me know if I'm going too fast.

Um, so here's an example that I did before. Um, output student instructions introduction to audio podcast audio quality. So, I made a I made a um overview of the assignment.

So, part two, now that you have your initial thing, um I have mine on my phone. So, add guard rails to improve your AI output. So, you're going to iterate on it and improve something about it. Might be the tone it comes out in or maybe change something for your audience. um because sometimes it comes out to be either too complex or not complex enough. Add accessibility to that. So any of these things and run it again and see what you get. I'm going to do that on mine. They should get a quite a different response.

So, here's a continuation of what I did. I did actually better one today. So, um I redid the prompt and then it made it more kind of specific and then added the learning outcomes. But I just did one where I did microphone patterns and I asked it to make it more student friendly and it made like all these really cool references to things in the lesson instead of it being so dry. So I like that it was I said uh what how did I reprompt it? Anybody else get like really different results on the second one? Oh yeah, I said make it more fun for students. It came out really good.

Okay, so here's where we get into kind of the workflow thing. So now you're going to create a prompt that other people could use over and over. So, rewrite your prompt to include your clear test definition, the limits of the prompt. Um, tell it not to make up stuff. That's a good one. Um, tell it that uh you want your output to require your own review and then run it again and see how that comes up. And that will be your approved prompt. So, you're basically creating your workflow um manually. That's what we're doing. This is fantastic.

> Oh, thank you.

> I've learned

### Why workflows matter now

### Workflows are the thing

> *“Workflows are the thing. If I had a banner on my new LinkedIn, it’d say workflows are 2026. Cool.”*

That was the note to end on: workflows are the thing. The idea is that, right now, there is a moment opening up where people are being pushed back toward actual making and doing. One speaker connected that to teaching robots and to students asking to be taught robots, and then widened it out into a bigger cultural shift. Their son-in-law thinks everything is going to crash in three years, but at the moment the trajectory seems to be toward a future where people spend more free time on things like knitting, woodworking, and cooking — going back to the kinds of real activities people grew up doing, instead of spending all day in cubicles and in front of computers, especially in industry over the last 30 years or so. In that sense, these tools are helping bring people back to being regular humans instead of sitting at computers all day.

#### Evaluating the workflow

The closing guidance was practical: evaluate your judgment. If something is not accurate, go fix your workflow. Someone should always be reviewing this stuff so it is not just autopilot. That review step matters because the process is not supposed to be blind automation; it is supposed to be checked, corrected, and improved.

Documentary workflow came up again as something already covered before, but still especially important right now while everyone is learning how to do this. The other key piece is sharing out: design these workflows so they make your work easier. The speaker mentioned that the presentation included some AI properties and principles, as well as some things on copyright.

Then came the closing invitation: if you can think about it, you can build it. What will you create? The point is to create things that make our lives easier and keep us spending less time in front of computers. That is what the whole workflow conversation is about.

#### Help, tools, and experimentation

The team made it clear they are available whenever people need help creating something and are not sure what tool to use, whether something is working right, whether it is not working, or whether there are new tools available. They mentioned tools like the ones in Microsoft 365 that can create workflows for you, whether you are doing it manually or using something like `Opal.google`, where workflows are being created and turned into beautiful apps that are easy to use. The request was simple: just let them know, email them. They are happy to help at any time, and they are there for the audience.

The handoff then moved toward Judy or Marcy, with one final question about whether they had anything else to add before that transition.

#### Manual and automated workflows

The response clarified that what had been shown was manual workflow creation: you are part of the process, and you are refining the output yourself. There are also newer tools that automate workflows for you. A guide from both Google and Microsoft had just been uploaded for those automated workflows. Those are the more advanced examples, the kind where, every Friday, the tool looks at your emails and responds to ones you have not responded to yet, and similar tasks.

That guide was uploaded specifically for the automated cases, and the distinction was emphasized: this is a much more advanced way of building workflows.

There was one important caveat about both tools: they are both in beta, meaning they have not been fully launched. Because of that, they may change. If you learn to use them one way, they will probably change before they are made more broadly available. Sometimes beta tools even go away entirely. A tool may be launched, not work well, and later disappear. That is the risk of beta. The only caveat is that you are experimenting alongside Google and Microsoft as they build the tools. Still, they are very powerful, and the team has been using them, so they believe they are going to be around to stay.

#### Industry use and session resources

There was also a note about how industries are already using these tools for data. One speaker said that at every industry meeting they attended last year, people were already relying on these tools for all their analytics. They thanked Judy for the guide and the explanation.

Then came a reminder for everyone: the session is recorded and will be posted on the LARK AI learning library. A link had already been placed in the chat. As had been mentioned during the presentation, some additional attachments and resources were shared, and when people click on the YouTube video they will find any attachments that were shared in the presentation in the description of the video.

The speaker said it was really fun to get in there and try these things alongside everyone, and encouraged people not to be afraid to experiment and see what the power of AI can do to help them and their work.

The next section was announced for January 30th, next Friday, and the link was also in the chat. Everyone was encouraged to share it, register, and attend. The session closed with a sendoff so people could get on with their day and try the new things they had learned. Thanks were given again to Marcy, Raina, and everyone who joined, with a look forward to seeing everyone next week and a final wish for a great weekend.