Transcript

Hailee Zapata: Welcome to our webinar on Google Cloud Next 2024 and the Gemini era. I'm Hailee Zapata. I'm the Alliance Marketing Manager here at Promevo, and I will be your host and moderator today.

So, over the next 40 to 45 minutes, we're going to talk about what are our biggest takeaways from Google Cloud Next 24, and we're also going to do a quick overview of all the announcements and updates. Then we're going to move on to the Gemini era, all things Gemini. And of course, we're going to end with our live Q&A with the presenters.

So if this is your first time with us, I do want to do a quick overview of who Promevo is.

Promevo is a Google premier partner. We sell service and build Google products. We are a hundred percent Google focused and we specialize in ChromeOS, Google Workspace, Gemini, Google Cloud, and we have our own proprietary Workspace management tool, gPanel.

All right, enough about us. Let's go ahead and meet our stars today from Promevo. We have John Pettit. He is our Chief Technology Officer. Colin McCarthy, our Change Management Leader. And Aaron Gutierrez, the Practice Director for Data Engineering and Analytics.

All right, John, I'm going to hand it over to you to start talking about Google Cloud Next.

John Pettit: All right. Thanks, Hailee. Thanks everyone for joining us today.

Google Next this year was a massive event definitely back in person. There was over 30,000 people there present at the Mandalay Bay. And I think most of us walked probably over a mile, 26 miles over a couple of days. It was definitely an energizing event to be there.

And true to form following last year's announcements of Gemini and pushing to AI, Google basically put Gemini on everything, right? So everywhere you turned, you saw Gemini in every product and every presentation, every partner booth. Definitely front and center.

What I think was very interesting to see is the shift from people talking about Gemini to having very real experimentations, plus pushes into real production workflows. And we're seeing that happening at big clients transforming their operations, whether it's in retail or banking or automotive. Just about everywhere. This is definitely something that's here and continuing to grow and evolve.

There was a lot of interesting innovations in Workspace that Colin will talk about, and I think some of those push the boundaries of the next generation of productivity, not just tools that you use for Documents and Sheets, but how these things all really start to come together and connect into to drive more productivity and automation in your workflow.

Behind the scenes to support all this, google's invested heavily in TPUs, tensor processing units, mega compute instances using NVIDIA's latest technology, and they've also reinvented the entire kind of AI stack to make AI training more efficient and faster. Just reengineering everything from the hardware up to the software layer.

They announced the public preview of Gemini 1.5 Pro and Imagen 2.0, and we'll talk a little bit more about that as we get deeper into the conversation. But the race to produce more powerful AI driven by more compute power continues to happen.

So they said, they could continue to throw more hardware and data at it, and it would get smarter and better. And that continues to happen. But, they're also taking advantage, as we see, of this multimodal capability of AI to not just interpret text, but interpret video images and speech across the, across all the data that you can throw at it.

Across migrations, we still know that people are out there on legacy systems, mainframes, sometimes data centers, and they've invested more into capabilities and tooling to help people get old VMs across or get out of data centers or to just bring data into, the cloud where they can do more rich analytics. So if you have an old mainframe system being able to migrate that data across.

And I think lastly, the thing we've seen is on the security side, Google continues to invest in security. So the acquisition of Mandiant has really started to come together into Security Command Center Enterprise.

So you can manage, bring together threat intelligence, plus the always on security scanning, plus the ability to use Vertex or Gemini to give you more insights and intelligence to the and response into the security events that you see in your systems.

So I'll hand it off to Colin McCarthy next to talk about the Workspace announcements, which there were a lot of, and then we'll continue on and go through data and cloud further on the presentation.

Colin McCarthy: Hi there. Great. Thanks, John. So yeah Google Workspace, Chrome, and Gemini announcements at Next.

A lot of IT admins and Google Workspace administrators are obviously very happy and excited about Next, and it's only been eight months since the last one. And obviously this is the a new format that Google are doing, but with Vegas and far more attendees.

I wasn't able to go, but there was a lot of great coverage available. Streaming watched their keynotes, the developer keynotes, and then Catching up with a lot of the sessions. It looks like virtually all of the sessions were available online, which is great.

As John said, the big announcements was Gemini and the introduction of two new SKUs. So for those of us that have ever looked in our admin panel under billing and subscriptions there's sections of Google Workspace you can click on and find out additional SKUs that are available.

So before we had Gemini business and Gemini enterprise, but Google announced two new SKUs, the AI Meetings and Messaging, and also AI Security. And if we go to the next slide, we can see it a little bit more detail about what is in those SKUs.

The most interesting one I think is the AI Security. Which is going to really help organizations automatically classify their sensitive drive data.

So those of us that have used labels and DLP rules know that there are some limitations and restrictions for how that information is classified. With the AI Security, the large language models will actually be able to be trained and fine tune to your specific needs, and we'll be able to automatically categorize and label your drive data, and then you're going to get better accuracy with your DLP rules and also context aware access policies.

So that's a very interesting one. And is a very good justification to have that. I would be hard pressed to find any CISO that would be able to argue against having that sort of capability in your platform.

The next one, which a lot of us have been using is AI Meetings and Messaging.

So this is... some of it is cosmetic, I feel. There is a lot of studio lighting adjustment, optimization of your camera. There's some smoothing of your image so you can always seem your best. If you have a call at 9am or 8am on a Monday morning and you're tired from the weekend, you can click a little button and remove some of those black bags from under your eyes.

The most exciting part of this SKU update is the the notes function within Calendar, Google Meet where it will actually create notes for you during your meeting or document everything that was said and then also create action points as well. And we've been trialing this out for a few months as a trustee tester and an alpha, and it is really good. So that's one well worth one investigating.

I guess the biggest announcement was the announcement of a brand new product. So, vids.google.com, the AI powered video creation for work which has come out of Threadit. Everybody believes, which was an area one 20 product.

If anybody's knows about area one 20, it's this sort of skunk works experimental part of Google Workspace where they test things. And all the same sort of video functionality seems to be in this, although the platform's not available at the moment, still are coming soon.

This is the biggest announcement or the first new product since 2016 when Jamboard was released. So we do rarely get a new product within Google Workspace, but very excited to see how this is going to change, how teams communicate, collaborate.

I know certainly on the project management, change management, I can see myself using this when we're doing updates to the client about the state of their that their project it might be more dynamic to watch 90 second or a three minute video rather than read a long report. So I can see there'll be a lot of grounds for that.

The next big update, which some of us might've missed, it was announced, I think, in the keynote was Chrome Enterprise Premium. So it was announced as a new product. Although if you look under the hood and you check a few links, it looks like this is more of a rename of BeyondCorp Enterprise under the Chrome Enterprise Premium banner and also the Chrome Browser Cloud Management part of Google Workspace, which I always found difficult pronouncing. It's probably why everybody calls it CBCM has been renamed to be Core and Chrome Enterprise Core. So very excited to see what's happening in the enterprise browser market.

Obviously, there are a lot of other players out there that are using the pure Chromium to build their own Chrome based or Chromium based browsers. There's a lot of lot of players in that new enterprise browser market, and I think it's great that Google is bringing everything under the Chrome branding, but I'll be very interesting to see what's in the Chrome Enterprise Premium, if it is a brand new browser, not a consumer browser with some of the other components in there, but a true enterprise about a browser. So that's very interesting to see.

And then some of the other key announcements within Workspace. All, a lot of these are still Gemini AI based and a big focus on mobile. There's been a lot of improvements to the mobile app in iOS and Android and how you could work with Gemini.

You can now do a speech based prompts, which is great for those of us that have been already used to talking to our phones and entering texts. Now you can enter that prompt within Gemini, on your phone, in an email, and it will help you, right? So really great when you're on the move.

Also a new building block blocks and notifications in Sheets. Now I know a lot of people are very excited about this. As a PM, I'm very excited about conditional notifications. That is really going to revolutionize how change managers and project managers utilize Google Sheets to track all of their projects.

And then the next update is something that other admins are actually pinging me about and asking questions is a Docs. Now that you're going to be able to get multiple pages through Docs almost taking on OneNote, maybe Notion, for note taking and have additional tabs, pages, tabs to really build out Docs.

And then lastly, Google Chat. They've increased that with some quite large capacity. I can't believe anybody would want to be in a busy Google Chat with half a billion members, but it'd be great for a, an announcement channel and also streamlining the communication between other platforms with their partner Mio, where you can actually communicate with Slack and Teams. So, really bridging the gap between all of the other platforms that are out there.

So that was very exciting. And I know a lot of us here at Promevo are excited to play with some of the things and use them in our own enterprise.

John Pettit: Hey, Colin. I think, yeah, when I've talked about this next generation productivity at the beginning, this kind of stuff is, it seems minor, but the stability to have tabs inside of docs and building blocks in Sheets, it's really building on the Google story, right, what they started last year with web versions of Docs. It's not just formatted as something you're going to print, but this like collaboration and these things become really integrated applications to you managing your work.

Where do you think they're gonna go beyond that?

Colin McCarthy: Yeah, it's interesting how you main mention page lists in Docs.

I've been using that a lot and made that my default 'cause realized I wasn't printing out any documents anymore and had all of that wasted space. I've got extra white space to build out my documents from going, vertical as they used to do in a, and in the old typewriter days, when actually printing, typing something out to send it having something more dynamic, it'll be...

Yeah, I don't know, it's gonna be interesting to see how vids.google.com blends in and if that continues to be a standalone product or is more experimental and then elements of that get integrated into Slides or Google Docs. I'd be interesting to see if Docs and Slides didn't need combined. And certainly with tabs within Docs, you could build out your your presentation within Docs.

Now, certainly with the full bleed cover images, I think was what they say. So you can have a presentation almost like a fully branded Google Doc. So yeah, that, that might happen.

John Pettit: As they add this additional content too, it's interesting to see how the sidebar has started to come into play where in their demos at Next you could do the @ and then reference a Drive doc and then have that be the context for your Gemini prompt.

Colin McCarthy: Yeah, like smart chips, smart clips is incredibly powerful. I have started to use them more and more. I think, they have just, it's such a wealth of data that's out there and I think all of these abilities, within the whole of the Google Workspace suite to be able to call on different elements and bring it all into one, just makes it so much easier to manage what you have.

Certainly be interesting to see what happens with between now and even IO on what might get announced then.

Next, I think it's back to you and or Aaron about Google Cloud Platform and Gemini.

John Pettit: Yeah, I'll talk about the Gemini pieces and then Aaron can go into more depth on data.

So if you move forward one slide, I think, the biggest announcement at Gemini Pro is really this content window. So the ability to have 1,000,000 tokens and hour video or 11 hours of audio or 700, 000 words, right? It really changes what you can analyze and how far you can go. But beyond that, under the hood for Gemini 1.5 Pro, they also talked about native speech recognition, being able to natively look at the images in a video and treat that as content that you're analyzing and using as your prompt.

And they also added in a file management API, which means I can send a file to Gemini Pro as the prompt without having to pass in additional text, right?

So I think this, beyond just the reasoning capabilities that have improved across these models, the content window has gotten much larger for bigger problems to solve.

So you start getting away from this need to do like LangChain and chunking and all these approaches to solve large data problems that you want to inject context into the, these engines to get really powerful summarizing and reasoning out.

Also, slightly lower in the mentions was that you can just tell it to give you JSON back. So you can tell it to respond with a JSON object. So for developers, you want to use this as an engine and you don't want to have to try and do weird prompt manipulation where you're passing in a JSON template and telling it to respond with this thing. It natively will handle that now. So Gemini Pro is moving the bar forward, but Google also realizes it's not always the best solution for every AI problem.

So their integration with Hugging Face APIs, their integration with Anthropic APIs for Cloud Opus, or even just being able to get cloud models in the model garden. It's pretty impressive. And depending on whether you're just trying to summarize, certain types of data, those may be better solutions at times.

So not every problem needs a large language model. Sometimes you need a small language model, and they're being open to letting developers use their stack and their technology from a Vertex standpoint to manage. All their problems within Google. So that was a really cool announcement. Really neat to see some of the demonstrations.

One of the things they showed was like, okay, you could have it live analyzed, streaming game, gaming data. So I don't know, like I have kids, right? You're playing games. It's cool. It's I'm an hour younger and you have to do like game guides. You're like, Oh, how do I beat this thing? But just imagine this world where you can have, AI always on watching your data and giving you analysis is getting closer and closer to reality, which is pretty cool.

Aaron Gutierrez: This this expansion of the amount of tokens you can use with the language models reminds me of the era when RAM started exploding for for your, just, you could consume your computer. Like, before you had to be real careful on what you could run, you had to look at the back of the box.

Remember when software came in boxes?

John Pettit: Yeah.

Aaron Gutierrez: Yeah. Look at the back of the box and say, Oh, my computer will maybe run this. I don't know. But, now you can just assume whatever program you use on the web or whatever you download works on your computer because everyone's got, eight gigs of RAM, 16 gigs of RAM.

This explosion of the amount of tokens that could be consumed by a language model is to me feels the same way in that regard.

Like you mentioned chunking and when you're a developer, sometimes you had to take your files apart and make them small and send them one at a time to be processed. Now you can just be lazy and just throw your whole huge file at Gemini and it'll process it.

I don't know if that's totally good. You might become a little lazy in your in your development practices, but it does feel like it's opening this huge resource pool up to, to all the possibilities.

John Pettit: Yeah. Yeah. And just seeing them increase the modes of data in this multimodal world where it's just text, right?

There's all kinds of image, video data out there, audio data. And the more these language model engines can handle, like the more problems you can solve with it. So it's pretty neat to see that.

I'm sure you'll talk a little bit more about some of the other kind of tooling that they're putting in behind the scenes when it comes to models, when you get to your piece.

But, just hearing about some of the things around like prompt engineering and fine tuning and the things that they're putting into the stack around that makes it not just a, oh, it's an API to a language model, but I have all this other stuff behind the scenes that Vertex is going to power for me.

Need to hear this when we're working together. It's really becoming more cohesive that you're in.

Aaron Gutierrez: Yeah, now they can consume all sorts of file types that's just going to open up worlds of routes you can take when you know, using your data. You don't have to transform it all into a table anymore. That, before, was what you had to do with everything.

You had an image. Somehow, you've got to change that into a vector to put into the table and then work with it there. That's not necessary anymore.

John Pettit: Let's jump forward to the next slide. So from tooling standpoint as a developer there, they've rebranded the duet. For developers and all the stuff, it's all Gemini now.

So they have Gemini Code Assist, which is the like VS Code plugin or JetBrains plugin that allows you to look at your existing code and provide suggestions for tests or code or how to, how to write some of your code to get you started. It's definitely expanding that 1 million token windows that starts to get in there is interesting because it could start to look at. Probably like your whole code base for context and start getting, way better in terms of what it does.

They've also rolled out their application life cycle, which is basically like a cloud assist inside of GCP console to help you more effectively manage your cloud operations and ask questions of the data and the logs.

They added on their security expertise, which is what we talked about a little bit with Security Command Center Enterprise and be able to ask questions around events and things that are surfaced in the tool.

And then they have a bunch of data tools, and I don't want to jump in too deep, but just like I said, Gemini on everything. So Aaron will talk about some of those in a minute on the slides.

Let's jump one more forward the hyper computer stack again. So as they look at this, you have performance optimized hardware, the open source software they're contributing to for you to solve the machine learning problems, and then the scheduling so you can optimize spend at the top of the layer.

But with their new performance, they've basically doubled or tripled performance from where it was before in terms of the overall stack. And that's, computer optimization, hardware optimization all the way down.

What's interesting though, is when you look at some of the public studies, like the machine learning training consumption needs are going up in like a log scale. So they're going up 10 X. So all of this is really mandatory just for people to continue to follow this a I boom and growth. So this is just the beginning.

And this benefits all of us, whether it's AI or not, just the amount of investment and cloud hardware, infrastructure, and things that support for the applications and workloads that we want to run.

It also brings the cost down for all of us is Google's investing in this and the AI boom. All the other stuff has become more efficient and cheaper for us to run. So just interesting to see the the change in evolution of cloud from what it was to simple computes and VMS to becoming really AI purposed hardware and software workloads.

We can jump forward. So this is Aaron, I think it's going to kick over to your section to go into stuff I've been teasing. But yeah, lots of exciting stuff

Aaron Gutierrez: Yeah, there was some pretty cool stuff out there for for data people, for analysts, for people that are behind the scenes on some of this stuff.

The first thing I wanted to bring up, something that was really cool. It's a hybrid tool. I don't know if you guys saw this in any of the keynotes or presentations, but they've developed this, this BigQuery, this canvas tool, which I don't know if you guys have worked with databases much.

But especially for a new employee to a company, like sometimes you're just staring at the stack of data tables everywhere, right? Like you've got this data set, you've got this project, whatever, like it can be overwhelming. In the old days, what would you do? Just start select star, show me what's in the table. Look at this, look at the schema, right.

But this data canvas tool, it's pretty neat. They've wrapped Gemini into it where you could start asking it, tell me a place where I would find my order information, tell me where I can find, details of my customers and it'll find the appropriate tables.

And then it'll like if you see on the screenshot there, it'll show you like, the schemas and then allow you to just start pivoting off and joining onto other tables using natural language, like now that I have my customers and my orders, can you show me like, who are all the top customers in Texas or whatever, and it'll actually do a little visual join on the screen.

And to get even further into that, you can say, all right, that's cool. Build me a bar graph from that. And it'll, on the same canvas, I don't know if that little animated GIF does it justice, but it'll just keep expanding, like Colin was mentioning, onto a... what do you call it, Colin? Paperless? Boundless?

Colin McCarthy: Pageless, yeah. Pageless. In Docs, yeah, it's called pageless.

Aaron Gutierrez: Yeah, so the canvas works in the same fashion. It'll just become pages and it will expand off to the side stacking nodes on top of each other. And one super cool thing about it is if I was digging in the weeds in this data and I found something cool, I could use like the built in Google sharing stuff like you do with a Doc or Sheets. Just share, Hey, John, look at this. And John can come in and start playing alongside with me at the same time, as you do with a Sheet or a Doc and start building onto this little, ad hoc analysis that you're working on.

And from there you can just, you can uncover a lot of things and as an analyst, like sometimes that's the hardest thing to do. Like, you get stuck in boxes doing the same analytics over and over because you really haven't had time to sit down or the tooling to sit down and just look through this stuff, dump all the puzzle pieces out on a table and see what you got to work with. But this BigQuery canvas gives you that power.

I don't think I've seen many other tools like it in other areas, yeah. A lot of the collaboration exists in some of the other tooling, but not in data yet. I'm really excited to play around with this. And John, once I built some cool bar graphs or findings, I'll send them over your way.

It also has natural language query building. So when you tell it, Hey, show me the top customers in Texas, it will show you the query, it will build the select from so and so table, with the where clauses and grouping and all that. So if you actually want to tweak that, you can, and we run the little module, it'll spit out the new result.

But I think that's a really cool thing for probably new users, for sure. You want to see how to build a specific query. It'll show you, but also for an old guy like me. Maybe you've always been doing something a certain way. BigQuery will show you a couple of variants of how it will build that query and then maybe teach you some new tricks.

Yeah. Go back to that slide, there's one other thing I wanted to mention. We, I skipped over at the top. BigQuery has, I don't know if it was directly at this Next, but maybe in a prior next they've integrated Jupyter style notebooks into the BigQuery environment.

Now they've rolled out their custom BigQuery data frames, which is, if you've used pandas or polars or py, anything, I guess any of those libraries in Python. Those capabilities are built into BigQuery now. So if you prefer developing in notebooks, it's a lot of data scientists and analysts prefer that's now, built into the BigQuery platform, which is making it super robust in terms of a toolkit for data analysts.

John Pettit: That was one of the things I was most excited about is this BigQuery Studio becoming more than just a query tool. Everything that they've brought in here with the notebooks and like being able to find the schemas, it's super powerful.

Aaron Gutierrez: Yeah, it's just robust. It's a nice tool set. Now it's not just a little SQL IDE window where select whatever, that's all I can do, right? Yeah, now you can build your notebooks and you can start, prototyping things and working on models and a bunch of cool stuff inside of BigQuery. You don't have to take it off into some other tool. You can just do it all in the BigQuery environment.

So, all of that stuff looks really cool. And that was one of the huge things I took away from Next.

All right, now we can hit the next slide. One of the other things that they showed off at Next was how Gemini is also working its way into the BI tool Looker and Looker Studio.

One of the cool things I wanted to point out to listeners on this call is they have this Looker Explore Assistant extension available right now. So if you if you wanted to go give it a shot, you could go to the GitHub listed there. This allows your your Looker platform to start taking advantage of Gemini.

And I, and I, and working with your data with natural language. It gives you a little screen where you just type in your natural language query alongside of your Looker Explorer. It'll generate, it'll generate the looks for you. It'll show you insights to your data. You can ask it questions about an existing dashboard, and it'll summarize it.

So it does a lot of a lot of cool natural language interactivity inside of the Looker platform. So again, they're really working on like bringing. Gemini to the tools, which to me is really awesome.

There are two options if you don't have your data warehouse sitting inside of big query, which a lot of people do. You can use the cloud function method, but they also have a BigQuery deployment option.

So if you wanted to just integrate everything totally within the GCP ecosystem, You have that option as well. So that's something we're really looking forward to the couple of our customers have asked about it. So we're trying to get those tasks out the door installed into their systems.

Colin McCarthy: Aaron, I know a lot of people love the explore feature within Google Sheets. It'd be great, and I think that has been depreciated or probably removed. Now I'm going to be replaced with Gemini features. It'd be great to find out, and I'll have a look if this explore feature will be available in data Looker Studio.

For those of us not using the full Looker, but using Looker inside Google Workspace.

Aaron Gutierrez: Yeah, I believe I did see some demos of that at next I'll keep an eye out on that just to see if I can send that your way, but I know like you and John both mentioned, they're really trying to push Gemini into a lot of these other tools and BI for sure is one place that can really use that natural language prompting. It's easier to work with the data when you just ask the questions, right?

Colin McCarthy: Yeah, or even get started. And, sometimes you don't know where to get started to build your report. But if you can articulate and speak your report requests in the natural language, then have the data to refine and tweak. It's a great benefit.

Aaron Gutierrez: Yeah, the cool thing about natural language ability is that you can take very informal approaches.

You can probably say something like, show me something cool about the order data and it'll, it'll try its best to give you some meaningful insights. And then you can always just say, no, I don't like that. Try it again. And, keep tweaking your prompts until it gives you. A good starting point to begin your analysis, right?

Colin McCarthy: Almost like the I'm feeling lucky button in Google search. I'm feeling lucky with my data. So tell me what I don't know about my own data.

Aaron Gutierrez: Yeah. I take that button as a challenge.

Colin McCarthy: Yeah. Yeah. If you were trying to hide something from the federal government, please create a report.

Aaron Gutierrez: Yeah. It'll do what you tell it. So it's pretty cool.

A couple of other things that. I wanted to touch on not specifically data, more of the machine learning type of products were at the keynote or the new features of Imagen. And I don't think our slide deck shows it very well, but this, this new functionality of Imagen was pretty cool where you could just tell it a prompt, the first one's supposed to be show me a flower blooming and it'll show you a little, it'll create a video clip of a flower blooming. And to me, that was pretty, pretty cool to see how realistic they can generate these images.

If you need some marketing material, just to show like a little splashy for example, the bottom right is a I think it's a curry. If you're trying to build something for your website or something to show, your mini offerings, this is something that you can just quickly go in and pop out some, some cool collateral materials.

It also rolled out ,I put it in quotes because I think magic erasers, like the Samsung branding of what it's called. But like they've added two images, the magic eraser functionality, and they've also added a, I didn't know what to call it, magic adder.

They showed a scene and they wanted to enhance some of the mountains in the background. So it was pretty cool. They just used their cursor, and like it was able to generate like a huge mountain range behind the image as opposed to a little hill that was there.

But they are rolling out a lot of cool features for Imagen, some watermarking tech. So if you are creating assets, you want to make sure they're, they're linked to you.

So prompt evaluation tools are pretty neat to see. They, they have this I guess this console where you can start asking a bunch of prompts and it'll tell you, it'll give you grades and allow you to grade the results as well.

To say, no, this is not good. This is good. So it'll save your prompt history. So it actually is teaching you how to build better prompts.

John Pettit: Yeah. They talk about that. The using the LLM to look, evaluate prompts and give people feedback. So basically, like you said, it's we're not really training the AI. The AI is training us now.

Aaron Gutierrez: Yeah, exactly. And I like the console street we had, where they had your historical results and showed you, Hey, this style of work. And maybe I think you could see other users as well. Like John writes, his prompts, very professional. Colin's got this, he's got this personal manner to him, but Colin gets better results. Maybe this is the way I need to try it.

John Pettit: The the prompt engineering is something you can't overlook, right? Like even as we've used Gemini internally, we'll compare results sometimes and people will be like, what did you use? How did you get that result? And it just, that prompt really makes a difference of what you put in is like positive and negative references to things.

Aaron Gutierrez: Yeah I, my personal style is very caveman. Like I type a word or two words.

Colin McCarthy: I think prompt will be 2024. And I think one report said the average number of words people are putting in is nine, but the optimum is 21.

So you really do have to rewatch a few episodes of Star Trek Next Generation and pretend you're on the bridge and you're talking to the computer.

You've really got to be specific or you can be very conversational and, and the more information you give it, with the persona, task, and the context then, and the better results you're going to get.

John Pettit: Yeah. So somebody is going to address this as a negative prompt for Halloween. Is that what I'm hearing?

Aaron Gutierrez: Hey, that's a good idea. Use this stuff to get Halloween ideas.

John Pettit: That's really exciting to see how creative you can get. So they imagine stuff. I know like it looks Oh, it's just generating images, but just how much more rich you can make your content. And seeing some of this come into vids is exciting because some of us are, got a creativity challenge. I think so like myself at times, so being able to make my stuff look nicer, more professional is always exciting.

Aaron Gutierrez: Yeah. It's just amazing that they've done all this with, if you think about the math under the hood, this is all just, vectors and neural networks, just doing this stuff. It's amazing.

Hailee Zapata: Yeah, I wanted to let everyone know when we send out the actual deck, I would love for everyone to go look at this slide because you can see the flower blooming and the curry boiling, and it's so neat. I don't know why it didn't transfer over to StreamYard, but I highly recommend everyone to go look at it.

As a marketing person, I'm super excited for this tool. It's going to be so helpful with creating images and just the details that goes into it. So I just wanted to pop on and say that real quickly. I'm going to hand it back over to you there.

Aaron Gutierrez: Yeah. Yeah, it's impressive for sure.

Colin McCarthy: Yeah. Yeah. And for a marketing person, the watching the demo of the creativity for marketing, the tent and the whole flow they went through doing that, was really good.

So I think it's going to be really good for any size organization, certainly for those small businesses that, want to create their own engaging, unique marketing, it's very good.

Aaron Gutierrez: Yeah this is cool stuff. Especially like John said, if you're not artistic, this is just amazing to me that this stuff can be created with words.

Okay. We'll move on to the other final really big thing that I noticed. They're really, I guess they're triple dipping back into the agents.

The, they used to call them speech and conversation. Is that what they were called? John?

John Pettit: Yeah. Speech and search API.

Aaron Gutierrez: Or something. Yeah. Yeah. The agents were a huge focus of the of the keynote because these are tools that are really adding a lot to, the customer experience for for the, for Google clients.

And they've actually created these these six archetypes that are very specialized to specific things. So, they're really making it, making the process more user friendly on the people that are interacting with the agents and as well on the developer side as well.

I, I personally feel like building an agent in the past with the prior versions of the dialogue flow systems were a little... they were sometimes a little cumbersome, a little, like you had to really get in there and customize all the routes and paths and pages that you had to do. And now they're actually letting the language models take a little bit more of the wheel on this and letting the intelligence of the archetype drive the user experience. So, it can tell when a user's asking a question that's related to, whatever the conversation is.

For example, they showed orders, sample orders being done in the presentation. If a user starts talking about something unrelated, it'll steer them back. You don't have to act. You don't have to build those routes. It will automatically drive the conversation back to the point of the conversation.

And they've done that by adding a lot of the other Gemini features into the agent toolkit. So it's able to ground the conversation, right? It's able to only speak with information related to the topic. It will steer the customer back onto topic. It's really adding a lot of this, the smartness that language models have into the agent system.

And on top of that, they've also shown us some pretty cool demos where you can actually pick up the phone, and it has a pretty realistic conversation back with you, and you can specify if you prefer Colin's accent to mine, to my nasal accent, you can have an agent that sounds a little more sophisticated. And or customized to your clientele.

So if you've got, if you have customers in England, they're probably going to be more familiar talking to an agent that speaks their accent.

So this was just very impressive in, in the resources they're throwing at the agents and how simple it is to, how simple the development process is now versus what it was a couple of iterations ago.

John Pettit: Yeah, that was a big part for me on this too is, if you're going to build a rag before and you had to set up a vector database and run your data through embeddings so that you could then search it and then bring it into the model. Having this, just be able to point at data types and that all goes away.

And then you can just pass in, like you said, any kind of mode, images, audio, whatever, and have it go and work together with the context of the data it finds and give you back answers is really neat. Also, that kind of like off branch of it interprets a request that needs to call one of your functions that you've described.

Aaron Gutierrez: Oh, yeah, that was cool.

John Pettit: Custom code function is pretty neat, right? It's more modes of data that you could expose to the agent that you couldn't before.

Aaron Gutierrez: Yeah, you can write a custom function to talk to any API that has that functionality. Yeah, that was really cool.

Colin McCarthy: It will be interesting to see how all of us get to benefit from this as end user consumers, because everybody hates dealing with those bots.

And the first thing you normally just, try and type in is speak to an agent. So now the agent can actually be useful and be helpful. And this could be a very good for everybody.

John Pettit: Yeah. I can't wait. Those other like old, like dialogue flow based ones that were so like rigid. I can't wait to see this stuff be way more adaptive and useful.

Colin McCarthy: Yeah, those ones were like an old text based adventure game where you didn't type in the exact prompt, you went nowhere.

John Pettit: Yeah, this is gonna be exciting.

Aaron Gutierrez: Yeah, so really excited to see how these things work, especially going forward, how much better they are than the previous versions.

Hailee Zapata: I want to thank you guys so much for walking us through all of that great information around Google Cloud Next and Gemini. I know I came out with a lot of questions from it, but we went over a lot of information.

Please feel free to reach out to us. Here's all of our contact information. If you just have a question about an update you can just email updates@promevo.com. We also hope you could check out our website. It has all of our solutions and services provided. And at the very bottom are links to our latest blogs on Gemini.

We have a ton of information, a lot of blogs on Gemini, actually, on our website as well. Feel free to check those out.

And then lastly, before we move on to our Q&A, I wanted to point out on the right side of the screen is a list of our upcoming webinars. Right now we have a lot of gPanel Office Hours coming up.

gPanel is our own proprietary Workspace management tool. So if you're interested in, or you have Google Workspace signing up for this monthly gPanel Office Hours is a must. We go over so many things coming up, we have gPanel API, gPanel automation, and then delegation. It's once a month, the second Tuesday of every month.

Okay. Now let's get into our Q&A. I know we had a couple of questions, some that came in previously before we had the webinar, we give people a chance to, at updates@promevo.com, send in questions. So let's start with our first question here.

So what advice do you guys have for organizations trying to evaluate Gemini for their business? And then how can Promevo help in this process?

Colin McCarthy: The best thing would be to reach out to us at updates@promevo.com, and we can talk to you about a Gemini AI pilot where if you do want to evaluate Gemini, we can work with you on a four week program to give you access to trial licenses and then develop a program to walk you through doing office hours, check ins, getting you to make the most out of Gemini "help me write," "help me organize," and "help me create" in Docs, Sheets, and Slides. So yeah, reach out to us about that.

Hailee Zapata: And you can probably show a company, because I know even in Promevo, we all use Gemini, like it can literally help every single part of your company, correct?

Colin McCarthy: Yes. Yeah. We'll do a discovery, understand what you're looking to get out of AI, how the business operates, what the challenges are and then, show you how you can use it to improve communication, analyze your data, get back some time, improve collaboration.

And then, there's so much more coming to Gemini within Google Workspace that it really does set, companies up for future success.

Hailee Zapata: Colin, this kind of leads into the next question. Or a question that was asked about what do we do to help properly analyze the tool within a company. I think we're talking about the workshops that we've started to offer. Can you touch base on those a little bit?

Colin McCarthy: Yep. So the workshops is A larger engagement than purely the pilot where we can help you do a rollout. We can manage all the change management, do a lot of end user surveys, feedback, evaluate where people are using it how they're utilizing it.

And as we said, I'm even amongst ourselves and as John's seen, it is that prompt magic that, although the prompt proficiency if we can call it that really gets the value out of Gemini. So it's helping users across the business to share information about how they're best use utilizing it and changing their workloads.

Hailee Zapata: Let's go on to our next question here. Okay. Since we're talking so much about AI, can you all speak a bit about Vertex AI and its capabilities? And how does it relate to other development solutions like AppSheet?

John Pettit: Yeah. So Vertex you can think of as just the overarching platform to talk to a bunch of machine learning.

Capabilities that Google has, and those could be Google's models or other models, but it allows you to solve just about any problem you have that comes to machine learning.

And it also is a gateway to Gemini, right? So if you have a problem that's appropriate for a large language model where you need more complex reasoning and you don't want to try it, it's not something you can easily build like a deterministic flow to, but that's a gateway to that.

AppSheet is a powerful tool that kind of empowers your end users to become citizen developers. So you start giving them technology to solve their own problems.

Now, what's interesting to me is when you take AppSheet and then you start connecting that to Gemini, now the users have the way to solve their own problems and include Gemini in some of the solutions and answers, whether that's auto filling fields, parsing data that people are inputting, or making a decision on the next step to happen in an AppSheet process.

So you can start giving people the power to be really powerful developers. Usually before, if you're a developer building something, it's I've got to break this down into an algorithm. I've got to figure out how I'm going to write the code out, how I'm going to construct all the error handling conditions. AppSheet takes that away, and then Gemini allows them to even take it a step further.

These things can work together, I think is the point on it. But Vertex is, if you have a problem you want to solve, like Vertex can solve just about any, complex reasoning problem or machine learning problem that you need to solve.

Hailee Zapata: So we actually recently had a client reach out and ask about a couple questions on Gemini, and one of them was, can Gemini create custom workflows and custom chatbots for specific use cases? Is that something that's going to be available soon? Is it something that is already live? Do you know anything about that?

Aaron Gutierrez: The agents definitely give you that ability to create chatbots.

I think that's really what their original bread and butter was just building chatbot systems for companies. Now that Gemini functionality has rolled into the agents that's going to give a lot more flexibility, a lot more human like responses. But yes, definitely that functionality is now available in the agents.

And that's something we can help with if you have a specific customer that reached out asking that question. That's something we could help implement for this,

John Pettit: Whether that's like an external chatbot or somebody wanted to put that into spaces. All right, so one of the interesting demos we saw next was them bringing like a dialogue flow chat based interface into chat spaces for even your employees.

So I want to ask a question and have a natural flow and kind of workflow built around that to solve problems, whether that's, Hey, like I need to get an approval request for a client to get a discount on this new deal. Something that simple. You could have that go and evaluate and ask questions of them and get down to a final answer and then just determine approval or not approval.

You can bake into something just by telling Gemini what your reasoning is and chaining the prompts together in the right way.

Yeah, lots of powerful things that could happen in terms of just internal workforce automation.

Hailee Zapata: So thinking about this, like how integrated Gemini can get and how it monitors and analyzes. Security is always something that comes up.

It's always something I like to ask about I know someone asked about specifics of the security add on for Gemini announced at Next, but also just security in general. What do you guys feel about the security of Gemini being business owners and things like that? How do you talk about that?

John Pettit: I'll just start with you. You're better off using Gemini integrated into Workspace than having users go off the range and use. A free version of Chat GPT or something else because you have zero control in that scenario of what kind of Information they're leaking out and you have zero control to monitor and manage the overall thing.

So Google's keeping your data private. They're not retraining this models on your data that you're using through these paid Gemini subscriptions. And the same thing happens when you use Vertex, if you're building models around there. So just from a base level of you don't want information leakage in your company, shutting off access to the, or the non paid versions of AI that people are using is a much safer place to be.

Uh, beyond that, you still have to do the basics of acceptable use policies for your employees training. And so they get the most, value out of it. And you still, don't want them to, I would say. You don't want them promoting other people sharing data over unsecure AIs, and you don't want them posting files to Chat GPT 3 5.

Just in general, you need to keep it locked down.

Hailee Zapata: Yeah, that makes sense.

Colin, what about the specifics of the security add on for Gemini Workspace that was announced at Next?

Colin McCarthy: I really think it is going to be its ability to better understand and interpret the data that you have, through its large language model and better categorize it labels, and it is a way of categorizing your data and creating DLP rules on it.

But then that's only as good as the manual labels that are assigned on the documents or the some of the limited automatic assigning of labels, the ability for the model to be trained and understand your data and understand what is sensitive and classified for your business and more accurately label that data, so the correct DLP rules get applied is absolutely, is really great.

And I think all of the security areas and functions will just get better, better functionality and features. And one of the good things that, you know, as we talked about with Aaron is that ability to use natural language in your reports.

So, instead of trying to dig into a report or the investigation tool to find the data that you want to investigate something that might be suspicious, you could just use native language. I think that's where it'll get so much easier for every admin eventually.

John Pettit: On that, the, since you brought up the kind of reporting that Security Command Center with Mandiant and the Gemini piece together as Security Command Center Enterprise does unlock some of that understanding and response and intelligence to give you more of a modern SecOps. And so I think that's going to be really powerful for people.

If you have any kind of application out there on the internet, just the amount of data that comes through the logs that you're supposed to be filtering, finding and responding to, right? Nations data tax and other things that continuously happen that you have to go block IPs. Having any kind of additional tool there can help you respond to that is, is definitely a valuable.

Hailee Zapata: Wonderful. That was great guys. I just, I know we had a minute left, so I just want to ask each of you real quickly, what is the announcement or product that you are most excited for coming out of Next?

I know that's hard cause there's so many, and we're all geeking out about what's going on, but what are you guys most excited for?

Colin McCarthy: I'm going to say tabs in Docs. I think that's really gonna change how we collaborate and store data.

Aaron Gutierrez: For me, BigQuery Data Canvas. Absolutely.

John Pettit: Yeah. You guys took the two best ones there.

I'm going to go with vids. I actually have a desire to see vids work out and become something that is just true to the product and helps with the creation. So I'm excited about the future of that just because I have to create a lot of content all the time. So I'm excited to see like where that goes.

And being just like a step above slides, right? Can we get them past the world where we just live in presentation decks and we have something that's more interactive? That to me is exciting.

Hailee Zapata: Yeah, I'm with you. I feel like I live in presentation decks. So I actually was, when we were talking about the vids, like we should do a webinar on that when it comes out and how you work it and all the technical stuff.

So yeah, that's awesome. Thank you guys so much. I know we're at time. Thank you everyone for joining us today about our Google Next recap and all things Gemini and the Gemini era. Again, please reach out to us with any questions at updates@promevo.com.

Aaron, Colin, John, thank you guys for your time today and walking us through all of this great information.

Thanks again, guys. Have a great day.

Colin McCarthy: You're welcome. Thank you. Bye.