So the story goes, as I had to jump on this call and I had just an hour or so.

And even though I knew what's happening at a high level, I didn't have all the details.

And they wanted an architect to kind of ask me detailed questions about how does one of our security features work.

So the best I could do was to use AI.

So in this case, I pointed it to a Confluence page and I say, hey, you're best technical product marketer and, you know, punchy messaging.

Help me put here a couple of slides for this particular topic.

So one of them is focused on value prop and it has all these key points that we should be highlighting about how our authentication works in the particular product you're bringing to the market.

Ali, welcome to the show.

Thank you.

Thanks for having me.

I'm excited.

This is really fun to do.

So you and I have known each other for for a few years and you know, you obviously being a technology executive for a while worked at a bunch of different companies.

Today you're at Solis.

Maybe we can start with, you know, what is Solis, what do you guys do and and also what it, what is your role there?

Solis is a company that's been around for more than 20 years now and we've been helping a lot of different customers in different sectors.

We started from the financial sectors, helping our customers to basically connect everything in real time.

And we've now expanded to many other verticals as well, including aviation, retails, what have you.

So lots of large enterprise customers is our space.

I am Head of engineering at SOLAS, so the entire engineering team and support team is under my umbrella.

We build and support everything that Solas provides to the market.

This is awesome and so I've been excited to have you on this show for a while because I know you and I had had this chat and it was in the, you know, LLMS we're we're just starting to take off.

People were just trying to use ChatGPT and I think the conversation was you recognize that you need to like immerse yourself all.

And then obviously you've done machine learning.

You've kind of like been in that world.

But as things were accelerating, I remember you saying that you had a new routine where I think every day between 5:30 and 6:30 AM was your like AI deep hour.

Is that correct?

It is, and that routine continues.

So I'm reading the book called AI Engineering as we speak.

So every day I wake up, I read through that, whether it's a book or a catching up on your podcast or other podcast, to kind of keep going with what's happening in the market, help other people are using.

AII think it's super important to stay up to speed and not only personally grow those skills, but also bring them back to your organization.

Super cool.

I mentioned that only because things are moving so fast, even for those who are super technical like yourself.

I mean, you are also investing a whole bunch of time just to keep up with all the, you know, the crazy pace of things.

There's a lot that we're going to go through today and I think you were telling me this story about how you had a customer call that was coming up and the customer specifically wanted to talk about about security.

And you know, you, you knew what was going on, but you wanted to prep for this call and you have a really cool AI demo to to show us how you prep.

So I'll let you take the.

Yeah, absolutely.

So the story goes, as you know, I had to jump on this call and I had just an hour or so.

And even though I knew what's happening at a high level, I didn't have all the details.

And they wanted an architect to kind of ask me detailed questions about how does one of our security features work.

So the best I could do was to use AI.

So, So we do have an internal tool that we use for AI.

We started building this a couple of years ago when there was a bit of paranoia about giving access to internal information.

And I think all of that has evolved, but nevertheless, here we are.

So I used this prompt and the tool has the capability to go to different internal tools that we use for different purposes.

In this case, I pointed it to a Confluence page.

So here and I say, hey, your best technical product marketer and you know, punchy messaging, help me put together a couple of slides for this particular topic.

So the tool went, learned about that particular security feature, came back with couple of slide options.

One of them is focused on value prop and it has all these key points that we should be highlighting about how our authentication works in the particular product you're bringing to the market.

You're getting on a customer call, you have an hour to prep and you know, normally I don't think most people would, would think, let me go create a, a slide deck because you, you typically can't really do that within an hour, especially you want to look good in front of the customer.

You don't want this to be like, you know, an ugly deck with the wrong messaging.

And it's amazing.

So you're first, like the first thing that you thought is, oh, I can do this because I can use AI.

And you have your AI connected to Confluence, which is your wiki, which has a lot of information.

And then AI talks to the wiki and figures out what the slides should be.

Exactly.

Not only that, but also it produced me this chart that, you know, if the person on the call happens to be an architect.

This is a sequence diagram for people who are not necessarily familiar with the format, but you know, it really shows all the components involved in this particular security feature, how they're interacting.

And if an architect happens to be a call, this is exactly the kind of thing they want.

Wow, it actually produced this chart.

It produced this chart automatically by reading through the wiki, understanding it, and then producing it.

Did you ask it to make this chart or it just decided to?

So let's go back to my prompt and see what I asked.

So I need one slide that shows the sequence, yeah.

OK, you did ask.

For it sequence diagram for the flow and then I need one slide to highlight the value of our solution.

As you go through it, see that that's the value slide and the other everyone is the sequence diagram.

Hey everyone, hope you're enjoying the episode.

1 of the common things that we hear is hey, you all talk about all these different tools.

You walk through these demos.

Sometimes it's hard for me to follow along.

We decided to do recently was take all the things that we talk about, put them in a weekly newsletter.

We literally list all the tools, we link to them and any demo that we walk through, we break it down step by step, put it in this newsletter so it's super easy for you to follow along.

And the other nice thing is that if there's someone on your team that you think could benefit, they're working on that subject area.

They could use a little bit of AI injection into their workflow.

Send them the newsletter, send them that particular episode.

We make it super easy to do that.

It's free to sign up.

All you have to do is go to thisnewway.com, enter your e-mail address, and you'll get this weekly e-mail from us.

Hope you enjoy it.

With that said, let's go back to the episode.

When you think about what this means, I don't know that people need prep time anymore or like you would have said, OK, I need more prep time, can we push it to next week?

Exactly.

Try to come up with a better schedule for the meeting or what have you.

So for those that that are watching, what is this AI tool that you're showing them?

So this is our internal AI tool, which by the way, after we built it, we decided to bring it to the market.

We call it the solid Asian mesh.

At surface, it may look similar to Claude or ChatGPT or what have you.

As you see it has a chat interface, but behind the scene it's a multi agent system.

So you have many different agents.

As an example, what we talked about was the Confluence agent.

You can have a GR agent or other types of agents integrated to different to different internal tools and then you can start interacting with the tool.

The tool decides what is the right agent that is supposed to solve your problem.

I would love to dig in further about this tool that you guys have built and, and all the, the unique things that it brings.

But I just want to emphasize for everyone who, who's watching, you can do something like this with, with Claude or Chachi BT at What's nice is that most of these tools, you know, Confluence included or whatever wiki you have, chances are there is a connector, like a native connector that exists with some of these tools or there's an MCP server that you can connect.

So the point is, I think that, you know, the conclusion is if you have a wiki and you're using cloud or ChatGPT, you should connect these tools.

And the next time you're going to do a customer call, you should do what Ali just showed us, which is you can literally create a slide deck just by prompting it in a clever way.

And and not only should you do it, you should get everybody on your team to do stuff like this too.

Absolutely, absolutely.

To emphasize on what you said, I think MCP is really evolving the way we are starting to use these tools because almost every single vendor is coming up with their own MCP tool, whether they provided remotely for remote access or enable you to deploy to your own tool.

It really opens up the world, the existing world to the world of AI.

And I think it's MCP.

The great job they did with MCP, it's coming from Anthropic, is that they came up with a clever, clever way to enable you to reuse your existing investment and bring your existing context to AI and make AI so much more meaningful as opposed to, you know, write me a story or, you know, come up with a funny picture or whatever.

Now you can actually use it in the context of business.

What did you do after this point?

So you have the the slide content and is there a tool that you can like take this and turn this into a slide deck or was it copy and paste from here?

So you could either copy and paste.

So let's let's try that actually copy and paste.

Now one thing that I realized as I was working on this, let's go maybe to I was.

Going to say these slides look pretty fancy.

You know these are not AI generated are?

They no no no, these are not AI generated.

But one thing I recently discovered is that Microsoft Copilot has this option that you can insert a slide using Copilot.

Interesting.

Let's try create slide for me about security with the following.

So I'm rolling the dice a little bit here.

We'll see what happens because as you know, yeah, because my has.

Their own AI too and and you know it's in PowerPoint, I think it's coming to Excel if it's it's not there, I've seen their CEO tweet tweeting about it about how you know now you can use AI inside of Excel too so.

There you go.

So you got, you got a starting point of a slide, which is actually not OK.

It uses the theme and everything and all you need is to move things a little bit around, maybe resize the image and.

And it created that image, right?

That's not your it.

Created that image on the fly, so some more formatting which I don't want to bore you with, but I'm off to races ready for my customer presentation.

Yeah, this is, this is pretty cool.

You know it, it kind of makes you think, right.

The next time there's, you know, you message someone on your team and, and, and you say, hey, there's a customer call in an hour and they're like, oh, I won't have enough time.

They're like, no, no, no, go watch my interview.

I'll just tell you, you have enough time.

Exactly.

Lots of opportunities for sure to not only, I think in my opinion, be more efficient, but also elevate the quality of the content and whatever we are doing using AI, right.

Totally.

I mean, I, I think like this is a big point because it's not just about speed.

I think when all this stuff first happened, everybody was thinking speed and less work.

But now I think you know more of the impact I'm I'm seeing at our company, Ali, and I don't know if you would agree is a lot of it is just we're doing things that we wouldn't have otherwise done, right.

So for example, maybe we would never have created a slide deck for every customer because it would have been too much work.

But now, now you can actually do those things.

So I find that we're getting to things that we just would have never gotten to before.

So it's just that the quality of the things we're doing is higher.

Yeah.

And I think the we are lowering the bar for experiments.

So like taking this to product development and experimenting with new features betters making making the overall product development process more efficient or the coding part of it more efficient.

I find sometimes it's easier to make a decision and say, let's just do an experiment with this feature as opposed to in the past.

You're like, oh, you know, this is going to take a month or so.

Am I?

Do I really want to invest that much to see if it works?

So now those kind of decisions are easier or even things like on the engineering practice, deciding to refactor a piece of the code that you think needs some love because it doesn't have the higher, higher high quality that you you want to have.

With AI assisted coding, it's easier to kind of jump into those kind of projects now because the cost again could be.

Could be lower.

It's really changing the game.

And so one of the other things and you and I have chatted in the in the past about is just again, Saul is a big company.

I think 500 or more people, you're, you're an executive there.

You have tons of meetings, you have lots of projects, you have to meet with customers, You have a lot going on.

And so I'm curious, what kind of stuff have you done to just get AI to help you just manage your day-to-day better as a busy exec?

You just saw one example of it, but let me show you a couple of more examples again using.

So here's another interesting example.

I'm sure a lot of you have had to book meetings with like 10 people who are busier than you and can be the hardest thing to do in the world.

So this tool again, can connect to our internal HR tool and it can look up all the vacations.

So as an example, I wanted to organize a leadership workshop.

And for the life of me, it was so hard to find the right time to do this.

So I wrote this prompt.

I'm like, hey, listen, you are my executive assistant.

I do want to organize a leadership workshop.

Don't book it the last week of August because I'm on vacation.

I don't think I've booked this in HR system.

Avoid the middle two weeks because I'm on a business trip.

Avoid Canadian holidays and by the way, check Mars Calendar.

The trainer was going to do the workshop for us and make sure that she's available and check all the other direct reports of mine.

Yeah, just you describing this makes me just like overwhelmed.

And I'm like, please, this would be like the worst.

This is like torture for me.

Just.

100%.

Listening to this.

So this thing and you see it on the right side, what it's doing, it's going to a chart tool, It's looking at all the individuals.

I'm not going to get into details of that and bore you, but it comes back eventually and says, OK, well, I've checked everything and I think your best bet is September 3rd to 5th because everyone is available, all 12 participants.

And here are the dates you should avoid because some of the other people are away.

And ultimately it gives me some recommendations like, OK, well that that was easy.

Yeah, it's amazing.

You can do anything now.

In the past, if I wanted to do that, it would take me like half a day and by the end of it I'd be disappointed because I probably wouldn't find the perfect time anyway.

This is pretty incredible.

I can just only imagine, you know, I, I saw another example of this which is so.

So again, this is everybody should go.

I think Claude by default for sure has a calendar connector like you can actually connect your Google account.

I I think you can do this in chat 2BT right now too.

But even something as basic as go classify my week, right?

Just you know how many external what percentage is external?

What percentage is internal?

What percentage are one on ones?

What percentage are group meetings?

And it can like do that analysis for you.

Absolutely.

Connecting these AI tools, whether it's Copilot, if you're a Microsoft customer, or other tools that connect to Google Calendar or your flavour, and then using them for all, all kinds of things is super powerful.

It frees up your time.

It literally acts as executive assistant and help you be much more efficient.

The other cool thing, I know Trachi PT has this where you can do these schedule tasks, right?

So you could do a thing where every Monday it can go analyze your calendar and like run the same prompt on your calendar and tell you, you know, last week you said you want to have less meetings, but it was actually a record week for meetings.

That's an example.

You could do all sorts of great, but yeah, this is really cool.

What else do you do?

Yeah.

So let me show you one more that I think it's quite fascinating actually how far we've come with AI.

So that the challenge here is I have to give a project status update to IRAP.

It's a body in the Government of Canada which provides funds for various projects and what have you for those who are not familiar.

But basically every once in a while we have to get together and give a progress update.

And the format of the monthly reports that we provide is unstructured takes.

So these are these Word documents, they're not necessarily stored in a certain structured documents.

So you can't use your regular BI tools to generate data visualization.

So what I did is I wrote this prompt, which is a little bit more involved than your usual prompt.

It uses this sort of XML format to guide the AI through extracting the data out of that unstructured data.

And then eventually it gives a specification of a heat map to give A, to give a status update of the project.

And this project, the way it's structured is it has a whole bunch of objectives and each of those objectives have activities.

So this thing went into that Word document, it's Ms.

Word document, It extracted all the statuses, the percentage of progress, and produced this heat map, which at one glance gives you a very quick update on the project.

Wow that's crazy.

I actually had no idea that you could create a heat map using an LLM.

Yeah, yeah, it's pretty cool stuff.

So in this particular case, what's happening is there's a tool that produces the visualization.

So what the LLM does is it knows how to interact and create specification for that tool and then it feeds that specification to the tool and that the tool as an agent renders the visualization.

What is the tool that's being called?

Here I think in this case we're using mermaid.

Mermaid.

OK, that's pretty cool.

Yeah, yeah.

So lots of examples.

And you know, back to your point about being busy and all that, imagine that day that I was prepping for a customer meeting, but also I had to give a status update later in the day.

In the meantime, I had to schedule that leadership training.

All of that stuff I pretty much got done in in a span of one hour when in the past it would have taken me half a day for each of these to accomplish.

This is pretty cool.

I mean, I would even say, for example, say that you weren't, let's say that you delegated this task, right?

And you know, you show up to the meeting and you have the person on your team who he worked on it.

If they showed up with a heat map like this, you, you'd be like, you are so above and beyond.

You're promoted.

I'm promoting you right now.

But nobody would have done it.

Nobody would have ever created like a heat map for a project status reporting.

So this is pretty crazy.

How do you so, OK, so you discovered this, you know, not discovered, but you know, this is a really cool workflow.

How do you get other people in your company to learn about these, you know, super interesting things that you discover.

You're on the edge, you spend a lot of time on this stuff, so how do you teach everybody else to think the way you do at the company?

So we are doing multiple different things.

One of the things that we've done is we've started this group called AI Champions at Solace.

So it's a group of, I would say about 20 people or so.

Everyone is super passionate about exploring and learning about AI, and you're opening up the group for everyone else to is also interested to join.

But the point of the group is for all of these people to be on the bleeding edge, explore, do different experiments, and then bring back those learnings to the rest of the team.

And they've done a whole bunch of things, including putting together trainings, creating Slack channels for continuously sharing what they're learning.

But we also have official lightning talks, So small little talks, not a lot of prep that people just show up and they show what they're doing.

I know you guys do the same thing in your all staff meeting where people share what they're doing with AI, Similar concept, but this is like a meeting dedicated to sharing about AI.

And we have many people showing up, learning and sharing.

So it's quite exciting.

Can you explain how that works?

So these are lightning talks.

Is it every week?

Is it every?

Yeah, So Lightning talks in general, like they've been around.

It's a concept that's been around for a while.

The idea is you just show up, you give 1015 minutes of talk, you don't have to prep anything.

You just show your workflow.

Like we just explored together a couple of workflows here.

So the lower the bar of prep, so more people participate in sharing, basically that's the idea.

And then anyone else who wants to learn shows up and learn from each other and create this culture and environment that everyone wants to share and learn from each other.

Super interesting.

And so is it an invite that goes out to the whole company and it's optional or?

Yeah, it's an, it's an optional invite at this point.

It's my organization.

But yeah, we have plans in the future to expand it further.

Hey everyone.

Just a quick pause on today's episode to tell you about my day job.

In addition to this new way, I'm the CEO of a company called Fellow dot AI, and Fellow is an AI meeting assistant.

It joins all your meetings.

It summarizes them, tracks the actions and the decision and does that better than any other tool that you've seen.

We've spent a ton of time making sure that the meeting notes and summaries and action items that come out of Fellow are the most accurate, the most precise.

It beats any human.

You have to try it.

But in addition to that, what makes Fellow different is that it is the first AI note.

Take care built from the ground up with security and privacy in mind.

This means that you can use it for all your meetings, not just the customer facing ones, but also the sensitive ones.

Things like one on ones and executive team meetings and those QBRS and everything in between.

It's got really good judgement.

So for example, if you start a meeting and you're talking about some social stuff that you don't really want on the record, that's going to get e-mail to everyone afterwards.

Fellow just knows.

It just doesn't include those things.

Or say you have something that you talked about and later on you realize, oh man, that shouldn't have been there.

Makes it really easy.

You can go back to the meeting, select that part, delete it, and then it's gone from the record.

It connects to all of the other tools you use in the organization, whether that's Slack or Asana or a HubSpot or a Salesforce or Linear or Jira or Confluence, whatever it is, Fellow integrates with all those things.

I think the best part is that Fellow also acts as, as an AI chief of staff because it sits on all the meetings and the conversations that you have access to.

You can ask it really cool questions such as what are the biggest opportunities in my company?

What are the bottlenecks in engineering?

And Fellow just pieces together information, sees trends, and they can answer those sorts of questions.

He can even do things like, hey, based on all of the one on ones that I've had with this particular person, can you create them a performance review?

And it can do things like that too.

And the Sky's really the limit.

What I really wanted you to do is have the opportunity to try out fellow check it out.

And we're making a special offer available to all our listeners.

So just go to fellow dot AI slash this new way to try fellow.

There's a discount code in there for you if you decide to continue with it.

Either way, I would love for you to try it and let me know what you think.

And with that said, let's go back to the episode.

One thing that we haven't gotten through yet is so you guys started, I guess built this, this internal AI and you know the, the beginning, I guess the, the reason in the beginning was you, you just didn't want all the organization's data to go to, I don't know, random LLMS all over the place.

You wanted to centralized and you start working on this and then you you turn it into a product maybe like tell us about the product and like what it does today.

Yeah, we started building this.

And then as we were building this, we step back and we are like everything we've been doing at Solace is actually relevant to AI because if you think about AI and how you might, you can make it useful in context of a company, it's all about that context, bringing that context of business to AI so AAI can do something meaningful for you.

And for the last 20 years, Solace has has been in the business of providing real time contacts and data movement, moving data from point A to point B.

So you felt this is a very good extension of our business and we can actually very meaningfully integrate some of these capabilities into our platform.

So that's exactly what we did.

In addition to that, we took a vary from a technical point of view, took an approach of using our existing platform to build this tool in a very scalable way.

So we are using event driven architecture behind the scene.

We believe the right way of getting agents, AI agents to talk to each other is through event driven architecture because it enables you to scale, it enables you to have many, many agents without worrying about those problems.

And we do believe the future is many agents.

We think any organization will have hundreds, thousands of agents.

And when it comes to our customer base, we think there's going to be an explosion of agents, which if it's not done the right way, people are going to pay the price for it.

As an example, just a parallel to that, the industry just went through 1015 years of building microservices.

And what we all learn and we have the scars to show for it is that if you don't get these microservices to communicate to each other the right way and the right way, again, you believe is event driven architecture, then you're going to have very fragile architecture that may explode at any time in, in the operational environment.

One microservice can bring down the other microservice and what have you.

So agents are, if you think about them, are just yet another type of microservice, which is why we think a lot of those lessons that we learned over the past 10-15 years, they the industry learned are very much applicable to this world of agents Takea.

For the non-technical audience, basic, what you're saying is that what what you've kind of your company is focused on is just communication of different systems.

And now we're going to live in a world where agents talk to each other.

And someone needs to orchestrate the way that they talk to each other so that they don't, you know, drop the ball, right?

You know, humans can drop the ball.

Agents can drop the ball too.

So you need to know that they're doing the things that they promised, that they're communicating back when they've done things.

You know all that stuff, right?

Or, as an example, what happens if an agent A agent A wants to talk to agent B and for some reason agent B is not available?

Are you going to lose that line of communication?

Are you going to lose that flow, agentic flow?

So how do you make sure this all works in a robust way?

There's something called an agent to agent protocol.

How does this relate to that?

So we support A to A.

So Google donated A to A's protocol that they come up with to Linux Foundation.

OK, the tool that we've built uses A to A protocol to do communication between agents, which we are super excited about.

In fact, if you go check out the Google's partner page H2H, we are on the partner page as well.

Oh, amazing.

So this is like an internal AI.

You can connect all the different tools in your organizations and then it just supports this more advanced way of talking to agents.

So everybody's always on the same page.

Balls never dropped.

So what kind of things can you do with it?

Do you have like a cool demo to show?

Yeah.

So I mean, you just saw some of the demos that I use on a day-to-day basis.

Those were all done on Solace Agent Mesh, but let me show you another demo.

I think one of the things that we are super excited about is what we call event triggered assistant.

And while you're putting that up, I think like the other thing I know about Solis is I would say, you know, critical systems, right, like banks, you know, our customers of, of Solis as an example.

So you know, when I think about this or I'm projecting, you can tell me if I'm wrong.

But if you have an agent that writes A blog post and then you know something goes wrong and the blog post doesn't go up, your business is not going to go bankrupt, right?

But if an agent is going to do something at a bank like transfer money or something, you want to make sure that happens.

Exactly.

A good example I was talking about with one of our airline customers yesterday is they have a claim system.

So you know, you and I travel a lot and you know when you lose your luggages, God forbid or something happens, you go to this claim system to to open a claim.

So what they want to do is to be able to automate all that workflow in in an agentic way.

So obviously they don't want to lose those claims.

They want to make sure this end to end workflow properly works and gets to a conclusion.

They want to make sure there is no hallucination along the way.

They want to make sure they can integrate this system into various different internal systems, such as the login system, to get the necessary information to process that claim.

So those are the type of use cases that you're talking about with customers.

When you say something like, you know to do it in an agentic way and plain English you know what does that mean in an agentic?

Way, yeah.

So let me show you an example.

So this demo is sort of a mini version of that, just from a time constraint point of view.

A lot of us know what is JIRA.

It's a ticketing system.

So in this particular case, what's happening is a customer comes in and opens a support ticket in JIRA.

They say I'm missing an order.

So you know, they purchase something, the order is missing, they provide a very limited information.

They say, hey, my name is Jim, Jeff, I'm from Midtown roster.

So let's take a look at the ticket together.

Very limited information.

They're not giving any order ID or anything like that, right?

So here's the agentic workflow.

So what you're seeing is that incoming ticket triggered an event.

It triggered an agentic workflow.

So the system is going through all these different agents.

There's ACRM agent, there's a support agent, there's an Atlassian agent.

You look at the relationship between the orchestrator, the manager of the system, and all the other agents on the screen.

So let's go back to this visualizer, which really shows everything that's happening behind the scene.

The CRM agent goes and gets more information about, OK, so who is Jeff?

Do I know anything about this?

Can I actually concretely find out who is this customer?

Then after getting some more information about that, it goes to the support system and tries to track down some of the orders that Jeff has put in.

And then it goes to the Atlassian agent and puts a comment on that support ticket that just came in.

Oh wow, that is really cool.

So that's an agentic workflow, right?

So you suddenly got back a comments.

So imagine in absence of that, someone in your support team had to go figure out who's this Jeff guy, find all the orders that they put in place and come back with the potential order that was missing.

So the agents did all of that work for you.

So they just came back.

They put a comment and say hey, I think this is the customer who is talking to us.

Here are the recent orders they had.

I think the most likely missing 1 is the following one.

Right here is some background for you and all that information.

And then not only that, it gives you some actions and also response that you can just copy paste and send to the customer.

So right there the customer support team saved probably hours of digging through orders, finding what's happening, and then well crafted message also done there and ready to go.

Wow, It's really cool.

It's like a luxury where I'm gonna go make a decision on this particular customer case, but I want someone else to go look it up and all the systems and do all The Dirty work and just bring me all the information and then I'm going to do the last 2%.

Absolutely.

One of our customers who rolled this system out, they rolled it out on top of their order management system.

And so the business challenge they have is they have about 100 to 150 people who their sole job is dealing with disorder management system.

The challenges the customers have and they claimed that they went from hours of investigation to answer tickets like this.

2 minutes.

Again, it's pretty crazy.

Where does this product go?

Where do you see it in in a year We continue to add.

Let you know, you mentioned earlier that there are many tools who provide different connectivity to different internal systems.

I think one of the things we are definitely working on is making sure we have a great set of robust connectivity to enterprise system, the systems that typically our customers use.

So they can easily bring all these context necessary to achieve the workflow such as the ones we were looking at.

But we can imagine these kind of work flows to be used in banking to be using aviation as we just talked about or any other retail business.

So I we think the opportunity for creating efficiency and quality of work in our customer base and internally at Solace is huge.

So.

A couple of interesting things that you you pointed out.

One is this idea of an orchestrator agent, right?

Just this again, you, you refer to it as as almost the manager of the organization who has specialized people any and then it decides who it wants to call on to do the next step of the process.

I mean, that's a that's a pretty cool thing and and you guys have built it in to your product, but what other kinds of scenarios Can you imagine you know, that you would use with an orchestrator agent?

I think the sky.

Is the limit really, because think of the way I like to think about agents is almost as digital employees.

That's one way you can look at it.

So if you think about an organization and how an organization grows, I think you can map that to the agentic world as well.

So let's think about that.

How would you scale an organization?

You start hiring people.

They all have different special skills and at some point you need someone to kind of get the project, break it down, assign it to the right people with the right skills.

And that's how you organize an organization of humans.

You can alley the same concept to agents and that's how you scale it.

O it's really important that you build a system that can support many agents.

You can have hierarchy of agents, you can have different domain expertise, maybe even for each domain you will have your own orchestrator.

So those are the kind of things we are thinking about, I'm sure people are thinking.

About quality, right, So you know, how do you make sure that like what kind of testing you put into place as someone who is deploying agents in their company to make sure that they don't screw up and don't give the password to the wrong person or no, I think that's really.

Important.

So, you know, if you look at people that talk to or listen to people like Andrew Inc, who is very famous in the agentic community, he always talks about evals as being one of the most important part of rolling out agentic systems.

That's been our experience as well.

You know, one experience that we had was even when you upgrade your models to a better model, sometimes they can surprise you because the prompt that used to work with an older model may not work as well with a new model, even if the model from a performance point of view on paper is supposed to outperform the older model.

So my biggest recommendation for everyone is take evals really seriously and an e-mail is like.

A test case, but just for AI so.

You know, taking back, taking it back to lessons that we learned from software engineering, think of eval as almost like an integration test.

It's exercising those scenarios that are important for your business and making sure that you get the response that you expect to get.

So being able to have this run in an automated way frequently and as you evolve the system is really important.

Otherwise, given the nature of the LLMS that are more predictive systems, this has become becomes very hard to test and make sure that they produce the the the results that you're looking for and then are are you guys?

Did you build your own eval eval framework or are you using something?

Yeah, exactly.

So the.

System that we have comes with some eval infrastructure as well that enables us to run those evals.

Oh, OK.

So the customers.

Actually, so you encourage them build this agent, but you can also build these evals using your tools.

So they can like build and test all in the same place.

Exactly.

I mean, this is pretty crazy.

I mean, all the things that it can enable really it's I think for everybody here.

You almost need to take a step back, evaluate your day, maybe use AI to help you evaluate your day too and, and the tasks that you're doing.

And then look at each and everyone and think, can I automate this or parts of this?

And start with yourself and then maybe then go go to your team.

We all have to reinvent the way we work 100%.

You know, back to the demo that we started with the preparation for that customer.

One of the things that I keep telling my team and I think everyone is starting to build a habit for is as a knowledge worker, whenever you want to start any task, try to have a trigger in your head, have a habit in your head to ask yourself, how can I do this better, faster, with higher quality using AI?

And I think if you keep repeating that question, in my experience, the answer is at this point, given how much, how far these models have come, the answer is yes, I can do something better now.

Maybe the first try is a little bit harder because you're stepping out of your comfort zone or maybe you haven't crafted the perfect prompt yet, but eventually you find the right answer and back to that heat map that we were just looking at, you know, I mean, I learned something I.

Didn't know you could do that.

That's pretty cool.

I'm very comfortable.

Now giving those updates and it takes me a minute to prep for them.

This is what I'm doing.

This afternoon is I'm going to go connect the mermaid tool and I'm going to start building charts.

I'm going to impress my team and they're they're going to think you became a designer overnight.

That's a beautiful chart you created.

Now this is super exciting.

So final question we, we always like to end on.

What is the thing you're excited about the most over the next year?

But I'm really excited about.

Is how fast AI is evolving and how I can integrate it into my work flows as an individual and also my team.

I think sky is the limit.

I think the opportunity we have to apply AI to everything we do in software development.

Work flows is huge, and I know a lot of people are focused on coding or coding assistant, but that's not where the bottleneck is anymore.

I think there's a lot of stuff outside of coding that could take advantage of AI.

So for me, looking at every single thing and see how we can improve it using AI is definitely rewarding.

And then, you know, using our own product to do that is of course even more exciting.

Yeah, this is great.

Ali, thank you so much for doing this.

Thanks for having me.

It was.

Exciting.

It's always a pleasure to talk to you.

And that's it for today.

Thank you so much for tuning into this episode of this new way.

If you like the content, be sure to rate, review, and subscribe so you can get notified when we post the next episode.

See you next time.