Hi, everyone. My name is Stacy. So usually I volunteer, but as Vineet said, I'm AI Operations Lead at HFS Research.

And I'll get into that in a bit, but I kind of want to spend a moment talking about MindStone. So if you're not familiar with MindStone, MindStone is like Joshua Wall has started it.

He travels North America and Europe, excuse me, doing events like this. But what he also does is he has a platform teaching people AI and how to do it practically.

So I encourage you to go to Mindstone's website and you can kind of learn and they have education tools and ways to kind of approach AI in a practical way.

All right, so with me, I just joined a research company. I've been in research for four and a half years, research and advisory, and I just joined the company HFS Research, which is a smaller firm outside of Boston. They also have offices in Cambridge and India.

And when I say research and advisory, research and advisory based on working with organizations on talking about new technologies, expertise with AI. The company also builds research around cloud computing, automation, and digital transformation. So HFS produces research, but then also does advisory and works with large enterprises as well.

However, in my role as AI operations lead, I don't do the research. What I am doing is I got hired on, it's like, okay, the research company is building research and talking to clients about AI and technology. What I'm doing is I'm trying to digital transformation within the company.

So how can you take a research and advisory company to grow with AI so they don't just talk the talk, they're walking the walk. And so that's why I joined HFS Research.

And a little bit more about research and advisory, how many of you guys heard of Deep Research that launched? few of you.

Okay, so deep research is essentially an agent. It started with, well, I think Google started with launching deep research, and it's basically an agent built within, you know, Gemini or OpenAI, ChatGPT, and Perplexity also has it, but it's an agent that

just became really popular in a technology about using a reasoning model, slowing it down and going through like the thoughtful critical thinking process and searching the web. Why am I bringing this up? This rattles research and advisory industry.

You know, people can actually just go to ChatGPT and get research, very well put together research, you know, right through with an LLM. So that's another thing that research and advisory companies need to think about. It's like you're not just competing against other research firms creating PDFs and doing research. You are now competing against these LLMs and platforms, these major players that are building AI technology.

With that said, I just kind of want to open it. It's changing a lot of industries, and I don't know if anybody thought of research and advisory, but it is definitely shaking research and advisory, and consulting as well.

What they brought me into was basically going, Stacy, we need to advance our operations and we need to not just talk the talk, but walk the walk. So in addition to building an agentic workflow strategy for the organization, we are also building a model that has the whole knowledge base.

So we have 1,700 pieces of research that this model is going to be trained on. So we have an engine that we're building. So in addition to doing the research, we're training a model on all this research.

So that's one area how we're evolving in this space and we will make that available for our customers so they don't have to download pieces of research, they can just talk to an AI agent and get everything they need.

The other area which I'll talk to you today about is internal. How do you bring agents into the workforce? And so I'm kind of walking you through a practical approach to that.

I am not necessarily showing you products. It's more like thinking, how do you bring it into the workforce?

So the first step is workforce readiness. Is the workforce ready? But at the same time, you know, when you were trying to bring it to the org, you know, what is the skill set of your current workforce at play? How do they feel about AI? How are they using it?

You know, I'm lucky, you know, our company is 75 people. It's smaller, so more agile. We write about AI. We know what's coming. So, you know, we're ready to take it on.

And you need to kind of Understand that first and foremost.

Then it moves into current AI usage. So is the workforce ready to be taken? How do they feel about AI? And then secondly, yes, how are they using it?

So mapping existing tools and identifying gaps.

So in my case, a lot of people were using AI within the organization, but it was all in different ways. It wasn't in an organized, structured way.

We have an enterprise license with ChatGPT, that they were using, but they were building a bunch of other little AI tools on the side.

So I'm coming in to kind of build a holistic strategy that can be used by everybody and kind of connect the dots. So it's not just a bunch of random agents and tools everywhere.

And then lastly, workflow optimization. So

And so once you go through those three steps of kind of having those conversations with various stakeholders, I highly recommend you don't do this in isolation. You need to talk to each department, right? You need to understand your entire workforce.

How do they feel? What use cases?

I sent out a survey to kind of understand everybody's feelings. How are they using it? What are their processes? What would they like to automate?

So just really understanding each department.

So after getting that, it's like, okay, do we build or do we buy? Do you buy the agents? Do you go out and go find ready-made agents and bring into the workforce or do you build? And when we were thinking about it, we're like, well, we don't really have a large tech team.

So building is going to be complicated. You know, we're pretty agile. But then by we also want to have some control, like we want to be part of building this solution.

So what we came up, we found a partner, Lyser AI. They're out of Jersey and they have offices in India. And is anybody familiar with Crew AI or Langchain? Yeah? Okay.

So how I would compare Lyser to Langchain or Crew AI. Langchain, Crew AI developers, they go after developers. Coding is key to build these type of agents. You need to have decent tech experience.

Lyser is going into the area of where they're looking for businesses and they will help you build. You know, and they're really big on safe AI and responsible AI. So, you know, we're working with them both on strategy and also build. And what they offer for free, just like if you build agents with Make.com or Langchain, you can build some agents for free, test it out.

They have a space as well that you can enroll in their agent studio. You can create your own agent. here and so they're set up a little bit differently right it's a little bit more no code friendly um you know if you can see can you can barely see there but like you can write your agent name you can give it a description select your llm provider

and then you know it's basically drop down boxes that you fill what is the agent role what list of instructions you can connect it to various tools that are listed here there's more tools that they have available that aren't listed but it's just a different approach that is more easy for somebody that doesn't have coding experience to go and build some agents another thing that i like is that As noted, they prioritize safe AI, so they do have toggles where if you are more focused on fairness and bias, toxic check, and then human in the loop. And they're also looking at redaction. So if you are building an agent, you're uploading information that is private, it also can redact private information.

So when we started talking with them, we like that. We like this ability where if you're a developer, you can go in, you can go deeper, you can also be more focused in using coding. But if you aren't, you don't code that much, you can also build these agents.

And I'm just going to go back. So we chose what we're gonna go with, where it's kind of this hybrid of build and buy. We want to bring agents into the entire workforce, right? So we felt Lyser was the best partner for that at this time.

And I encourage you to check them out. Like I said, if you're a developer or on the business side, it is interesting how these different companies are understanding the needs from different audiences and how they're addressing it.

So having the agent chosen, now it comes to implementation roadmap, right? So, you know, what do you do? Well, discovery. Define objectives and, you know, assess your data accessibility.

So another big thing, and this is huge, is what is your organization's tech stack? And you have the list, which is great, but you need to look at more deeper into what are the APIs? How can you integrate it with AI?

What is the data in, data out of your tech stack, your full tech stack? And so we had to list it all out. And then also, what potential AI strategies can you have with your tech stack in various ways?

HubSpot's our CRM. So what is HubSpot doing with AI? How can we use Lyser to work with HubSpot?

What is the best solution for a specific task when you have two different tools that say they can do that? So it's all that thinking of connecting it all.

AI is not necessarily part of your tech stack necessarily. It's much bigger, but how does your AI you know, work with that ecosystem.

So we had to work that out as well, and we're still working it out to this day.

Workflow Agents

And then lastly is the use case, you know, scaling the expertise and enabling cross-functional workflows. So you get a workflow up, and then it's working, and then what we're thinking is, and I'll show it to you, this could be more than just for the marketing team to work.

Content, everybody uses content. Right?

So kind of just thinking and mapping out like, okay, what I what you don't want is to have a bunch of different agents all over the place for everybody, right? It needs to be organized. So it's looking at, you know, cross functional workflows flows that you can create.

And one other thing I just kind of wanted to point out is the types of AI agents that we had to look at. And for some here, it's pretty obvious, right?

You have your task agent, very simple. It's just gonna handle single goals and tasks that are repetitive.

And so when you are going through this journey of assessing what you need, how complex does it need to be, right? So that's a checklist as well based on the needs and assessments.

Do you need a task agent? Is it something that's super simple?

Workflow agent is another category that we came up with, which is you're using only one AI engine, but it's connected to a lot of different pieces within your org. So tech, like your CRM or something. So it's more of a workflow and it doesn't involve a human checkpoint in the middle of it executing a task.

Then there's the insight agent. I kind of label this in our organization as pretty much your LLMs. I don't really even want to build one because they're so good at it.

So we have a chat GPT enterprise license. And so everybody's trained on it to be your strategy assistant.

So that's how they use that. We won't play with that area. If you want an insight agent, can't beat the LLMs themselves.

And the multi agent and I'll show you what we're working on right now building. Fortunately, I can't demo the output yet. But I can show you the workflow.

But yeah, going through the journey, this is where we're looking at a system of agents, you know, enabling collaboration between AI tools and humans and mapping out that workflow. So this is a multi-channel campaign workflow, primarily for content. So once again, I told you we are a research company. We produce 1,700 pieces of research.

And so what we wanted to do is that when a new piece of research gets published, we want it to output marketing content immediately and we train and we're training it to do so so if you see in the black square the research agent content strategy those are the agents those are the ai agents but when you go step one is our research analyst which is basically pulling

and key details from the research deck. Our research decks are 50 pages long, so it's pulling out what you need. And so we've trained that agent on that.

Then it goes to a quality check. That's the blue triangles there. Those are checking to make sure that it's following our brand guidelines, safety checks, all the stuff that we need for a QA assessment. If it doesn't feel that it hit the quality check, then it pings the human.

So then my colleague Aubrey gets pinged on Teams to say, hey, I don't think, can you just check to make sure that this sounds right? Once that goes through, if it needs to go through quality check, then it goes to content strategy. And this is where the research agent feeds it to the content strategy agent.

And we're training it to think of a content strategy. Here's the research deck. Here's your audience.

Here's the key findings of the research. I want you to build a content strategy. What can we do? Should we do a podcast?

Should we do this? What type of marketing, positioning? So that's filling the role of the content strategist. The content strategist, we go through the same phase too. We wanna make sure that it's quality and it's following our guidelines. And if it needs a human intervention, then it pings Aubrey on Teams.

Then it goes to a multi-format. There we go. Then here it goes to a multi-format content agent.

This is where it starts writing the content. So it's going to write content for all our channels. It's going to write landing page copy. It's going to write social posts, emails, sales emails, all content that we would use to market. and then it goes through the check and then we are training it to do visual and media assets.

So it'll create your visual content as well. And then once that's all done, it goes to the human for check and then it's approved for publication. We are not connecting it to our email or social media at this point, that's too risky. So right now we're going to keep it as it's going to deliver as a package to the human at the end after some checkpoints, if necessary.

So that's what we're working on right now. I just kind of wanted to take you through a more complex workflow and how you can make it as complex or simple as you want. But once we built this out, we realized that sales leads could use this. The research teams could actually use it themselves based on sharing it. So this is a workflow that we are likely going to take across the org and not just within marketing.

And I think I'll show you one demo. This is not attached, but working with Lyser. I have like two minutes, right? Awesome.

Just since I had time to show a demo, Lyser built this for me, honestly, in an hour. So what they did is they built it within their tool and then they connected with Lovable.

Lovable is a pretty cool app for the developers. I encourage that you check it out. It writes code and it can build an app within seconds.

So what he did for a user interface, he used Lovable, but it's powered by Lyser. So what I showed you in the back end of how to build an agent. And so like within an hour, I'm just gonna brief it.

This is a contents read. Again, this is not the workflow that I showed you because that workflow would have a lot of human checkpoints. But this is just a quick little tool that you can build within Lyser, connect it to Lovable, and create your own marketing strategy.

So I just copied and pasted a campaign goal and hit Generate. And the output is kind of outlining, this is my brief up here, but it outlines your marketing strategy. It gives you a one to 12 week campaign targeting working professionals through influencer marketing.

And it kind of just broke it up into sections. You have some tabs here that you can build. So you have your target audience, details on that.

Also, the different channels that you can use, influencer marketing, sponsored campaign trips, et cetera. Budget, they created an app for budget, and then time. And also, if you wanted to, you could just turn the toggle on and talk with the strategist. right within the app.

Like I said, very simple. I wanted to just show this demo, and this is something that they built in an hour with Lyser and Lovable. It's kind of interesting in how quickly you can build things.