Latest / Full Tech Ahead / The Importance of Model Context Protocol

Transcript

The Importance of Model Context Protocol · Full Tech Ahead

• 0:20
  SPEAKER_00: Hello and welcome to Full Tech Ahead.
• 0:22
  SPEAKER_00: I'm your host, Amanda Rizzani, and with me today, I'm excited to have Amit Sharma.
• 0:27
  SPEAKER_00: He is the CEO and founder of C Data.
• 0:30
  SPEAKER_00: How are you doing?
• 0:32
  SPEAKER_01: Doing well.
• 0:32
  SPEAKER_01: Nice to be here with you, Amanda.
• 0:35
  SPEAKER_00: Happy to have you on the show.
• 0:37
  SPEAKER_00: Can you share a little bit about C Data?
• 0:39
  SPEAKER_00: What services do you provide?
• 0:41
  SPEAKER_01: Absolutely.
• 0:42
  SPEAKER_01: So C Data is all about connecting data to AR.
• 0:46
  SPEAKER_01: The core capability we are, like as you think about AI has transformed all of our lives.
• 0:51
  SPEAKER_01: It's definitely in the workspace.
• 0:54
  SPEAKER_01: But the real power of AI is only captured when AI can actually connect to enterprise data.
• 1:01
  SPEAKER_01: Think about all the systems that modern enterprises use today, like Salesforce or accounting systems or CRM systems.
• 1:10
  SPEAKER_01: And the ability to be able to have AI to connect to this data set in a secure, governed fashion is transformational for most organizations.
• 1:19
  SPEAKER_01: And that's all we're about.
• 1:21
  SPEAKER_00: Okay, great.
• 1:22
  SPEAKER_00: Well, we're on the show today to talk about enterprise AI and infrastructure and the impact it's having on the world.
• 1:28
  SPEAKER_00: What are some of the AI issues or problems that you're seeing right now?
• 1:34
  SPEAKER_00: What are some of the toughest ones?
• 1:37
  SPEAKER_01: I think the toughest problem I'm seeing is like uh most people have had uh used AI and see how transformational it can be.
• 1:45
  SPEAKER_01: But for certain use cases that that involve enterprise data, we're still at an early stage in figuring out how do we do that?
• 1:52
  SPEAKER_01: How do we make sure that we can do this in a secure manner?
• 1:56
  SPEAKER_01: How do we make sure that all of our employees can connect AI to the data sets that they need?
• 2:02
  SPEAKER_01: How do we model the data so AI is efficient?
• 2:04
  SPEAKER_01: There's a lot of uh questions around token efficiency, et cetera.
• 2:08
  SPEAKER_01: Those are the problems I see in the industry.
• 2:10
  SPEAKER_01: And people are approaching it in different ways, but um, we believe that we have a solution that can be transformational.
• 2:18
  SPEAKER_00: So a lot of the problems I'm hearing about is the quality of the AI, um, the answers, and um where do you solve this problem?
• 2:30
  SPEAKER_00: It's a data problem, oftentimes.
• 2:33
  SPEAKER_00: So, what advice do you have?
• 2:35
  SPEAKER_01: So, yeah, I mean, when you think about the quality of uh the responses from AI, most people think that um that there might be a better model around the corner that's gonna fix the issue.
• 2:46
  SPEAKER_01: But I can tell you that the improvements we have seen in the recent models, starting from the end of last year with uh the release of uh new models from Anthropic and other LLM vendors, it is truly a step function above what we had before.
• 3:01
  SPEAKER_01: The models aren't the issue.
• 3:03
  SPEAKER_01: The issue is like how do we make the data and context available to AI?
• 3:08
  SPEAKER_01: There's a lot of context that we capture in our head that is not available to AI.
• 3:13
  SPEAKER_01: And if you can actually bring that to AI in a sensible way, if you can bring the data to bear to AI in a sensible way, uh it could truly be transformational in how it works.
• 3:25
  SPEAKER_00: When it comes to integrating AI tools, it there are so many AI tools out there.
• 3:31
  SPEAKER_00: So, what is the first step for business leaders as far as determining what AI to incorporate?
• 3:40
  SPEAKER_01: I'm sure many of your users would probably have heard about MCP.
• 3:44
  SPEAKER_01: So, like you could start with any of the model providers.
• 3:47
  SPEAKER_01: Like you need a you need a model that you want to work with.
• 3:50
  SPEAKER_01: You could start with Cloud, you could start with OpenAI, you could start with uh Gemini.
• 3:55
  SPEAKER_01: In fact, what I'm seeing in most organizations, people are using all three different set of users are using one model, and I think that's totally natural.
• 4:02
  SPEAKER_01: People have preferences, the models are stronger in certain areas.
• 4:05
  SPEAKER_01: So I would not try to fight that.
• 4:07
  SPEAKER_01: And if people want to do that, that's that's okay.
• 4:10
  SPEAKER_01: The second step is how do we bring more context and data into these AI models?
• 4:15
  SPEAKER_01: And that is where, like, if your users haven't heard this, MCP is the new wave and how to do that.
• 4:20
  SPEAKER_01: Lots of capabilities in the MCP.
• 4:22
  SPEAKER_01: MCP is the model context protocol.
• 4:25
  SPEAKER_01: It was designed to bring context into uh AI models.
• 4:29
  SPEAKER_01: There are various approaches to do that.
• 4:31
  SPEAKER_01: That is a second step that people should think about how what is my strategy to use something like an MCP to bring uh data into the AI models to uh to answer whatever business decisions are we're trying to make or questions we're trying to uh ask off of AI.
• 4:47
  SPEAKER_01: And once you have those two pieces, it's not enough to just get any MCP.
• 4:52
  SPEAKER_01: You have to think carefully about how that MCP solution is structured, how are you going to govern it?
• 4:58
  SPEAKER_01: How does it uh work so that you're not creating excessive use of uh tokens?
• 5:03
  SPEAKER_01: Uh so there are some problems, but I think those two or three steps would get uh your users a long way into thinking about the problem.
• 5:12
  SPEAKER_01: And then, of course, it needs to be some discovery in finding the best mix.
• 5:16
  SPEAKER_00: So it seems like the next step, everyone's moving towards agentic AI.
• 5:21
  SPEAKER_00: We're hearing agentic AI.
• 5:23
  SPEAKER_00: And so, what are some of the changes and some of the things to be aware of when using AI tools that are starting to take action on their own?
• 5:32
  SPEAKER_00: They're not just answering questions or uh responding.
• 5:37
  SPEAKER_01: Yeah.
• 5:37
  SPEAKER_01: So I think uh the big step change and when you're thinking about agentic use case versus conversational use case is the there's no human in the loop, right?
• 5:47
  SPEAKER_01: The agents are taking actions on our behalf.
• 5:49
  SPEAKER_01: Uh so the biggest thing, the first thing that people need to think about is the accuracy of whatever action they are taking.
• 5:56
  SPEAKER_01: Well, did they understand the problem correctly?
• 5:59
  SPEAKER_01: Are they taking the right action?
• 6:00
  SPEAKER_01: Because the cost of taking an incorrect action is much more larger than the cost of giving a wrong answer to a human being who can immediately detect it and ask for a clarification or or ignore the answer and do their own research.
• 6:16
  SPEAKER_01: That cannot happen in an agent tech system.
• 6:18
  SPEAKER_01: So I think what happens is the requirements for accuracy are much higher when you're thinking about agent tech solutions.
• 6:26
  SPEAKER_01: That is why this earlier this year, we published a paper on accuracy and responses in AI.
• 6:33
  SPEAKER_01: Uh, I would encourage your user base to go and check it out on our website.
• 6:36
  SPEAKER_01: But I think like the first thing to think about is the accuracy of what the agent is going to perform as an action is uh dominant.
• 6:45
  SPEAKER_01: The next few steps is how do you secure the agent so it can only do what it wants to do?
• 6:50
  SPEAKER_01: How do you govern the agent are also important uh following questions.
• 6:55
  SPEAKER_00: Absolutely.
• 6:56
  SPEAKER_00: And as they incorporate more and more AI tools, as you mentioned, many companies are using all the different AI tools for various purposes.
• 7:05
  SPEAKER_00: That's a lot of new technology to keep track of.
• 7:09
  SPEAKER_00: What advice do you have for business leaders as far as making sure they understand all the tools that they're using and protecting their the company?
• 7:18
  SPEAKER_01: Yeah, I mean, I think like uh if you think about the AI stack, so to speak, as different layers, uh, you could probably you should have a preferred tool.
• 7:28
  SPEAKER_01: Like, like as I said, I'm very open for people switching models.
• 7:31
  SPEAKER_01: People are very familiar doing that.
• 7:33
  SPEAKER_01: You could switch uh models based on use case, but you should think about your AI stack in different pieces.
• 7:42
  SPEAKER_01: So there is the data layer.
• 7:44
  SPEAKER_01: Uh, how do you you should think about the data layer?
• 7:46
  SPEAKER_01: How do you want to bring data to AI?
• 7:48
  SPEAKER_01: I'm obviously biased, and I think C data is a fantastic solution for that.
• 7:52
  SPEAKER_01: But regardless of solution you use, uh, you should think about how I'm gonna manage my data, how I'm gonna model it, how am I going to bring it to where to AI, and there has to be a strategy towards that.
• 8:04
  SPEAKER_01: On top of that, is there is the LLM layer where you're thinking about which LLMs will take actions on it.
• 8:11
  SPEAKER_01: And then you need to think about an agent orchestration platform that's going to orchestrate all the agents that you're going to be building.
• 8:18
  SPEAKER_01: Multiple solutions for all of those uh pieces, but I think like uh it is beneficial to think about the broad architecture first and then identifying the best solution for each segment of that architecture.
• 8:32
  SPEAKER_00: And let's talk about the infrastructure here for a second, too.
• 8:37
  SPEAKER_00: As more and more companies are using AI, and in fact, everyone is using AI, there is a concern about where all that data is housed, and we're needing, you know, more and more data centers as the solution.
• 8:50
  SPEAKER_00: So I've been hearing about actually um, you know, on-premise data centers coming back to companies.
• 8:57
  SPEAKER_00: What are your thoughts on this?
• 9:00
  SPEAKER_01: Absolutely.
• 9:01
  SPEAKER_01: I think uh data privacy and uh data residency are important issues for many organizations.
• 9:08
  SPEAKER_01: Uh, we actually believe that there are reasons, separate from how you're building the AI stack, decide where your data is uh resides.
• 9:17
  SPEAKER_01: And people shouldn't just give away up on those residency requirements very easily.
• 9:22
  SPEAKER_01: If they have a case for keeping data on-prem, they should keep the data on-prem.
• 9:26
  SPEAKER_01: We, in fact, uh favor solutions like virtualization, where you can leave the data where it is uh and be able to act on it at real time.
• 9:35
  SPEAKER_01: The alternative approach is, of course, to bring warehouse all the data in one place so AI can work on it.
• 9:42
  SPEAKER_01: So that involves like bringing all the data from various places into one central warehouse and then operate on it.
• 9:48
  SPEAKER_01: There are use cases where that is useful, but in our uh in our experience, oftentimes that's not necessary and it creates more problems than it solves.
• 9:58
  SPEAKER_01: Uh, so we we actually recommend leaving data in place where it is uh needs to reside.
• 10:04
  SPEAKER_01: As you mentioned, like a lot of organizations choose to keep that data on-prem.
• 10:07
  SPEAKER_01: So we're in favor of actually leaving it there and then building infrastructure on top that you can model it, virtualize it, and still connect it to AI.
• 10:15
  SPEAKER_00: Well, AI is advancing rapidly, and as fast as it's evolving, what do you envision for the future?
• 10:21
  SPEAKER_00: What is the next great wave that you're gonna see hit companies?
• 10:27
  SPEAKER_01: I think the next wave is going to be of uh of uh efficiency and more value out of AI.
• 10:34
  SPEAKER_01: So, like uh, I think most people are familiar about AI, but we have only scratched the surface in terms of uh how much uh value we can get out of it.
• 10:44
  SPEAKER_01: As people start deploying agents and start working with these technologies, they will find that they can do a lot more.
• 10:50
  SPEAKER_01: And I'm also a firm believer, like some people worry about the AI disrupting human capital.
• 10:57
  SPEAKER_01: I'm a firm believer that AI is going to enhance that.
• 11:00
  SPEAKER_01: Like it's not a question of uh I I if you go to any company, like uh they will tell you there is never a dearth of ideas that people want to implement.
• 11:10
  SPEAKER_01: So it's not a question of who can do this work instead.
• 11:13
  SPEAKER_01: It's a question of how much can of the future we can build, how quickly.
• 11:17
  SPEAKER_01: So I'm a firm believer in that.
• 11:19
  SPEAKER_01: So I think we will see immense change in organizations that are adopting AI and we will see uh then move much faster.
• 11:27
  SPEAKER_01: Uh things that they were might be thinking of doing in the next 10 years, they probably are going to be able to do in the next three to four years.
• 11:34
  SPEAKER_01: And even faster.
• 11:35
  SPEAKER_01: Yeah.
• 11:36
  SPEAKER_00: Well, if there was one key takeaway you could leave our audience with today, what would that be?
• 11:41
  SPEAKER_01: Uh the key takeaway is like uh if you are not already on this journey, uh, I don't I don't like to uh create a FUD factor and and scare people into doing this.
• 11:50
  SPEAKER_01: Uh there is not much to this technology.
• 11:53
  SPEAKER_01: You can learn it quickly.
• 11:54
  SPEAKER_01: Your organizations can left seen people not knowing it at all and being able to adopt it very, very quickly from conversational AI to agentic AI.
• 12:04
  SPEAKER_01: Take the first steps.
• 12:06
  SPEAKER_01: Uh, you will be amazed what it can do.
• 12:08
  SPEAKER_01: We have uh I've seen demos where people are brought into a room that have not seen the capabilities of AI and they are wowed.
• 12:14
  SPEAKER_01: Create that wow experience uh inside your organization so people are like what AI can do and that it can truly transform their work lives.
• 12:23
  SPEAKER_00: Wonderful.
• 12:24
  SPEAKER_00: Well, thank you so much for coming on the show and sharing your insights today.
• 12:28
  SPEAKER_01: Thanks, Amanda.
• 12:29
  SPEAKER_01: This was great.
• 12:30
  SPEAKER_00: And thank you to our audience.
• 12:31
  SPEAKER_00: If you have any questions or comments or concerns, put those in uh below in the comments, and I'll try to respond back as soon as possible.
• 12:38
  SPEAKER_00: Have a wonderful day.