Want a smart insight into your inbox? Sign up for our weekly newsletters to get the only thing that is important to enterprise AI, data, and security leaders. Subscribe now
The launch of something launched in November, the model context (MCP) has begun to collect a large number of consumers, but it has been guaranteed to adopt a large scale needed to make the industry standard.
But there is a subset of businesses that are no longer joining the hype for: regular industries, especially financial institutions.
Banks and other businesses that offer access to loans and financial solutions are not strangers to AI. Many people have been leading in machine learning and algorithms, even playing an integral role in making the idea of ​​using robots very popular. However, this does not mean that financial services companies want to jump immediately into the MCP and the Agent 2 Agent (A2A) bandwagen.
Although many regular companies, such as banks, financial institutions and hospitals, have begun experimenting with AI agents, they are usually internal agents. Regulated companies have APIS. Nevertheless, many of the integration of these companies have been tested years to ensure compliance and safety.
“These are very early days in the fast -paced domain, but there are blocks of some basic buildings that are missing, at least as mutual cooperation and communication standards or excellent methods.” Katina Labs. “In the early days of the web, there was no e -commerce because there was no HTTPS, and there was no way to make safely transactions, so you can’t make Amazon. You need blocks of these basic buildings, and now they have building blocks on the web, and we don’t even think about them.”
Fast, enterprises and AI platform providers are setting up MCP servers as they develop a multi -agent system that interacts with agents from external sources. The MCP provides the ability to identify an agent, which allows the server to determine tools and data, which gives access. However, many financial institutions want further assurances that they can overcome the integration and ensure only approved tasks, tools and information.
John Waldron, Senior Vice President AlonA subsidiary American BankIn an interview, Venture Bet told that when they are looking for MCP use, there are many questions around the quality.
Walden said, “There are not many standard solutions here, so we are still looking for many ways to do so, which may include doing this connection without the exchange of MCP if the agent’s technology is common between them and these are just two different domains.” “But, what is the search for the exchange of data without any other exposure in this message? What is happening in the MCP diagnosis right now shows that if the protocol is only handling the exchange and does not provide further risk leakage.
Models and agents are different
The financial institutions and other regular businesses are no strangers for AI models. However, when Robid Viewer – where the algorithm did not make decisions about financial planning and investment, which did not interfere, most passive investment increased. Many banks and assets organizers initially invested in natural language processing to enhance the performance of documentation.
However, Sales force Vice President and General Manager of Banking Industry Solutions and Strategy, Greg Jacobi, told Venture Bat that some of their financial clients already have a process to evaluate models, and they feel challenged to connect AI models and agents with their current danger scenarios.
“Machine learning and prediction models fit well with this danger framework because they are precise and predictions,” said Jacobi. “These firms immediately take the LLM to their model risk committees and have found that the LLM produces an undeniable result. It has been a crisis for these financial services firms.”
Jacobi said that these companies have the risk management framework where, if they input the models, they expect the same production every time. Any type of variations are considered a problem, so they need a method for quality control. And when the regulated companies have embraced APIS, with all the tests there, most of the regular companies “are afraid of openness, by putting something that they cannot control.
However, Elon’s Waldron does not exempt the possibility that financial institutions can work to support the MCP or A2A in the future.
“Looking at this with a business perspective and demand, I think the MCP is a very important part where I think business logic is going on,” he said.
Waldron said his team is in the diagnostic phase and “we have not yet made a server for pilot purposes, but we are going to see how to handle this boot -to -boot exchange of messages.”
Agent cannot do any other agent’s CYC
Katina Lab’s Neville said she was watching the conversation around the Inter -MCP and A2A, such as the interprants protocol, with great interest, especially since she believes that in the future, AI agent will be as user for banks as much as human consumers. Before Katina Labs started, Naville sponsored the USDC Establishment company, so it experiences the challenges of bringing new technology to regular business.
Since the MCP is open source and new, it still has permanent updates. While the MCP offers an agent identification, which is key to many companies, there are still some lost features, such as protector for communication and most importantly, the audit trailer. These issues can be solved through completely different standards like MCP, A2A or even Luka.
He said the biggest problem with the current MCP revolves around verification. When the agents become part of the financial system, even MCP or A2A, there is no real way to “your customer” about agents. Financial institutions need to know that their agents are dealing with licensed entities, so the agent must be able to point out, Navil said.
“An agent needs to be a way to say, ‘This is what I am as an agent, here my identity, my danger and who I am working.’ This certified identity can understand all different agents framework in such a way that they are key.
