|
Model Context Protocol (MCP)
|
|
Model Context Protocol
is a open source standard that
enables AI agents and AI embededded systems to interact
with external software applications, data sources,
business tools, and platforms. MCP was developed by
Anthropic in November 2024.
|
MCP Highlevel Architecture [reference
28]
|
| |
|
The LLM is used extensively to get answers to user questions
or prompts. This has been typically performed on AI tools
that provide solutions - language to language conversion
(SQL to python, pyspark, etc.), resolving code errors, and
for all other business requirements. The next step is to
automate the routine processes, embed humanless decision
making by autonomous code generation - create profit/loss
statements etc.
|
| |
|
Current State
|
|
Currently the systems that have AI-agents, store
workflows, dependencies, diagrams,
DAG,
etc. that relate to a data store (database table).
These provide high level diagrams of data lineage,
flow pattern, usage/user statistics. Some may also
store information about groups of data that exist
in a datalake. Code and API creation are based on
human developmental activities which involve
planning, complete SDLC, various other steps, that
consume large amount of time.
|
|
|
| |
|
Next State -
LCNC
|
|
The data store will be integrated with more details such
as codes used to describe a transaction (TAX, FREIGTH,
FEES, COST, etc.), that will be the data in the data store.
Next formula can be embedded as an Agentic-AI based on
API to compute business solutions. These API will be
accessible to intercompany LLM suite for
secure/proprietary transactional usage and
publishable/public
data can also be exposed to external agents or users.
By prompting the LLM suite, it can create python, panda
code and provide a NCLC solutions of the fly. The users
can fine tune the prompts and get best results. Typical
fixed code enhancements need full SDLC and is a long
drawn process - code maintenance, version control etc.
At a hight level following are the AI enablements.
|
-
Embed Agentic-AI with data lineage, workflow, dependencies,
API and other information
-
Agentic-AI autonomously maintains code for pay-off,
write-off, tax, etc. updating when new codes are
created.
-
By AI enablement, system constantly updates the
code list when new codes are created to be included
for computation
-
Agentic-AI maintains time sensitive data such as daily,
weekly, business day, calendar day activities.
-
Agentic-AI autonomously sends warning when gaps in data
arise and tries to auto resolve issues.
-
Agentic-AI will allow users to add new codes to analyze
data from a specific point in time (e.g. use new codes
from January 2026).
|