What is an Autonomous Agent
and why it’s not just ChatGPT or Claude
It is much more than simple access to an LLM – Large Language Model – typically obtained through a cloud-based cognitive chatbot or through the various web interfaces provided by global LLM makers such as OpenAI, Anthropic, Google, etc.
A BoK Autonomous Agent works on-Prem
Autonomous Agents in BoK operate within the BoK platform, which can be installed on-Premises or on our clients’ Private Cloud. This provides greater assurance of confidentiality in processing: organizations’ data is processed securely.
The typical elements of an Autonomous Agent are:
- Input Data
- these are needed to turn a processing workflow into an actual output
- The Autonomous Agent’s specific processing environment
- AI – LLM: the actual Large Language Model, for which the ability to draw on its own general knowledge of a given topic is deliberately limited — this being the main cause of hallucinations.
Furthermore, LLMs can be run either on-Premises (e.g., OpenAI’s GPT, Meta’s LLaMA, Google’s Gemma, etc.) or accessed as on-Cloud LLMs (e.g., OpenAI’s GPT, Anthropic’s Claude, Google’s Gemini, etc.). The choice between an on-Premises LLM, an on-Cloud LLM, or a combination of both depends solely on the client’s specific requirements. - Training and Domain Knowledge: these are needed to turn a processing workflow into an actual output
- Context Layer: these are needed to turn a processing workflow into an actual output
- Semantic Layer: these are needed to turn a processing workflow into an actual output
- Output generation instructions: these are needed to turn a processing workflow into an actual output
- AI – LLM: the actual Large Language Model, for which the ability to draw on its own general knowledge of a given topic is deliberately limited — this being the main cause of hallucinations.
- Output Data
- Layer di formattazione grafica e/o di predisposizione del formato di ouput previsto dall’utente: puro testo, HTML, un file JSON, una chiamata ad un’API, etc.
Internal and External Knowledge
In an Autonomous Agent, the solidity of the Internal Knowledge made available by the Client — in collaboration with R-Tree consultants — proves critically important.
What is an Autonomous Agent
INPUT
Data and/or documents
AI – Large Language Model
Training and Domain Knowledge
Output generation instructions
External Knowledge
Internal Knowledge
(Client and R-Tree)
AUTONOMOUS AGENT
OUTPUT
Data and/or documents
BoK as a platform for
Autonomous Agents
Each Autonomous Agent represents a way of accessing your own knowledge with the assistance of Generative AI.
What can you do with Autonomous Agents?
- Have a dialogue with your own PDFs
- Fill out checklists and validation lists, based on defined acceptance criteria, according to the contents of certain documents analyzed by the Autonomous Agent
- Search your own knowledge base in a conversational way
- Create in-house experts specialized in specific domains of your organization’s knowledge:
- Expert in researching topics from customer technical support tickets
- Expert in paint composition for a painting/coating plant
- Expert in managing production line faults/breakdowns
- Expert metallurgist and analyst of the chemical properties of metals
- Expert in managing Safety Data Sheets for chemical products
- Expert generator of marketing emails
- Expert in writing technical documentation/manuals
Need information?
Tell us about your context and goal:
we will reply with the most suitable path.
