GenAI and Governance

Satya Nadella's prediction of Business Logic Agents has created significant anxiety ("do you really want AI computing the payroll?"). Of course not. That would be madness.

The solution is to put the "Human in the Loop". Microsoft uses the analogy of a Pilot and CoPilot. To pursue the metaphor, you might think of AI as the plane - enormously powerful, but the pilot remains in control. They just go a lot faster.

Let's see how.

GenAI - Human in the Loop

AI Process: GenAI, Governance, Execution

Let's recall the basic process:

GenAI Creation – Users provide a natural language prompt in a browser. This creates prototype projects that execute as your agent, enforcing the rules you provide.
AI Governance – Human-in-the-loop validation is essential. Developers review and extend logic using standard Python libraries and IDEs, and then deploy standard containers.
Agentic Operation – The Business Logic Agent processes API calls and messages, applying deterministic domain logic to maintain data integrity, and integration logic to interoperate with systems and services.

What did GenAI Actually Do?

GenAI is active only in step 1, and performed (only) 2 tasks:

Schema Creation: it creates tables and relationships from your prompt, replacing the tedious DDL process we all know and, well, know.
Logic Translation: it translates natural language logic into Rules, as shown below.

Logic Transparency

When the GenAI process is complete, you can review the logic and inspect how it was translated into underlying Python rule code.

Governance - Verify and Extend (Human in the Loop)

Let's examine the governance step.

The created project is standard: open it in your IDE, as shown at right.

Logic Translation

GenAI does not invent rules. It translates your logic into Rules as shown here. You review them, correct any issues.

The rules are completely deterministic, like the formulas in a spreadsheet.

You can add additional rules in your IDE, using code completion.

Your Code

You can add your own code. It is not realistic to create complete complex systems from only Natural Language.

Logging and Debugging

You are running in a standard IDE, with full access to the debugger. The rules also provide detail logging that shows every rule that fire. This makes it easier to verify regulatory compliance.

Your Tests

You test the logic using your favorite test tools / methodologies (e.g, Behave BDD testing).

Standard projects/IDEs enables the use of complementary AI coding tools such as CoPilot, ChatGPT, Gemini, etc.

GenAI-Logic is Governable Agentic

TL;DR The creation process provides agentic suggestions. The created server satisfies many Agentic elements (autonomy, goal oriented behavior, context awareness, planning and reasoning, action execution). Support for self-improving requires user design and implementation.

There is much interest in a new category of "Agentic AI", and how it compares to "Generative AI (GenAI)".

Core Idea: creating vs. doing

According to Bernard Marr,

GenAI is about creating
Agentic is about doing

With GenAI-Logic, the created systems actively perform transaction processing, so they are about doing.

See this article, or this YouTube.

Degree of Autonomy

One key criteria is the degree of Autonomy. According to ChatGPT:

High Autonomy: If the system independently generates an entire functional application, including backend logic, database structures, UI components, and even deployment steps without additional input or intervention, it is highly autonomous.

Key Indicators of Autonomy

• Decision-Making Ability: Does the system make architectural, design, or functional decisions without user input?

• Self-Optimization: Can the generated application adapt or improve based on inferred needs?

• End-to-End Automation: Does it handle everything from code generation to deployment autonomously?

By this definition, GenAI-Logic is highly autonomous.

Key Features of Agentic AI

We then asked ChatGPT to list the Key Features of Agentic AI, and then filled in our view of GenAI-Logic:

Summary

GenAI relieves you of the many tedious steps:

Database Creation (SQL syntax, etc)
Project Creation (initial dependencies)
Client creation (JavaScript and HTML)
Rule syntax

You control the creative process by reviewing the created project. Deploy when and where you choose. Final execution is 100% deterministic.

Page updated

Google Sites

Report abuse