Success with AI

The purpose of this page is to describe how the chuck-stack uses AI to help small and medium organization improve operations.

Generative AI (specifically aichat) was the catalyst to create the chuck-stack. AI, in all its forms, is what drives conversational enterprise computing. We now have the ability to put more tools with less effort in the hand of the people who perform organization's actual work.

AI in the chuck-stack

The chuck-stack uses AI to help organizations is three high-level ways:

Data insights - using traditional BI and AI analytics to drive insights and operational objectives
Real-time assistance - helping users receive the right information, make the right decisions, execute the right actions, and eliminating mistakes
Role augmentation - helping users automate and monitor their role's tasks without the involvement of IT (tool generation)

AI Data Insights

Traditional/classical AI analytics methodologies continue to drive an organization's success. The datasets are getting bigger, but the analysis tools remain largely the same. Here are the high-level steps:

Extract, transform and load the data into an appropriate columnar storage and processing tool (Spark, Snowflake, Redshift, BigQuery)
Work with a qualified data analytics professional to analyze the data using best practices
Establish the behaviors that create the most beneficial outcomes
Build tools that offer real-time assistance and augmentation (next section) to drive future behaviors

AI Real-Time Assistance and Augmentation

A chuck-stack success factor is getting AI to help organizations think and reason about operations. Here is how we use our AI Brain to think:

Working memory (cli and shell session engagement and current context)
Episodic/augmented Memory (attribute tagging important observations and references)
Semantic memory (work instruction retrieval)
Procedural memory (scripts and workflow)

References:

AI Principles

Here are the AI principles that drive our decisions:

Simply the AI tool sets to the extent possible
Bring the model and code to the data
Make the models and code useful to those who actually perform the work

AI as a System

This video does a great job of characterizing our approach to AI systems deployment. Any one model by itself does not help an organization. Instead, the chuck-stack's use of work instructions, database conventions, tool deployment, multi-model prompting, and best practices drive success with the chuck-stack.