From: redpointai

Measuring progress and justifying investment in AI initiatives requires a rigorous approach focused on Return on Investment (ROI) [00:00:02]. The journey should begin small and incrementally scale up [00:00:01].

Starting Small: The Experimental Approach

Given the difficulty in predicting AI’s effectiveness for specific use cases [00:00:18], it’s recommended to adopt a scientific, experimental mindset [00:00:25].

Initial Steps

The initial phase can be as simple as:

  • Spending a minimal amount (e.g., 20 cents) on platforms like OpenAI or Llama on Databricks to “litmus test” AI’s capabilities for a particular task [00:00:12].
  • Setting up experiments to maximize chances of success, trying the “best possible model” available [00:00:30].
  • For preliminary testing, even basic prompting or manually supplying a few relevant documents to the model can suffice, rather than immediately implementing complex systems like RAG (Retrieval Augmented Generation) [00:00:34].

The goal is to determine if there’s any inherent value or “there there” before investing further [00:00:41].

Iterative Progress and Benchmarking

As you progress, ensure that advancements are justified by demonstrable ROI and contribute to “things that matter” [00:00:03].

Evaluation and Refinement

  • Establish Benchmarks: Create specific benchmarks to test AI’s performance [00:00:08].
  • Iterative Improvement: Recognize that initial benchmarks may be inadequate and continuously build better ones [00:00:09]. This iterative process is crucial for evaluating AI systems.

Scaling Up Complexity

If initial experiments show promise, the next steps involve increasing complexity:

  • Implementing RAG: If the model requires access to internal data, it’s time to implement a robust RAG system, as the model won’t inherently know enterprise-specific information [00:00:43].
  • Fine-Tuning: If significant value is being generated, fine-tuning the model can embed more specific knowledge, leading to better quality, though it incurs more upfront cost [00:00:53]. This falls under AI production techniques.

NOTE

The principle is to constantly test, evaluate, and scale up only when justified by tangible progress and value creation.