Privileged Insight: the Key to AI Success in Mid-Sized Firms

You've had the key to AI success all along. Privileged Insight is the data only YOU have by virtue of being in business, and it's usually OK if it's unstructured and messy.

Leaders in mid-sized firms are under pressure to show how AI is creating new business value, but they lack the budget to fund big IT projects and the spare time to become an expert on all things AI. Here's what I'm seeing that's real and practical. Everything here is about and for mid-sized companies (let's call it $100M to $1B in annual revenue)...

1. They say AI but they mean ChatGPT. Right now, most practical successes are coming from LLM's (not ML or other AI) and most LLM projects are ChatGPT.

2. Leaders worry about their data being too dirty or unstructured to yield good results.

3. Ambitious data cleanup projects are distracting, expensive, and never seem to finish.

4. If you Google "AI strategy advice" you'll see tons of stuff from the heavy hitters but their advice seems targeted at really big phased projects that would require a big consulting company (surprise, surprise) to manage.

...organizations should rely on their own business data to build genAI models... Forrester Research

But here's the really good news.

1. Today, the key to AI success is something called Privileged Insight (PI), and you probably have a lot of it. BCG calls it "proprietary pools" and McKinsey calls it "proprietary data." What they are talking about is that data that by virtue of being in business, only YOU have. That data could be about the marketplace, or customer satisfaction, or internal workflows, or hiring trends, or customer support cases, or renewal rates, or job application resumes. Simply put: It's the accumulated data about your business that only YOU have, and ingesting it is what takes LLM's to the next level of value-creation.

2. Your PI is likely unstructured and messy and that's usually OK. You probably don't need a giant cleanup project. Instead, you need a data person with the right data skills to explore your data reality, think creatively, and assemble and augment what you have into a useable form. (Full disclosure: that's what I do.)

3. Recognize that the number one use of LLM's today is "Therapy and Companionship." That means that LLM's have the power to coach your people in a way that makes them feel better about themselves and more excited to work. For example, in a sales coaching GPT I recently configured, here's a sentence from the GPT to the sales rep after they discussed a compelling narrative crafted for a specific customer: "You got this: with that narrative and the recent news headlines, your offering is compelling and you’ll sound sharp and relevant from minute one."

4. Lastly, recognize that the top two functions for creating AI value are 1) Operations, where AI drives cost savings, and 2) Sales & Marketing, where AI drives increased revenue. When in doubt, those are the rocks to look under first.

   
   

If you look over my work samples, you'll notice that each one calls out the specific Privileged Insight data source that was put to work. All of those projects started with a concern that the data was too messy, and ended with an appreciation for how powerful it can be.