Data and AI
1. Data and AI
Are data and AI separate things? What do you think?2. Data is AI
No, data is AI. We look at AI as if it is a separate entity, but really, the AI component is just a model that converts data into a new form. The code is the vehicle to do that. In the digital paradigm, code is the entire product. When it comes to AI, the code and even the math are not the product. Those are only what we use to transform the data into a new format where it can deliver value and meet business or customer needs.3. A shallow moat
We don’t monetize the code or untrained model. Businesses cannot use either one as a core differentiator or competitive advantage. As Google’s famous line goes, “We have no moat!” A moat in business strategy is a competitive advantage that protects the product from being easily copied or replaced. The company was referring to LLMs and Generative AI failing to deliver protection from copycats. Those aren’t moats because constant improvements are being made to model architectures. There's always something better on the horizon. As a result, we must find something unique or hard to duplicate that a business can monetize. It's not the code, math, or the pre-trained model architecture, so what is it? Data and AI products are more than code and algorithms. At the heart of every data and AI product is the data. When Google said we have no moat, they lied, didn't they? They do have a unique data set that most other companies don't have access to. Even using the same models, Google will achieve more reliable, functional models than their competitors. Why? Google has unique data and lots of it. It isn't just the smart people at Google that make AI products like Bard. Amazing, isn't it? Without data, there is no AI, so data is the differentiator. We think about data in legacy terms, and it's tough to monetize data when we think about it in digital terms or BI terms.4. Harsh realizations.
In business intelligence (BI), displaying the data to users is the only means of value creation. In data, analytics, and more advanced models, we must have some sort of differentiation. It must be more valuable, or what's the point? Why should we use analytics? Why would we use models if there isn't an improvement over BI and simply displaying the data back to users? For the additional investment necessary to deliver analytics and models to make sense, we must explain to the rest of the business why they have a higher monetization potential. Data and AI are more expensive to deliver. Novel or difficult-to-source data must be formatted for analytics and machine learning applications. Core realization number 1, and the bad news I often have to give clients, is that most of your data is not valuable. When enterprise data is built for digital use cases, displaying it back to the user, it isn't sufficiently labeled and structured for analytics or machine learning.5. Context is king
We need something new called context. Context comes in two types: business and customer. What does context mean? It is the domain knowledge that supports some sort of functionality. Business domain knowledge is what’s required to run the business. As I discussed in the last part, the workflow, what domain knowledge is necessary to complete a workflow? When people do the work, there is an element of domain knowledge that supports each step. What's the customer context? Products and services deliver experiences and outcomes to customers. Domain knowledge or expertise is necessary to deliver high-value customer outcomes. Those connect to the workflow or set of steps that must be completed to deliver the experience or outcome. Context is the domain knowledge required to complete a workflow that delivers customer experiences and outcomes or helps run the business. Without properly formatted data, that context does not exist. As a result, we need so much more data to learn the domain knowledge, and we must rely on all of these really complex model architectures to tease it out of all that data.6. Intentional data collection
Gathering data intentionally focuses on data-generating activities like workflows. It targets extracting the knowledge required to deliver customer experiences or outcomes and run the business. If we can capture data in connection with workflows, business knowledge, domain knowledge, and expertise, it's less expensive to develop models. Models are also more connected with internal user needs or customer needs. Both help focus data and AI products on creating and delivering value to customers. Creating that connection lowers the price and delivers products that will monetize better than BI. The goal is not just capturing data, writing code, and building pre-trained models. We must capture data with business and customer context for it to translate into AI that the business can monetize successfully.7. Let's practice!
Create Your Free Account
or
By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.