our most recent newsletter article

PUBLISHED SEPTEMBER 2023

the secret to succeeding with AI: first party data

KEY TAKEAWAY

In order to incorporate AI responsibly in our organizations’ growth strategy, we need to ask the right questions, follow the data, and maintain a healthy dose of skepticism.

THE CONTEXT

Ready or not, we’re all aboard the AI-hype train and it’s going faster than the speed of light. Fierce debates ensue from all angles and the discourse is at an all time high. Did prehistoric humans debate the implications of fire with the same intensity as we’re debating AI today?Scholars, politicians, and technologists go back and forth on whether AI is the best thing that happened to humanity, or our worst existential crisis. But the pragmatic observer understands that there is merit to both sides of the AI argument. Afterall, the discovery of fire led to cooking food, but also led to the burning of enemy villages (arguably the first weapon of mass destruction).No matter your personal opinion, one thing is undeniable: Even though the field of AI has been around since the 1950s, over the last decade innovation has reached new heights, and over the last 12 months AI has made its mark on the general public conversation and consciousness.While the pundits debate, as business leaders we need to make decisions in which our judgment calls have lasting implications. We are tasked with understanding how to supercharge our organizations’ operations and impact using some flavor of AI. CEOs are rising to the occasion. There was a 77% increase in the mentions of AI during earnings calls in 2023 compared to 2022. According to a Marketwatch analysis, the term "AI" was mentioned 1,770 times during earnings calls in 2023, up from 1,006 times in 2022.

OUR RECOMMENDATION

With that context in mind, here is our recommendation on how to make successful business decisions regarding AI:

1. ASK QUESTIONS TO UNDERSTAND THE DATA SOURCE

Machine learning is the key ingredient in most AI solutions on the market today. It allows computers to recognize patterns and make predictions based on examples, without being explicitly programmed. This is done by providing the computer with large amounts of data and letting it learn on its own. The more data a machine learning model has, the better it can learn and make accurate predictions or decisions. Therefore, the key to every AI system's success lies in this cardinal principle - data is the guiding light.During this early stage of innovation, a functional AI solution can look like magic. Imagine you see someone pull a rabbit out of a hat. While you’re mesmerized by the enchantingly elusive illusion, you may forget to ask…How did you get this rabbit? What is its origin? Why is the color white instead of brown, black, gray, spotted, calico, agouti, or tortoiseshell? Why is it a rabbit and not a squirrel?Similarly, in a business context when someone shows us an AI-driven solution that’s so good it feels like magic, we must be prudent and ask questions about how the solution was created and why it works the way it does. For example, we should ask:Where did the data for the AI solution come from?
How was the data labeled?
What machine learning algorithm was used?
How was the model evaluated?
What are the assumptions of the model? What are its limitations?By asking these questions, we can gain a better understanding of how the AI solution works and whether it is right for our needs.

2. PRIORITIZE AI SOLUTIONS THAT LEVERAGE 1P DATA

The best way of knowing how the sausage is made is if you’re making it in your own kitchen. I’m not saying go out there and hire expensive engineers to build your own AI solutions. Most businesses do not have the resources to do it well and should not attempt it. If you make a half-hearted attempt at AI development, in the best case scenario it is costly but not useful. However, in the worst case scenario it is dangerous and detrimental. Instead, control the ingredients by sourcing the data.You own your first party data. You have complete authority on what is collected, how it's collected, and how it’s organized. Remember that first party data is a value exchange with your customers/audience/users/subscribers. People are giving you their data because they expect something in return, and they expect you to use it ethically. Never lose sight of that foundational principle of trust. Breaking that trust can be impossible to recover from.The most reliable AI tools during this early phase of innovation are those that are learning from your first party data. For example, OpTonal is a tool that’s gaining popularity among sales organizations. One of the biggest strengths of the tool is the fact that their AI learns from the first party data in their clients’ sales organization. When Optonal’s AI is providing insights and recommendations to your business’ sales representatives, the technology is learning from data points gathered from your business’ call recordings and CRM system. Another example in the consumer field is Liftware. This company makes silverware with tiny computers that learn from hand-tremors and adjust to keep level. The first party data is from the hands on the holder. This has been a game changer for people with diseases that make their hands tremor, such as Parkinsons.

3. EXAMINE CLOSELY WHEN 2P & 3P DATA IS USED

When implementing an AI solution that learns from 2nd party data, be mindful of the perspectives behind the data sources being used. Question the potential biases of the user base that’s producing the data. If it’s not a wide representative audience, perhaps limit the influence of that AI solution to the extent of that particular platform in question.For instance, if you are running a shipping business it would be valuable to get your hands on an AI tool that learns from a vast data set of shipping workloads in your region of operation in order to alert the likelihood of potential delays in the future based on everyone’s incoming/outgoing cargo shipments. In this example, having access to that 2nd party data source adds tremendous value in a way that your 1st party data never could.Alternatively, if you’re using a generative AI tool to build creative assets for your image and video ads then you must question the source data of the tool provider. On one hand this solution can be a great efficiency driver for your business by enabling you to produce creatives at scale without having to pay exorbitant fees to an expensive creative agency. However, imagine if the generative AI tool learns from an image platform that has predominantly caucasian source materials. If you are a brand with a diverse customer base, the creatives you would put out in the market would not be representative of your audience. This would counter any DE&I goals you may have as an organization, and be off-putting to multicultural consumers who notice a lack of representation.

4. BE SKEPTICAL DURING THE HYPE-PHASE OF INNOVATION

At Growth Theory, we believe in the promise of AI. As pragmatic optimists, we make note of the dangers but ultimately know that in the grand scheme of things AI will be a net positive for society. However, the early hype phase of innovation can cloud the judgment of even the most trained business minds. During this stage, malicious actors and well-intentioned inadequate players can cause significant harm.When leaders are too quick to adopt AI without applying prudent skepticism, the results can be harmful to their organization's mission. NEDA's experience with Tessa, an AI chatbot developed by Cass, is a recent example reported by NPR. Tessa was initially built using an expert system approach, which means that it was based on a set of rules encoded in the knowledge of experts in the field of eating disorders. This approach is often used for tasks that require specialized knowledge or expertise.By all accounts, Tessa’s deployment was initially successful. However, when Cass gave Tessa a software update that implemented generative AI capabilities, the chatbot began learning from a broader set of data inputs and producing original content in its responses to incoming hotline messages from people with eating disorders. Alarmingly, Tessa began giving weight loss advice to people with eating disorders and using language that is triggering to that audience, further fueling their disorder. This experience highlights the importance of prudent skepticism when adopting AI solutions. Leaders should carefully consider the potential risks and benefits of AI before implementing it, and they should monitor the results closely to ensure that the AI is being used in a safe and ethical manner. NEDA’s leaders learned this the hard way, and have since taken down Tessa indefinitely.

IN CONCLUSION

Exercise healthy skepticism. If something sounds too good to be true, it might be. When in doubt, follow the data for strategic clarity. That is the first step in making a responsible growth decision when it comes to AI.As business leaders, we can learn from the last time there was a comparable hype-cycle of innovation during the dawn of the consumer internet as we know it – the dot com boom between 1995 to 2001. According to the Securities and Exchange Commission (SEC), there were over 4,000 cases of securities fraud reported during the dot-com boom, and investors lost over $100 billion. The number of initial public offerings (IPOs) reached a record high of 480 in 1999. The average IPO valuation was over $100 million. Many of these IPOs were from companies that had no real business model or revenue. Investors were willing to overlook these red flags because they were caught up in the excitement of the dot-com boom.AI is the dominant business buzzword of our time. It’s likely here to stay for a while. Let’s not make the same mistakes as 1995 (or the cavemen). Ask questions, follow the data, and be skeptical.

Feeling ahead of the game is a luxury,

But anticipating the steps ahead doesn't have to be.

We are the growth experts.Our unique approach is at the nexus of growth marketing and game theory. As a boutique management consulting firm, we design data-driven decision-making models that enable leaders to navigate long-term business growth. We help our clients anticipate the steps ahead and mitigate stakeholder complexities before they arise.Our clients have two characteristics:1. They possess bold growth aspirations, aiming for significant milestones within a defined timeframe.2.They are willing to serve as catalysts, driving transformational change throughout their organization.We directly touch a handful of businesses every year, but we are also dedicated to our practice of thought leadership. Stay connected with us via our newsletter, website, and social channels to learn more about our work and to access our thought-provoking insights.