Prevent AI backlash: invest in people
Much has been said and written lately about the need to govern the prolific output of content from GenAI systems that are intended to support business goals, but too often create as many problems as they solve.
Among the chief challenges involved are ensuring that AI models are effectively trained to be free of bias and hallucinations, and that the content they produce is consistently compliant with a brand’s quality standards and policies.
Embarrassing and downright dumb examples of GenAI producing cringe-worthy mistakes abound, among these Google Photo infamously categorizing people of African descent as gorillas. Or more recently, an article in Microsoft Start’s travel pages referring to the Ottawa Food Bank as a “tourist hotspot,” advising readers to visit on “an empty stomach.”
Moreover, when it comes to the esthetic quality of images it produces, today’s GenAI typically tends toward a schlocky plastic look, containing basic flaws that violate common sense and design principles that no human who’s a qualified designer would ever let slip past.
Don’t Believe the Hype
Challenges posed by the current AI tech revolution resemble what happened in the market 30 years ago when desktop publishing solutions first emerged, putting tools into the hands of anyone with a PC. What resulted was a flooding of the market with poorly designed, inexpensively produced, but crappy content that mostly served to cheapen brand values.
Eventually, it dawned upon anyone responsible for building and maintaining a brand’s image that all the hype accompanying new desktop publishing tools was just that—hype. While it’s true the technology did enable a trained designer to efficiently produce quality content, without the human talent and training to operate them the tools were like typewriters to monkeys.
The rapid emergence of GenAI now requires revenue growth and marketing leadership to take a step back and ask themselves not just what are the potential benefits of the technology, but also taking a close look at its costs and limits as well. This means starting with a careful strategic assessment, before adding yet another platform to their existing tech stacks.
Dropping Back a Step
Marketing tech providers have traditionally offered solutions that only support specific functions like content personalization, lead management, or customer analytics. At many organizations (small and large), the situation has created a confounding mess of disconnected systems, siloed data, and ultimately a poor customer experience—all of which negatively impact a company’s revenue.
Even when they’re aware of the potential backlash that GenAI can produce, many chief revenue and marketing officers don’t fully grasp that success in today’s business climate—awash with tsunamis of data—won’t be achieved by simply adding an AI platform to their already bloated tech stacks.
Instead, achieving the holy grail of an AI strategy that produces high-quality, brand-compliant marketing content on a massive scale, using key data points and performance indicators that support a positive customer experience, requires a fundamental rethinking by company leadership.
Minimum Viable Data
Perhaps counterintuitively, one of the biggest drivers of unnecessary costs associated with deploying GenAI into an organization’s revenue chain results from an attempt to mine too much data with the system—more than required to yield useful outputs.
Not only can an overambitious, data-greedy approach generate inaccurate conclusions about likely customer preferences, etc., it inevitably results in excessive data storage and processing costs. Most large companies, for example, keep large quantities of data stored across various departments, including previously abandoned pet projects, which are irrelevant and should be excluded from a new GenAI model.
Based on my recent experience consulting with many CROs and CMOs at numerous Fortune 500 companies, I have observed the most effective leaders recognize the importance of taking a more minimalist approach when deciding which data sources to tap with a GenAI model.
In this regard, allow me to offer a simple formula to keep in mind when planning and deploying a new GenAI solution into the organization’s revenue chain: Minimum Viable Data = Minimum Valuable Product (MVD = MVP).
Starting with People
Effectively applying the principle of MVD = MVP to GenAI entails leadership first bringing all the revenue chain’s stakeholders together in the planning-stage work group, with the goal of answering an initial question, “What is the MVD for this proposed solution?”
My experience has led me to observe that some stakeholders typically covet certain data points they believe to be essential for success, based on their unique but limited perspective on the entire revenue chain. Often, their preferences reflect a bias of some type and their desire to emphasize this in a GenAI’s model, which leads to ill-advised outputs, not to mention unnecessary data storage and processing costs.
On the other hand, the best insights in terms of essential data points for achieving revenue-chain success using GenAI naturally come from stakeholders who are working daily in the trenches. So, it’s incumbent upon the CRO or CMO—whoever’s ultimately in charge—to listen thoughtfully to all stakeholder perspectives, while maintaining a laser focus on separating wheat from chaff in defining a project’s MVD.
It may sound like a cliché, and yet it’s demonstrably true: The most valuable resource in any company are its people. This holds true when it comes to leveraging GenAI to optimize an organization’s end-to-end revenue chain, from IT to product design, marketing, sales, and customer satisfaction. Investing in the people who perform these critical roles daily and valuing their practical experience and knowledge is the place to start.
# # #