thirtymlmusings

In 1787, Russian Empress Catherine II along with several members of her court planned a tour of Crimea. To show that the area was more prosperous than it was and impress the Empress, the local Governor, Grigory Potemkin, used painted facades and set up mobile villages along the Dnieper River.  As soon as the barge carrying the Empress arrived, Potemkin’s men dressed as peasants would populate the village, once the barge left, the village was quickly disassembled and rebuilt downstream overnight. The term “Potemkin village” has since come to mean a façade created to impress or hide reality. 

Today many AI initiatives are becoming the corporate version of Potemkin village. According to a Financial Times report, employees in Amazon used AI tokens for unnecessary tasks to meet their manager’s expectations of AI usage. Employees were indulging in what’s called “token-maxxing”, a performative productivity tactic involving deliberately inflating token usage to impress managers or meet corporate expectations on AI adoption. 

AI’s siren song seems hard to resist. Companies are contorting themselves to attach the “AI” moniker to their business because of the financial incentives attached. Tech companies, VC funds and Wall Street – all have a vested interest in letting this frenzy continue. 

Unfortunately, AI adoption in many firms is not based on strategic need but due to competitive pressures and the fear of being left behind. This panic-driven adoption leads to perverse adoption targets, unrealistic expectations and misguided pilots that don’t scale. And when actual business problems are not solved the investments don’t yield any ROI.

While only a Luddite would deny that AI unlocks new opportunities, the current lemmings-like rush needs to be replaced with a more nuanced, deliberate and strategic approach. 

Grow A Pair. Leaders need to grow a spine, demonstrate courage to look beyond the hype and temper Board and investor expectations. Leaders need to consistently remind stakeholders that AI adoption must not be “skin-deep”, business integration yielding sustained benefits needs patience and a longer-term view.

When AI Mandates Go Wrong. AI adoption is being forced top-down in many companies through daily targets for using prompts, consuming tokens or linking usage with bonus payouts. Such mandates have unintended consequences. When Meta, Shopify and Salesforce created internal leaderboards for their employees to track their weekly and monthly token usage, this competition led to “token maxxing”. Employees prioritized volume over quality of output, they intentionally sent out longer or unnecessary prompts, used higher-end models for simple tasks and ran multiple agents in parallel to maximize their usage numbers. All of this led to high compute bills for these companies. Such instances are a good reminder of Goodhart’s Law; when a measure becomes a target, it ceases to be a good measure. 

Focus On Business Value. It’s important to eschew “AI theatre” where an AI-project simply ends up as a shiny showcase but stays disjointed and not integrated to business. The fundamentals of business; revenue, cost and margins are immutable truths, your AI pilots must test ideas that deliver value across any of these dimensions. 

Back To School. Learn more about your business, get into the operational innards, truly understand how your business makes money and learn everything you can about the cost and margin structure. In parallel, invest in quality education on AI for yourself and your team, understand how AI-tools work, what it can do and what it cannot. When you combine AI-fluency with deep business knowledge, you are equipped to unlock value. 

Data Still Determines AI Success. The #1 reason for failure of AI projects is poor-quality, outdated, siloed and incomplete internal data. Invest time and effort to streamline internal data collection pipelines, improve data quality and robustness to reap benefits from modeling. 

Amp-up Your BS Detector. The American astronomer, Carl Sagan said, extraordinary claims need extraordinary evidence. Don’t fall for outrageous claims on LinkedIn, verify rigorously and form your opinions based on facts. Be a curious skeptic, this trait when combined with your deep knowledge of the business empowers you to call out charlatans and snake-oil salesmen masquerading as AI-thought leaders. 

Lastly, Don’t Forget The Human Element. While it’s easy to be hypnotized by the AI Pied Piper, you can break its spell by remembering that people buy, cookies on computers don’t. The curve of advances in technology is often steeper than the curve of human adoption and AI-outputs can be human-like only but can never be human. In almost every business, conversations beget trust and trust builds relationships. And many business relationships have value that go beyond mere transactions. 

AI is no silver bullet; it is a strategic amplifier. Businesses that chase optics will achieve little more than an illusion, but businesses that ground AI in strategy, quality data and integrate it with business and human judgment will unlock its true potential. The choice is between building something of substance or constructing the next Potemkin village. 

First published here in BW Businessworld.