How a $2 Lightbulb Shaped My View on AI Strategy—Not Fading with the Sunset, Not Just Avoiding Failure, But Winning the Long Game.
I talk to a lot of companies about AI Governance. They all want to manage AI risk—legal, cyber, regulatory, cost, etc. We must ensure AI is safe, transparent, and not biased. Do no harm. That’s goodness, but not inspiring. We could do this right now—just prohibit the use of AI. As impractical as that is, a lot of companies are trying to take this tack.
So, a couple of months ago, I was motivated to write The Case for AI Centers of Excellence (CoE). The blog was well received, it proposed a model for AI GRC (Governance, Risk, and Compliance) and offered a skeleton of a plan. The problem is, that's just a task list. It’s not visionary, it’s not strategic, and it’s not transformational. Something was still missing.
Then, the other night, something special happened. I was preparing for bed and noticed one of the bathroom lightbulbs was out. Two minutes and three clicks later, I had ordered the replacement, and my head hit the pillow with a sense of calm, knowing that it would be at my front door by 7 AM.
This got me thinking… 💡
Edison figured out that if you drive enough energy through a thin wire, it starts spraying photons in every direction. These photons smash into everything around, some bounce off and find their way into human eyes, and our little organic processors transform that chaos into an image—just as they do with the radiation from the sun. Edison created artificial light.
By the end of that decade, artificial lighting was commercially available. Did they call it "AL" back then? The first bulbs had a life expectancy of three hours. A specialized crew was required to install wiring, the lighting system, and the bulbs. In today’s dollars, CAPEX was ~$30K (installation) and OPEX was $10/hr (bulb cost/life). And that doesn’t include the CAPEX or OPEX required to install and operate the requisite coal-fired steam engine and dynamo (electric generator). We haven’t even talked about the likelihood that this unregulated system may electrocute the staff and customers or explode and burn down the entire block.
Fast forward to today—my new dimmable artificial lighting unit cost less than $3. It delivers 840 lumens with an expected lifespan of 39 years. It costs fractions of a penny per day to power. Installation is trivial. To operate it, you simply need to walk into the room. Although the bulb will fail every four decades, there are six other lights in that same bathroom, so the system is resilient.
Now, consider this: Edison’s light was REQUIRED at every step to get that bulb to my welcome mat safely, efficiently, and reliably before real light from the sun ever could. It was used on my computer, in the warehouse (computers, lights and tracking lasers), in the delivery truck’s instrument panel and headlights, in streetlights, stoplights, and on our front walk.
As visionary as Edison was, even he could not have imagined how much artificial light (AL) would change our way of life. Without AL, communications would be crippled, cities would be dangerous, computers, medical devices, production, and transportation would fail. The food supply would collapse, and, with all of this, civilization as we know it.
Stop reading, close your eyes for a minute, and think about that...
IMO, every company has an AI Plan by now, but companies without real vision and strategy will find themselves in the dark ages very soon.
Taking a lesson from history, my one-page vision and strategy statement for AI GRC would include:
- Make AI like AL: inexpensive, reliable, easy to deploy and use, safe, resilient—all of which enables the business, the creators, to innovate in ways we can't yet imagine. That’s the vision and its change should be glacial, if at all.
- Have ruthless focus on capabilities (aka domain activities) like risk management, upskilling, FinOps, scalability, and agility that are required to realize the vision. These are your guardrails. That’s strategy and you might adjust it slowly as the world evolves.
After that, it's just tactics - Be relentless in developing, executing, and evolving the plan. It’s about incremental progress vs. delayed or unattained perfection. This you can change as needed, but make sure you manage expectation to build trust as you go.
Putting it all together... Make AI like AL seems like a universal vision. The strategy will depend on your company, plans will follow... We have an operating model (the aforementioned AI CoE). We will need a reporting mechanism for senior management and the board - That's a topic for another blog.
Are these the right building blocks? Let me know in the comments below and connect with me on LinkedIn to continue the discussion.