I keep getting versions of the same question from business owners and leaders: how do we adopt AI in a way that actually works? And then, usually in the same conversation: but how do we do it responsibly? They have seen the headlines about the water use, the emissions, and the effect on jobs, and they do not want to be the person who approved something connected to those impacts.
Here is what I see when I look at their organizations. AI adoption is already happening. Just not strategically, and usually without anyone tracking the consequences. Someone in marketing is running campaigns through a personal ChatGPT account. Someone in customer support keeps a Claude subscription open as an informal knowledge base. The owner is using Gemini to draft proposals between meetings.
The pattern has a name: shadow AI. About three-quarters of knowledge workers now use AI for their jobs, and roughly half are using tools their employer never issued. Most companies have no formal AI policy. Small and mid-sized businesses are no exception. Adopting AI this way creates real risk, and it leaves real opportunity on the table.
What I hear from leadership is that they feel stuck. They can see the need to do something. They can also see the risk of doing the wrong thing. So they do not do much at all. And while they are not deciding, the business moves ahead anyway, without guardrails or coordination.
Being behind might be an advantage
Some businesses are behind on adopting AI at all. Most are behind on using it deliberately. AI is being described as a force more powerful than the industrial revolution, arriving many times faster. If that is even partly true, it deserves more care than a personal ChatGPT login. Many of the decisions that matter have not been made yet: which vendors you trust, what data you let them see, who signs off on AI output, and which workflows you actually redesign. If that describes your business, it may be an advantage.
The companies that raced ahead are now dealing with the bill: data security gaps, quality-control failures, and incidents that cost them customer trust. A business that paused, even by accident, can do it right the first time. But that is only an advantage if you use the window to make these choices deliberately, before the business makes them for you by default.
Making responsible choices starts with understanding the impacts of AI and being accountable for them. I group them in three categories.
Environmental. AI has a material footprint across water, energy, and emissions. Data centers consume water to cool their chips, draw power from regional grids, and produce emissions that depend on whether that power comes from coal, gas, hydro, or renewables. Providers vary widely in both their footprint and their willingness to disclose it. (The next post puts numbers on this.)
Social. AI affects your people, your customers, and the community you operate in. The roles most exposed to AI displacement are concentrated in clerical, administrative, and support work. A 2026 Brookings analysis found that the U.S. workers most exposed and least equipped to adapt are 86% women, concentrated in those roles. Meanwhile the data centers AI runs on are increasingly being built in water-stressed parts of North America, competing with residents for the same water and power.
Governance. Governance is the work of deciding what is in bounds, naming who is accountable when AI causes harm, and staying current as the tools and rules change. In the U.S. there is no comprehensive federal AI law yet, but a fast-growing patchwork of state rules, with Colorado’s AI Act among the first. Canada’s federal attempt stalled in early 2025. And the EU AI Act’s transparency rules for AI-generated content and chatbots, landing in 2026, reach any business that markets into the EU, wherever it is based. The landscape is unsettled, which is exactly what governance has to navigate.
Each of these deserves more than a paragraph, and the next two posts go deeper. For now, the point is that AI has impacts worth acknowledging and addressing.
I have had this conversation before, about flights
I keep getting déjà vu in meetings about AI. Not because the technology is familiar, but because the conversation is.
Before I advised on AI, I spent a decade as a climate and sustainability strategist, and I had a version of this same conversation about air travel. Whole industries and economies built themselves around cheap long-haul flights, then realized late that the carbon math was going to catch up with them. Nobody was acting in bad faith. The tool became cheap and scalable, and everyone rushed to use it before the full cost was priced in.
Air travel and AI are not the same thing, but the two conversations have the same shape. Both tools are powerful. Both can be used frivolously or for something that matters. Both have individual best practices for using them more responsibly, and both have system-level problems that take collective action to solve.
In aviation, roughly thirty years passed between long-haul flight becoming cheap enough to scale and the industry seriously measuring its emissions. During those decades, businesses built themselves around exactly the thing we now need them to reduce. AI is moving far faster than aviation did, so we do not have thirty years to get this right.
AI is at the moment right before the dependencies lock in. Which vendors do you commit to? What kind of energy runs the data centers behind them? How fast do you roll the technology out? What guardrails do you put in place, and how do you support the people whose work changes? What your business decides in the next year will shape the years after it.
Where I am with this
A note on my own practice, since I am asking you to look at yours.
I do not have all of this figured out. As a society we are still learning what AI does to us, to our work, and to the systems it touches, and because the technology keeps changing, we will not have the full picture for a while. That is not a reason to wait. It is a reason to proceed deliberately, and be proactive about the risks.
AI is a genuinely powerful tool. It can solve serious problems and create real value. It can also do real harm, sometimes on purpose, more often as a side effect of how it is built and deployed.
Here is where I land personally. I use AI every day for research, writing, and thinking out loud. I have stopped using some tools after running their numbers through the AI Environmental Impact Calculator I built. I keep a working policy for what I will and will not use AI for. I have revised it several times, and I expect to keep revising it.
A good place to start adopting AI responsibly is simply knowing what you are working with. So here is a question to locate yourself: if a client, your insurer, or your own team asked how your business uses AI and what it costs, what would you say?
If the honest answer is “we would have to find out,” you are in good company. Most of the businesses I talk to are in the same holding pattern. None of it is locked in yet, which means you can still decide where your business lands. The only question is whether you make those decisions, or whether they get made for you.
The next post walks through what AI actually costs, and shares a tool to help you put a number on it.