AI. AI. AI. AI. Sometimes it feels as though every business conversation starts and ends with the same two letters. I use AI every day. It makes my practice faster, lighter, and more capable. It can analyze information, challenge my thinking, improve a draft, test a model, and absorb hours of mechanical work. But AI is not what a client hires me for. Clients hire me to understand the situation, connect what different stakeholders are seeing, recognize the decisions that matter, and help people move through them. AI helps me do that work. It is not the work itself.
That does not mean organizations should retreat from AI. It means compliance and business leaders need to create responsible ways to use it. My governing principle is simple: the client's classification of the work determines how AI touches it, if at all, not my preference or convenience. General work may be appropriate for a commercial AI service with the right contractual protections. Proprietary work may need to remain inside the client's approved environment. Highly sensitive work may require a tightly controlled local environment, or again, no AI at all. The choice should be explicit. The client should approve it as clearly as they approve the scope, staffing, and deliverables.
Responsible AI use begins with five questions: what is the sensitivity of the information, where is it permitted to go, which environment meets the client's requirements, what decision will the output influence, and who remains accountable for verifying it. Governance should not be a vague reason to say no. It should create a safe and understandable way to say yes. There is a world of efficiency waiting on the other side.
When I was a new manager, I spent late nights working alongside senior partners on the eighteenth floor of a midtown Manhattan building. Takeout had been ordered. A car service was ready to take people home whenever we finished. I was the person suggesting copy as we agonized over the language in a proposal. I made sure the fonts were consistent and the text boxes aligned. I reran the numbers until the deal model reached the thresholds required for approval. Then we rehearsed. Who would open? Who would speak to each section? Where would the handoffs happen? How would we answer the questions we expected? What did we know about the stakeholders, and what did we understand about their values?
The formatting and calculations were part of the job. The real curriculum was learning how experienced people prepared for decisions when the stakes were high. Even the sharpest newly minted MBA could not arrive with that experience. It had to be earned in the room. Today, AI can accelerate much of the work that once consumed those nights. It can produce alternatives, find inconsistencies, summarize information, test calculations, and turn rough thinking into a cleaner starting point. I welcome that. Spending less time aligning text boxes leaves more time for the judgment that clients actually need.
I am also not trying to recreate a global systems integrator inside my practice. I do not carry layers of management, enormous delivery teams, or the pressure to place available people into an engagement. AI gives a small, experienced practice leverage that once required a much larger bench. That creates a useful seam. Clients can have senior experience directly in the work, supported by modern tools, without paying for the machinery surrounding a global firm. That leverage comes with responsibility. Some clients may be comfortable with properly contracted commercial AI. Others may require that every piece of work remain inside their own technology environment. Some may decide that particular information should not touch an AI model at all. Those are not obstacles to work around. They are client decisions to understand and honor. My responsibility is to offer the right posture for the work, make the boundaries clear, and remain accountable for everything I deliver. AI can help me move faster. Experience tells me where to go, what deserves caution, and when moving faster is not the right objective. That distinction is the practice.
Takeaway
Early in my career, I learned the mechanics of consulting by doing them. I reran the numbers, fixed the fonts, aligned the text boxes, revised the proposal, and listened as senior partners debated a single word. AI can now do a meaningful share of those mechanics. That is real progress, and organizations should pursue it. Most of my daily AI use is deliberately ordinary. It helps draft standard invoice language, organize internal notes that contain no confidential client information, structure CRM tasks within approved systems, prepare outreach, refine public writing, build checklists, and test different ways to express an idea. It handles first passes and administrative friction. I review the work, make the decisions, and own the final output. What AI cannot reproduce is the experience happening around the task: why a word mattered, which executive should open the meeting, who should answer the difficult question, what we knew about the client's stakeholders, history, and values that changed how the recommendation should be presented. An AI model can process the record of experience. It cannot have the experience, read the room, or accept responsibility for the decision.