The questions CEOs ask before hiring Eyzaguirre & Co. The answers are Alan’s.
Our AI pilots aren’t working. Why? They were built on an outdated plan. Big consultancies are paid to keep clients happy, while AI demands rethinking the whole sector. I read that shift on blog.eyzaguirre.co from daily use, not analyst ratings. Three things changed.
The cloud is no longer the default. Most companies were sold AI as the next step of the move to the cloud. The SaaSpocalypse broke that story, and keeping data in your own warehouse is back on the table.
The market moved years early. We all expected local models to need a few more years to catch up. Clever engineering has already closed the gap on disk space and computing power. Agentic AI arrived faster still, and your kids probably understand it better than your organization does. The question is no longer whether AI gets a spreadsheet right. It is which part of the business AI can run first.
The scorecard changed. The KPIs, the ROI math and the dashboards are different now. Vendors would rather you did not notice that Porter’s five forces are back on your side: buyers have power again, outcome-based pricing is the goal, and the market has little patience left for overpriced per-token pricing. Some of it I wrote down months before the market caught up: in September 2025 I argued for agents working under structured permissions, a month before the leading coding agent shipped a sandbox.
We already spent our AI budget. Why spend more?
Then the first job is making the money already spent pay. Stop the licenses nobody opens and the pilots nobody owns, and put that cash into the one project that can return it. New spend waits until a project can show its return.
Why hire another consultant into a process already full of them?
Because my job has always been to find what isn’t being said: the mechanics of the boardroom, and what an audience really thinks behind what it says. I spent years on your side of the table. At RingCentral the CEO treated the Magic Quadrant as the scoreboard, and the company lined up behind it. At PayPal I built analyst relations before the pandemic made it urgent. At Cisco I took on Responsible AI when nobody else would, and wrote the position that changed how the company saw itself. In each of those places I ran the consultants for the executives I served and kept them in check. I know where a deck hedges, where scope creeps, and when a twelve-week assessment is a status update in disguise. Today the only way to hold a consultant to account on AI is to know it at every level: the PyTorch code, the architecture, the safety, how people actually use it, the long-term economics, and the story. I build at each of those levels myself. The sixth is the one most firms skip: a client is not rewarded for putting state-of-the-art AI into its key workflows. It is rewarded when investors, the board and employees believe the plan and see it working. That is why strategic communications is part of every engagement. And I have skin in the game: my fee is fixed before I start, and my name is on the result.
Isn’t this our IT team’s job, or our software vendor’s? IT keeps the systems running, and a vendor sells its own product. Neither is paid to ask which processes would move your margin if AI ran through them. That question belongs to the CEO, and it is where I start.
Why now? We could wait a year.
Because the models already handle most back-office writing, research and triage well enough to use. What is missing is the work of putting them into the business, and that work takes months, not a license.
What do you actually deliver?
The three services on the services page, in the order the business needs them. Every engagement ends with something working, not a deck.
Our most experienced people are retiring. Can AI keep what they know? Yes, if it is captured before they leave. Their decisions, documents and reasons go into one searchable record the whole company can use, with who said what and when. It does not replace them; it keeps their judgment in the building.
Why does a consulting practice build its own product?
Because advice on AI is only as good as the adviser’s own work with it, and we’re all sensing that AI is reaching a tipping point: the tools are powerful, and few people can say what they touch. My career has been a story of becoming the same three things: an innovator, starting with my first company out of college; a leader, from Keynote at Apple to the teams I ran after it; and an empath, because every one of those products worked only when I understood what its users feared. Mariwave brings all three to AI. The structure that keeps an enterprise’s AI safe also serves a one-person studio or a broker, and I explain why in One structure keeps AI safe at every size of business.