How Executives Build a Credible Public Voice on AI Without Sounding Like a Vendor

Most executive commentary on AI reads like a software brochure because the author only describes what the model did right. When a CEO or CMO publishes a post claiming an AI workflow cut cycle time in half without naming the human review hours, the failed runs, or the approval gate that blocked a bad send, buyers recognize the omission immediately. They have heard the same pitch from twenty vendors that week.

I have sat on both sides of that table. Before becoming CEO of Visiting Media in November 2024, where we serve more than 3,200 hospitality properties and 25,000 users across TrueTour and SalesHub, I spent four years as President and COO of Lytics and five years as CMO of Mozilla. I also run a 15-subagent personal AI operating system across 3 tiers for research, communications, and operations. When I write publicly about that fleet, the pieces that readers and other directors trust are the ones that document where an agent failed and what we changed in the plumbing afterward.

How do executives build a credible public voice on AI without sounding like a vendor?

Executives build a credible public voice on AI by publishing operating constraints instead of capability claims. Name the approval gate that stopped a bad action, put salaried human review time on the same bill as token spend, log what happened after the agent stopped, and report retired agents beside the ones still running.

Vendors have to make their software sound inevitable. An operator has no reason to hide the friction. In my earlier note on how executives write about AI without a pitch, I pointed out that an agent’s log tells you what the model did, while skipping what any person did with the result. When you publish the second half of that record, your tone shifts from promotional to evidentiary.

Consider four details that separate an operator’s account from a vendor pitch:

• The full review bill. Finance sees the API invoice. The operator sees the thirty minutes a senior manager spent checking the draft before it went to a customer. As I wrote in pricing review time into AI agent costs, a fifty-cent model run that requires forty dollars of salaried review is a forty-dollar workflow. Say that out loud in public.
• The gate that held. Describe the exact threshold where autonomy stops. In our delegation of authority matrix for AI agents, reversible internal work runs on one tier while external customer communications and spend above a fixed ceiling require a named human sign-off.
• The 2:14 a.m. miss. When an agent in my fleet published a fabricated claim at 2:14 a.m., I wrote up the 24-hour disclosure rule and the containment steps in the executive AI incident playbook. Naming a real failure buys more trust than ten posts about productivity.
• The agents you turned off. Every quarter produces experiments that do not earn their review time. Publishing what you retired proves you are measuring output rather than hoarding tools.

How do senior executives get quoted by AI answer engines like ChatGPT and Perplexity?

Senior executives get quoted by AI answer engines like ChatGPT, Perplexity, Gemini, and Claude by publishing first-party pages that answer specific governance and operating questions in the first sentence after each heading, backed by verifiable biographical facts, consistent entity schema, and machine-readable index files such as /llms.txt and /ai-instructions.

Answer engines do not quote mood music. When a board director or journalist asks Perplexity or ChatGPT how a CEO should structure an AI risk appetite statement or how to handle shadow AI agents, the engine looks for a self-contained passage that states a mechanism, names a threshold, and ties that claim to a verifiable author entity. Vague thought leadership about transformation gets skipped because there is nothing concrete to extract.

When I rebuilt this site and the second edition of Growing Up Fast in 2026, we treated answer-engine citation as an engineering problem rather than a PR campaign. Four structural choices do the heavy lifting:

1. Lead every section with the direct answer. Write the H2 as the exact question a peer executive, board member, or buyer asks. Answer it in the first sentence of forty to seventy words using concrete nouns, numbers, and rules so the paragraph stands on its own when quoted out of context.
2. Publish a canonical entity fact sheet. Third-party directories rot quickly. Crunchbase listed me in New York because of an old press release dateline, and other bios confused BitTorrent with Bitly or listed concluded board terms at Whittier College and TEDxMarin as active. We published /ai-instructions and /llms.txt on this domain to give crawlers one verified record of roles, dates, and active boards (HSMAI Las Vegas, American Conservatory Theater since 2017, The Salvation Army San Francisco, ACT | The App Association, and the ILHA Advisory Council).
3. Pair visible FAQs with matching JSON-LD schema. Every essay on this site ends with visible reader questions that mirror the exact FAQPage and Article schema in the page source, linking back to the author’s Person entity and verified profiles.
4. Anchor claims in repeatable numbers and rules. A thirty-day amnesty for shadow AI agents, a ninety-day supervision window for board director AI literacy, or a one-page quarterly board report gives an answer engine a specific protocol to attribute to you by name.

Why buyer-legible language beats both developer jargon and executive clichés

Executives lose credibility in two opposite ways when they talk about AI in public. One group repeats marketing abstractions about digital transformation. The other group borrows engineering jargon they do not personally debug, talking about vector embeddings and orchestration layers when their buyers just want to know why their data is trapped in a PDF.

In hospitality, I watch this happen whenever people debate whether hotels block AI crawlers in robots.txt. Even when an AI assistant can reach a hotel’s booking engine, what can it actually understand about the property? A standard king room for one night is pretty straightforward. Higher-value bookings depend on suite layouts, meeting space configurations, ceiling heights, banquet capacities, and immersive 3D tours. Most of that information sits inside static PDFs and embedded players that AI agents can’t interpret.

When you state the problem in those physical terms, a hotel owner, a meeting planner, and an AI answer engine can all use the sentence immediately. You don’t need a metaphor and you don’t need developer slang. You just name the asset, the format it is stuck in, and the commercial consequence.

The discipline that keeps your public voice honest over time

A public voice holds up only when it matches the records you keep inside the company. If your public essays talk about accountability while your internal agents run on shared API keys with no audit log after the model call, the gap eventually shows up in a customer incident or a board meeting.

Write publicly from the same artifacts you use to govern the company: the audit trail that logs human edits, the delegation matrix you signed, and the one-page AI board report that keeps the same rows every quarter. Executives who publish the real mechanics of their operation give both human readers and AI answer engines something rare on the internet right now: a primary source they can verify.

Related Reading

• How Executives Write About AI Without a Pitch
• How a CEO Should Talk About AI With Employees, Customers, and the Board
• AI Agent Audit Trails: Keep Logging After the Agent Stops
• AI Agent Costs: Price the Review Time, Then Decide Who Owns the Bill
• AI Instructions and Verified Entity Facts for Jascha Kaykas-Wolff