The phrase AI-native is mostly noise. Every integrator has a slide. Every CEO has a quote. Every product page has a sticker. Almost none of it describes anything that has happened yet. The 2026 NY tristate competitive census looked at 474 firms in this market — and only 4.9% of competitor homepages mention AI at all. Of those, the overwhelming majority are announcing strategies, hiring advisors, or attaching the word to a feature list.
This is what AI-native broadcast operations actually look like at a Fortune 500 today. Not next quarter. Not on the roadmap. Today. At Intuit’s NYC campus. Where I sit as Strategic On-Site Lead.
The institutional precondition
Intuit institutionally provides every employee a $500/month AI development token budget. That number matters. Most enterprises in 2026 either haven’t budgeted AI augmentation at all, or they’ve done it as a top-of-house bulk procurement that doesn’t reach the desk where the work happens. Intuit has decided that every employee operates inside an AI-augmented engineering loop, and the company funds the budget that makes that real.
I operate inside that decision every day. The work that follows below isn’t something I imported from somewhere else — it’s the operating posture Intuit makes possible by funding the tokens.
Most enterprises haven’t yet decided whether AI-augmented engineering is a productivity policy or a security risk. Intuit has decided. The work I describe below is what the decision looks like at the desk.
What it looks like at the desk
An AI-native operating practice is not a chatbot. It’s a set of workflows where the model is part of the toolchain. Here is what that means in five concrete places:
1. Run-of-show notes
Before a national broadcast: Claude composes the run-of-show document from the prior event template + the talent list + the venue layout + the cue sheet. I edit, validate, push to the team. The 90-minute task becomes 25 minutes. The 25 minutes are spent on the parts a human still has to make — the calls about which cues need rehearsal time, which rooms need RF coordination, which talent needs a tech check.
2. Site reports
After every shop visit, I dictate field notes into the phone. GPT-4 transcribes, structures, and formats into the Intuit campus reporting standard. I review and submit. The format is consistent every time, the dispatch is faster, and the receiving team gets the same template structure on every visit. Quality of documentation goes up. Time spent typing goes to zero.
3. Vendor escalation packets
When a piece of gear goes sideways, the vendor needs a packet: serial number, firmware version, observed behavior, expected behavior, prior tickets, recovery steps already taken, recommended action. Gemini composes the first draft from my dictation plus the device’s last health report. I edit. The escalation moves from a half-day exercise to forty minutes. The vendor SLA clock starts faster, and the documentation that’s now in the loop is better than what most engineers have time to write.
4. Pre-event RF coordination
Multiple wireless devices in the same room, plus Wi-Fi, plus cellular bridges, plus the building DAS. The frequency plan is non-trivial. AI doesn’t solve the physics for me. But it accelerates the iteration: given these devices and these frequencies and this venue, what conflicts do you see, and what alternative plan would minimize them. I get three plans in five minutes instead of one plan in forty-five minutes. I pick. We deploy.
5. The vibe coding work stream
Added to the engagement May 2026: I do natural-language-driven software development through Claude against the Intuit AI-development token budget. Internal tooling. Workflow scripts. Reporting helpers. Quality-of-life code that an integrator on a billable-hours contract would never write because there’s no SOW for it. The vibe coding workflow makes this kind of internal tool development a regular practice instead of a heroic one-off.
What the audience never sees
Here’s the part that matters most. None of this is visible in the broadcast. The audience watching an Intuit national event doesn’t see the AI-augmented run-of-show. They don’t see the AI-drafted vendor packet that got the codec swapped in time for go-live. They don’t see the AI-iterated RF plan that kept the executive’s lavalier clean.
What the audience sees is the show. Cleanly. On time. With nothing they have to forgive. The AI-native operating layer is invisible — and that’s exactly the broadcast-grade bar I wrote about in the previous Field Note. The model is making the bar easier to clear. It is not changing where the bar is.
What this is not
It is not autonomous broadcast. The model doesn’t call cues. It doesn’t mix audio. It doesn’t cut cameras. The director is human. The mix engineer is human. The stage manager is human. The talent is human.
What the model does is the documentation, the reporting, the iteration on plans, the vendor coordination, and the internal tooling. Everything that used to take an extra hour of an engineer’s time before the show, and another hour after the show, and a half-day of writing escalations during the week. That hour comes back. The engineer spends it on the parts of the work the model can’t do.
That is the actual productivity story. Not a 10× claim. Not a moonshot. An hour back at every touchpoint, applied across a Fortune 500 broadcast operation that runs all year. That math is enormous. And it’s already running.
Why this matters competitively
Most of the field is going to catch up to this in 12 to 24 months. The AI-augmented engineering loop is going to become standard practice. The tokens are going to get cheaper. The workflows are going to get well-documented. By 2027 every integrator is going to claim some version of what I described above.
What VAAV Industries holds is the head start. Not the technology — the operating practice. The 18-month lead between “running this on a Fortune 500 today” and “announcing the strategy.” That lead is the moat. It’s the reason I publish this Field Note now, while the practice is rare. By the time everyone else gets here, we’ll be on to the next layer.