When a winery owner asks me to help with the first AI rollout, the second question is usually some version of who should run it. The first question has already been answered for them — the IT-adjacent staffer, or the marketing coordinator who reads tech newsletters. By the time the owner is on the call with me, the chosen person has already been told they will be the champion. Almost every time I have walked into that setup, the pick was wrong.
The right first champion at most wineries is the tasting room lead. Or the wine club concierge, or the hospitality manager — whichever role at your shop is the one staff already go to when a guest is upset or someone needs to swap a Friday shift. The person at the visible center of how the team handles trouble is the person the team will follow into a new way of working. Anyone else is a guess.
That claim runs against the way most rollouts get assigned. The rest of this post is the case for it and the four-criterion order I use when a winery owner asks me to help pick, grounded in a worked example with our fictional demo winery so the criteria stop being abstract.
Why the wrong person gets the job by default
The wrong pick happens for a believable reason. When a leader thinks about AI, the picture in her head is a screen full of typing and a chatbot answering back. The role she imagines is closer to a software admin than a hospitality lead. So she scans her team for the staffer who looks the most software-admin-shaped and hands them the work.
That scan tends to surface two profiles. There is the technically inclined staffer who already runs three productivity apps for personal use. And there is the ambitious early-career employee who has read enough to know AI is a career-defining wave and wants the visibility of being the in-house expert. Both will say yes when asked and will probably enjoy the assignment. Neither is the person the rest of your team will copy.
The companion to this post — the wrong-champion failure mode — covers what happens after that pick gets made. Brennan McDonald documented it as the most-cited failure pattern in AI rollouts. The short version is that the team's distrust attaches to the messenger before the message arrives. The team is not refusing AI; they are watching a coworker they do not already follow and deciding whether to follow her into this. The decision usually lands the wrong way, quietly.
So the operative question on day one is who the team will follow into a new way of working. The technical-skill question shows up much later, and matters less than most owners assume.
The four criteria, in priority order
Hartz AI's published guidance puts the cohort ratio at one champion per fifteen to twenty employees. For most wineries that means one to three champions total, and getting the first pick right is structural. Here is the order I use.
Peer trust comes first. Peer trust is the property of being the person other staff already go to when something has gone sideways. The lead host the night shift texts about a difficult guest. The club concierge the marketing coordinator pulls aside before sending a tricky cancellation email. The cellar hand the production assistant asks about a stuck pump while the winemaker is mid-fermentation. This is who the team has already chosen as their internal helpline. The role is unassigned and informal, which is why most leaders miss it.
Watch your team for a week and you will see this person. They get pulled into side conversations. Other staff ask them questions in passing rather than waiting for the manager. When they say something casually, the rest of the team takes it as guidance rather than information. That is the asset you want pointed at your AI rollout. When that staffer says she has been using a tool for her Friday wholesaler emails and it is saving her an hour, the rest of the team hears a coworker reporting a useful trick. When the volunteer enthusiast says the same sentence in the same room, the rest of the team hears a sales pitch.
Curiosity comes second. Once you have a shortlist of peer-trusted staff, look for the ones who already ask is-there-a-better-way questions about their own work. The tasting room lead who built her own visit-recap template in Notes because the official CRM is clunky. The concierge who keeps a private list of email templates because the macros never fit. The wholesaler-rep coordinator who has experimented with three different ways to track sample requests. These are the staff who already iterate on their own workflows when nobody is watching. They are the ones who will fold AI into work that exists rather than treating it as a separate exercise.
Curiosity matters because the champion has to keep using the tool past the novelty phase. The 3x-a-week-for-8-weeks habit window Microsoft documented is real and unforgiving. A peer-trusted staffer without curiosity will use the tool dutifully for two weeks, then quietly stop. A peer-trusted staffer with curiosity will hit week six, find a use case nobody scoped for her, and that use case is what carries the team across the threshold.
Willingness comes third. The third filter is whether the person wants the role when you explain it to them. Some peer-trusted, curious staff will hear the pitch and say no. That answer is information. Run the credibility audit on the next candidate. A reluctant champion forced into the role becomes the surface-compliance pattern McDonald describes — opens the tool once a week so the dashboard looks alive, types a low-stakes prompt, closes the tab.
Willingness is also where the involvement gap shows up. Keith Ferrazzi's HBR piece cited the number that ought to keep leaders up: only twelve percent of employees feel involved in their employer's AI deployment decisions, against seventy-eight percent who fear AI will take their job. The willingness question is the moment where you cross that gap with one person on your team. Make the conversation a real one. The right champion will tell you what she is worried about. That conversation is also how she becomes the person who later answers the same worry from her coworkers, in language the rest of the team already trusts her to use.
Technical skill is fourth, and a distant fourth. Almost every leader I work with would have put this first if I had let them. The reality of AI tools in 2026 is that the technical bar to start using one well in a hospitality role is below the bar most of your staff cleared the first time they learned a new POS or CRM. The skills that matter are reading a draft critically, knowing what a good club email or visit recap sounds like, and being willing to redo a prompt when the first output is off. Those are hospitality skills wearing different clothes. A peer-trusted, curious, willing tasting room lead with average computer comfort will outperform a technically fluent staffer with no peer trust by the end of week two and by a wider margin by the end of month two.
There is a tail risk in over-weighting technical skill. The technically fluent staffer will sometimes build a workflow nobody else can use because she optimized for elegance rather than for the version her coworkers would copy. That workflow gets the right answer once, then dies the moment she goes on vacation. The peer-trusted champion will build a clunkier workflow that survives because three other people understood it well enough to keep it running.
A worked example, with the fictional demo winery
To make this concrete, here is what the pick looks like at Not Really Wines — our openly fictional demo winery at notreallywines.vercel.app, which we run as a working showcase of the kind of vault and workflow we set up for real clients. NRW has a thirty-person operation split across production, DTC, hospitality, and wholesale. Maren Holloway, the CEO, asks me which staffer to point at the AI rollout first.
The candidate she has in mind is Marcus, her operations and finance lead, because he runs the spreadsheets and would be the staffer she trusts to manage a software vendor. Marcus is competent. He is also at the center of nothing on the hospitality side. The tasting bar staff have never asked Marcus a question about a difficult guest, and the club concierge has never run a tricky cancellation email past him. Marcus belongs on the implementation side — license seats and vendor agreements. The champion side is a different job, because that role lives or dies on peer trust the hospitality team has already given to someone else.
The right first champion is Riley Park, the tasting bar lead. Riley sits at the visible center of how the front-of-house team handles trouble. The other hospitality staff text her on weekends when a regular asks for a wine the inventory system has no record of. The mobile trailer crew calls her first when a private event runs long. Tessa Brennan, the hospitality director, has been quietly funneling the trickier guest scenarios to Riley for a year because Riley keeps a clearer head than anyone else on the floor.
Riley also clears the curiosity bar. She maintains her own visit-recap template in Notes because the CRM template buries the fields she relies on. She has been experimenting with three different ways to organize the regulars' preference lists. Nobody asked her to do any of that. She does it because the official setup has gaps she keeps tripping over.
The willingness conversation with Riley takes twenty minutes. She has questions about whether the tool will replace any of her work. She has more questions about whether her coworkers will think she is the one pushing it on them. Both questions deserve real answers. The first is short — the goal is to take the parts of her job she already dislikes and compress them so she has more time for the parts she likes. The second is where Maren has to make a call about how the rollout gets framed to the rest of the team. The framing is that Riley is trying something on her own work and the rest of the team will see how it goes. The framing matters because it is the truth.
Technical skill is the fourth conversation, and the shortest. Riley uses Instagram and the winery POS. She has never written a prompt. By the end of week one she will have written forty of them. The technical question turns out to be a non-issue, the way it almost always does.
Two months in, the play has done what the play does. Riley's visit recaps come out in three minutes instead of twenty. The Wednesday follow-up note to regulars who came in over the weekend goes out the same week now, when it used to slip to the following month. Tessa Brennan, who was the most skeptical of the rollout going in, asks Riley how she handled an email after a difficult event — and Riley shows her on the screen. Tessa is the second champion, picked by Riley's behavior, not by a staffing decision. That handoff is the proof signal Maren needed.
What this changes about your next step
If you are about to pick a champion, run the credibility audit before you assign the role. Stand somewhere visible for a week — the break room before opening, or the cellar floor at end of shift — and watch which staffer keeps getting pulled into the difficult conversations. Pick that person, in that order: peer trust, then curiosity, then willingness, then a brief check that she is comfortable enough with a keyboard to start.
If you have already picked someone and the picture I described in the worked example does not match what is happening at your winery, the move is to re-pick. The companion post on the wrong-champion failure mode covers the handoff in detail. The short version is a candid conversation with the original pick, a quiet transfer of the work, and a deliberate decision to under-announce the change. The team will read a clean handoff as a normal workflow adjustment. They will read a reset announcement as a referendum on the first person, which is the one outcome you want to avoid.
A small note about the language you use when you make the pick. The word "champion" is fine internally and useful in writing about the play, but when you are talking to Riley, do not call her your AI champion. Call her the staffer trying something on her own work. The label adds pressure she does not need and creates a category the rest of the team will start measuring her against. The version of the pick that survives is the one that looks, from the outside, like a normal staffer doing her normal work with a different tool. That is the version her coworkers will copy.
For the broader frame on which moves to skip when you are getting started, five things wine industry leaders should stop doing in AI rollouts covers the patterns that most often sit alongside a wrong-champion pick.
A note on the webinar
I am running a webinar that walks through the full 60-day play this post sits inside — the credibility audit, the four-criterion pick, the second-employee proof story, and the habit-formation window that decides whether the rollout sticks. If you are getting ready to assign a first champion at your winery or you have already picked one and the rollout has not moved the way you hoped, the session is built for that decision.
Check the webinar schedule for the next date and to register.
Part of a 20-post series on employee AI adoption for wineries — see the full series under AI Adoption.