The usual assumption is that selling voice AI into a restaurant group means convincing the person who signs. That person is frequently the easiest one in the building. They see the missed-call number, they do the arithmetic, and they say yes in a twenty-minute conversation.
Then the deal sits for two months.
It sits because the signature is one of five or six approvals, and the other four are held by people who were not in that meeting, have different jobs, and have each been handed a piece of technology that made their life worse at some point. Knowing who they are, and what each one is actually worried about, is most of what determines whether this takes six weeks or six months.
The seat that signs is not the seat that decides
At a group with three to ten restaurants, the owner or managing partner signs and often decides. Above that, signature authority moves to a VP of operations or a CFO, and decision-making spreads out.
The pattern is consistent enough to plan around. Someone becomes the internal champion, usually because the phone problem is on their desk. Someone else has veto power for reasons unrelated to the phone. And a third person, typically a general manager at a specific store, has no formal role at all and can still sink the whole thing by telling the ops director that the pilot was a mess.
Map those three people by name before your second vendor call. Not by title. Titles at restaurant groups are unreliable guides to who holds what, and a director of operations at one twelve-unit group has authority a VP at another does not.
What each seat needs to see
The mistake is showing everyone the same deck. Each of these people is answering a different question, and material that satisfies one of them is noise to the others.
The owner or principal
They need the money framed against something they already pay for. Not against a hypothetical. The comparison that works is an answering service invoice, an overflow line, or the labor hours currently spent with a phone against someone's shoulder during a rush. Plans start at $250 a month per location, and that number only means something next to a line item that already exists. That framing is worked through in getting budget approval from an owner.
They also want to know what happens if it doesn't work. Answer that unprompted.
The operations leader
This is the one who blocks, and it is worth understanding why. Their team absorbs every complaint about how the restaurant sounds. They have watched a POS migration go badly. Their instinct is that automating the phone will produce a wave of guest frustration that arrives on their desk and not on the desk of whoever bought it.
Accuracy percentages do not move this person. Recordings do. Play them three real calls, including one that went sideways and got handed to a human, and let them hear the handoff work. Then show them the escalation rules and let them edit one. A person who has changed a setting owns the outcome differently than a person who was shown a result.
Finance or the controller
They want the unit economics per location and the contract terms, in that order. Monthly cost, what happens at renewal, what a location costs to add or remove, and whether anything is metered in a way that spikes during a holiday rush. Give them the whole pricing structure rather than a monthly figure, because a finance person who finds a cost you did not mention will assume there are others.
IT, or whoever holds the credentials
At fifty locations this is a real IT function running a security review. At eight locations it is an office manager and a phone bill. Either way, find them early. They control POS access, the phone account, and number porting, and every one of those is a week of delay if it surfaces late.
At the larger end you will also get questions about where call recordings live, who can access them, and what happens to the data if you leave. Those are reasonable questions with real answers, covered in data ownership and privacy for voice AI. Do not improvise them in a meeting.
Marketing or the brand owner
Present at franchisors and at groups with a defined brand voice, absent otherwise. Their question is whether the agent sounds like the brand. Show them the greeting and let them write it. It costs nothing and it converts a potential objector into someone with a stake in the rollout.
How the block usually arrives
Rarely as a no. Almost always as a delay with a reasonable justification attached.
The security review that needs to be scheduled. The wait until after the holiday season. The request to revisit it when the POS upgrade is finished. Each is individually defensible and collectively fatal, because a restaurant group's attention is a scarce thing and a project that slips twice usually slips permanently.
The counter is a pilot at one or two stores, run and finished before the committee ever meets. A 30-day evaluation plan at a single location produces something a committee cannot argue with in the abstract: this store's missed calls, this store's average phone ticket, this store's GM saying whether it helped. Committees are good at debating proposals and bad at debating results from their own restaurant.
Pick the pilot store carefully. It should be busy enough that the phone genuinely rings, and run by someone the rest of the organization respects. A pilot at your easiest store proves nothing to anyone, and a pilot at your worst store proves something you did not intend, because a location with a broken menu and three open shifts will make any system look bad.
Two stores is better than one if you can manage it. A single store's results get dismissed as that store's manager being unusually good at this, which is an objection you cannot answer after the fact.
Sequencing so nothing gets re-litigated
Run it in this order. Champion identifies the problem with real numbers from real stores. Pilot at one or two locations with the operations leader involved from day one, not briefed afterward. IT and legal review in parallel with the pilot rather than after it, because that review is dead time and dead time is what kills these. Finance sees the pilot results and the full pricing at the same meeting. Owner signs.
The version that fails runs the same steps in the opposite order, starting with a signature and ending with an operations leader hearing about it secondhand.
One more thing worth doing before any of this: ask each person in the process what would have to be true for them to say yes. Most will tell you plainly, and a few will name something you cannot deliver. Finding that out in week one is much cheaper than finding it out in week nine, and it is the same discipline that makes vendor evaluation useful rather than ceremonial.