Wyndham at HITEC: Real-Time Transcription on Every Inbound Call

I am sitting two rows back in a HITEC breakout room when Josh Dow leans into the microphone and says the quiet part out loud. Wyndham, he tells the room, is now running every inbound call — every one — through real-time transcription, with contextual data surfaced to the agent while the guest is still on the line. The Salesforce co-presenter nods. Someone behind me writes the word “superpowers” in a notebook because Dow has just used it, twice, and it is the kind of word that gets underlined.

Here is the contrarian read I want to put on the table before anyone else: the interesting thing about Wyndham’s rollout is not the AI. It is the decision to make transcription the default for one hundred percent of inbound traffic. Most call-center AI pilots I have toured in the last eighteen months are sampled — a percentage of calls, a subset of properties, a queue or two. Wyndham’s posture is the opposite. Transcription is the floor, not a feature. Everything they build next — agentic workflows, suggested replies, post-call summarization, escalation routing — sits on top of an assumption that the transcript exists. That is an infrastructure decision dressed up as a productivity decision, and it is the part operators should be copying.

The infrastructure bet underneath the demo

Dow’s framing during the session matched the argument Hospitality Technology made in its piece on AI-powered hospitality starting with infrastructure and evolving with people: the AI is not the product. The data layer is the product. Wyndham’s stack, as Dow described it, is built on Salesforce Communities, which is a deliberate phrase. Communities is not a call-center product per se; it is a customer-data and access-layer product. By anchoring transcription, surfacing, and agent tooling on the same fabric that owns guest identity and franchisee data, Wyndham collapses the integration tax that has historically made call-center AI a brittle bolt-on at hotels.

The mechanics, as I understood them from the session and the hallway conversation that followed: an inbound call hits Wyndham’s contact-center routing. Transcription kicks off the moment the agent answers — not after, not on a queue. As the guest speaks, named entities (property, confirmation number, rate plan, loyalty tier) are recognized and lifted into a side panel pulling from the Salesforce-resident guest record. If the call references a stay that has not happened yet, the system surfaces the reservation context. If it references a stay that already happened, it surfaces the folio and any prior service interactions. The agent does not type a search query. The transcript types it for them.

That is the “superpower” Dow keeps coming back to. He is not claiming the AI is replacing anyone. He is claiming the AI is removing the seven-second cold start at the top of every call — the alt-tab, the lookup, the “can you give me your confirmation number one more time” — and giving the agent back attention to spend on the guest. In a brand portfolio the size of Wyndham’s, where call volume and franchisee mix make uniformity hard, that seven seconds compounds into something real.

Why this reads as agentic readiness, not an agentic launch

Mark interpretation: Dow did not announce agents. He announced the substrate that agents will eventually run on. That is the right order, and it is the part most hospitality vendors are still getting wrong. Agentic workflows — the kind where a system books, refunds, or rebooks without a human in the loop — only work if the inputs are clean, complete, and live. A transcript-on-every-call posture is the cheapest, most honest way to guarantee that the inputs exist before you ask an agent to act on them.

You can see the same pattern shaping up across the category. A forthcoming May piece in this column walks through Toast IQ’s positioning around restaurant-level intelligence — same logic, different surface area: own the data plane first, ship the agent later. An upcoming May piece on OpenTable’s deeper integration with Booking is doing a version of this on the distribution side, where the “agentic” framing is really a data-merger framing. Wyndham is doing it on the voice channel. The throughline is that the operators winning the AI conversation right now are the ones whose CTOs can answer the question “where does the transcript live, and who owns the schema?” without a follow-up call.

What I would watch for next from Wyndham, based on what Dow stopped short of saying: post-call automation that closes the loop without an agent click. Refunds under a threshold. Reservation modifications inside policy. Loyalty point gestures. None of those need a human if the transcript is clean and the policy is encoded. Dow’s framing of “superpowers” is generous to the agent today; the same substrate is going to be generous to the system tomorrow. That is the deal operators should understand they are signing up for when they buy in.

The honest critique I would offer Wyndham — and any chain copying the playbook — is that one hundred percent transcription is also one hundred percent surface area for compliance, retention, and consent questions. Real-time transcription of guest voice data is not a neutral act. The franchisee conversation about who owns what comes out of those transcripts is the next conversation, and it is not going to be as fun as the one Dow had on stage. But the infrastructure is the right call. The rest is governance.

— Naomi covers hotel F&B and operator tech for TableTransfers. Tips: [email protected].