01
ask arthur

Building the GTM Engine for Ask Arthur

An AI voice agent for restaurants needed to know exactly who to sell to and how, not just that everyone with a phone was a prospect.

clayclaygentfoursquare places apifind jobsclay sequencer

situation

Ask Arthur builds an AI voice agent that answers restaurant phones, takes orders, and books reservations. They wanted help with GTM and outbound infrastructure. Every restaurant with a phone is technically addressable, which makes the TAM enormous and useless. The real task: find exactly which restaurants Arthur wins fastest and keeps longest.

the build

process map

1Pull the restaurant list
2Drop the ones Arthur won't win
3Score the rest on buying signals
4Sort into the three buyer groups
5Find the right contacts
6Personalize with live facts
7Send the sequenced outreach
8Replies route back by group

impact

  • Three tailored conversations instead of one generic pitch — each built around how that type of restaurant actually decides to buy, so the team wastes less time figuring out who's even a fit.
  • Timely openers that earn replies — a public job posting becomes a same-week email grounded in something real and current about the business, not another cold template that gets deleted.
  • Turnkey handoff — the team runs the whole system without me in the room, and with no ongoing cost.

tech stack

currently using
  • ClayGoogle Maps enrichment, scoring and segment logic, native sequencer
  • Claygentchain and channel classification
  • Foursquare Places APIfoot traffic signal
  • Find Jobsfront of house hiring signal
phase 2
  • Web scraping to backfill reservation platform and menu data
  • A paid firmographic source for faster disqualification
  • Slack alerts on entry to Urgent
  • CRM sync so segment and score travel with the account
  • Geographic territory logic for field reps

reflections

  • The tooling fights back, and handling that is the job. I paid for a data source (Foursquare) before confirming it worked; it broke because the endpoint had moved and Clay's AI request builder silently generated a bad call I only caught by reading the raw response. The gap between clicking a "native" integration and actually solving the problem is exactly where GTM engineering starts.
  • The client made the model better. My first thesis assumed a fast, one-decision sale. The founder pushed back — he'll take a longer cycle if it means onboarding several locations at once. That one comment turned a yes/no filter into three distinct buyer groups. Being wrong in a way the client can correct beats being right alone.
← back to case studies