Conmetior Connect
When someone needs a dentist they search Google and call one of the first practices that looks active. The Google Business Profile is the storefront that decides whether that call comes in. Most practices have not posted to theirs in months, not because it does not matter, but because the work is weekly, unrelenting and nobody's job.
Conmetior Connect is a subscription product that keeps that profile working without adding anything to the practice's week. It writes their posts in the voice detected from their own site and reviews, drafts replies to reviews in whatever language the reviewer used, monitors profile health daily, and publishes only what the practice approves. Approval takes about a minute a week from a phone.
The product is live and functional. It has not yet been sent to its target list, so the numbers below measure the build and the targeting research rather than adoption.
The compliance gate
Dental has a constraint most content tools ignore. Healthcare advertising rules differ by market, and a post that promises an outcome or quotes a price creates a real problem for the practice rather than just a bad impression.
Every generated post is graded against a dental advertising ruleset before a human sees it, with a stricter European variant outside the US. The grader fails closed. If it cannot run, nothing publishes. The product's whole promise is that a practice can stop watching, so the failure mode had to be silence rather than a bad post going out under their name.
Two models split the work. Sonnet 5 writes anything a dentist or a patient will read. Haiku 4.5 handles classification and extraction, where quality is invisible and volume is high.
Ten languages, not ten translations
The product ships in ten languages across 7,600 keys, but the localization goes past the interface. Generated content is written natively in the practice's language rather than translated afterward, review replies match the language the reviewer wrote in, and the compliance ruleset switches standard by market. A validation script runs in the build and blocks retired pricing, unapproved claims and stylistic violations from reaching any locale.
Targeting research: 73,821 contacts down to 300
The client had a customer list of 73,821 contacts across 31,916 business domains and no idea which to email. Sorting by contact volume would have surfaced the DSO chains, which are the opposite of the buyer.
I wrote a pipeline that filtered out chains, government, universities, suppliers and international domains, then visited all 16,784 remaining dental practice websites and scored each on how visibly unattended its presence was and how likely it was to buy. Practices already running Birdeye, Podium or Weave were pushed down automatically, which found 1,899 of them. The output was a ranked first send of 300 with a plain English reason attached to every row.
Rebuilding the funnel around a free audit
The first version asked a cold prospect to create an account, connect their Google profile and enter a card as the first step. Real clicks produced no signups.
The replacement leads with a free audit that asks only for a URL, returns a graded report on how the practice looks to a new patient, and writes two real posts for that specific practice in exchange for an email. It also solved a writing problem. Research on a stranger cannot be used in a cold email without sounding like surveillance, but the same specificity is the product once they have asked for the analysis.
