How an outpatient mental health clinic in Los Angeles added about $2M in a year

Its phones were staffed from about 5 AM to 3 PM Pacific. Everyone who reached out in the evening, overnight or on a weekend got voicemail or nothing. The Tars chat agent now answers them at any hour and sends each one to the team by morning.
At a glance
- Who: an outpatient mental health clinic in Los Angeles (name withheld)
- Problem: 100,000+ website visits a month, phones staffed about 10 hours a day
- What Tars built: a website chat agent that answers questions, captures name, email and phone, and sends each lead to Salesforce
- Live in: about 3 to 4 weeks
- Next: the Tars voice agent for the after-hours phone line
The problem: 100,000 visits a month, 10 staffed hours a day
People rarely decide to look for a therapist at 10 AM on a Tuesday. They look in the evening after work, late at night when they can't sleep, or on a Sunday when there's finally time to think.
This clinic's website drew more than 100,000 visits a month. Its intake and sales team worked from about 5 AM to 3 PM Pacific, early enough to cover part of the East Coast morning. People who reached the team in those hours were looked after well. The problem was the other 14 hours of the day.
After-hours calls went to voicemail. From 3 PM until 5 AM, and all weekend and on holidays, nobody answered. Many callers hung up without leaving a name or number.
Website visitors had nowhere to ask. Someone reading the site at 9 PM with a question about programs, cost or insurance couldn't get an answer on the spot. Most left.
The next-morning callback came too late. By the time the team called back, many people had moved on. For the ones who left no details, there was nobody to call.
The clinic was attracting the demand. It was losing much of it in the hours when people most often look for help.
What Tars built: a website chat agent that never closes
Tars put an AI chat agent on every page of the clinic's website. It runs at all hours, and it does its most important work when the office is closed.
It answers questions in plain language. It answers from a knowledge base built from the clinic's own website: services, how programs work, what to expect and how to get started. Only from that approved content, not the open web.
It helps first, then asks. Once it has answered the visitor's question, it asks for a name, email and phone number so the team can follow up. No form wall in front of an anxious visitor.
It sends every lead to Salesforce. Each lead lands in the clinic's Salesforce CRM automatically. The team starts each morning with a list of named leads instead of anonymous voicemails.
Going live took about three to four weeks: building the knowledge base from the website, connecting Salesforce, testing the agent and its guardrails with the clinic's staff, then adding one small snippet to the site. Monthly reviews over the first quarter tuned the answers and the moment the agent asks for contact details. Ask too early and people leave. Ask too late and they get their answer and go.
The results, quarter by quarter
About 1 in 10 conversations ended with the visitor leaving their details. The team turned about 100 of those leads into new patients each quarter.
- Website conversations: ~14,000 a quarter, ~56,000 a year
- Leads captured and sent to Salesforce: ~1,200–1,400 a quarter, ~5,000 a year
- New patients: ~100 a quarter, ~400 a year
- Added revenue (at ~$5,000 per patient): ~$500,000 a quarter, ~$2M a year
What's next: the phone line
The chat agent caught website visitors after hours. People who called after 3 PM still reached voicemail. The next step for the clinic is the Tars voice agent on the after-hours line: it picks up, answers questions, takes the caller's name, phone and email, sends them to the CRM, and can book a callback or transfer to staff when someone is available.
If we started this project today, we'd cover the website and the phone line from day one, so every channel is answered at every hour.
How we counted
All figures are approximate and rounded, covering about the first year the chat agent was live. Revenue counts every new patient who came through a chat lead, at about $5,000 per patient. It is revenue the chat agent influenced, and it doesn't subtract people who might have reached the clinic another way.
Want to see where your own center stands? Take the 2 a.m. test →

My name is Vinit Agrawal and I am the Founder and CTO of Tars. I have been working on the Tars Chatbot Builder since 2015 and my core strength is in building software products with simple and functional user experiences focusing on bringing some core business results. My current role in the company is a mix of Product Manager, Engineering Manager and in Business & Marketing Strategy.
Chosen by 800+ global brands across industries
Recommended Customer Success stories

How Rate My Therapist Doubled Visitor Engagement by Making an AI Agent Its Front Door

How Gama Hospital Turned Its Crowded Lobby Into a 24/7 Multilingual AI Front Desk









