A Daily Sales Coach Built on Eight Years of CRM Data

The company manufactures barber chairs, salon furniture and spa equipment and sells to businesses and individual customers in Turkey and abroad. Leads arrive from the website, social ads, WhatsApp and export inquiries. A team of six sales reps works these leads inside the company's CRM.

Management could see how many leads came in, but not what happened next. Who answered first? How long did customers wait? Why did some leads move between reps three times? The CRM stored every action, but nobody could read eight years of it.

The owners also wanted fairness. A rep should be measured on working hours, not blamed for a message that arrived at midnight. Customer details had to stay private, even from the people reading the reports.

We connected to the CRM in read-only mode, imported the complete activity history and set the system to refresh every two minutes. On top of that data we designed five screens, each answering one question a sales manager asks every day.

Screens shown with sample names and figures. The live system shows record numbers only and never displays customer names, phone numbers or emails.

The first numbers looked alarming: the median reply time was 84 minutes. Measured within working hours (Monday to Saturday, 08:30 to 18:00) the same data showed a median of 6 minutes. The team was fast. The delay was the messages that arrived in the evening and waited until morning.

That finding shaped the alerts. Instead of blaming reps, the system collects everything that arrives after hours and posts it as one message at 08:30, so the first job of the day is clear.

We added a dedicated assistant to the sales team's group chat. It posts at most one alert per person per day for each of four behaviours, so it stays useful instead of noisy:

• Late reply. A customer wrote and no one has acted for 20 working minutes.
• Unusual hand-offs. A rep passes their own leads to colleagues more than five times in a day.
• No activity. Two hours into the working day with nothing logged.
• Low usage. Fewer active hours than expected by 17:00.

Every working day at 18:00 it posts a summary. Managers can also ask it questions in plain language, and it answers from the CRM together with the company's ad performance by country.

The system can read but never change CRM data. Customer names, phone numbers and emails are not shown anywhere; reports use record numbers with a link back to the CRM. Message contents are not stored. Automation and admin accounts are flagged and kept out of the team comparison, so no rep is measured against a bot.

We show the limits on the screen itself, so nobody reads more into the data than it holds.

Management now opens one screen instead of asking around. Reply speed is discussed with numbers the team considers fair, after-hours messages are handled first thing in the morning, and the AI manager raises waiting customers while they are still waiting.