With My experience i am telling you You can make more on Money on building Agents for Non-Tech Startups
retained text
When I joined an IoT company, the founder asked me to work with the sales team. I’d spent the previous three years building SaaS products, and I wanted to understand an industry I hadn’t worked in before. The company already had customers and was getting a lot of inbound enquiries through Google Ads. They had been around for about six years, but the sales team was still managing everything in spreadsheets. I started with a CRM and a Kanban board. To decide what to build around it, I had to understand how an enquiry became a customer. Every new lead needed an initial call. The salesperson asked a few qualification questions about the customer’s requirements. We had a minimum ticket size, so smaller opportunities went to sub-dealers while our team handled the larger ones. That initial conversation became one of the first parts I automated. I built an AI voice agent that called someone as soon as they submitted an enquiry with their phone number. It asked the qualification questions, and we used an LLM to evaluate the answers and score the lead. The call was connected to the next part of the workflow too. We tracked how many leads each salesperson already had and used that workload, along with deal value, to assign new leads. Higher-ticket opportunities went to more senior people. So the enquiry could move through an initial call, evaluation, scoring, and assignment before the salesperson continued working on it. I enjoyed building that connection because the team needed the outcome of the conversation to be useful inside the CRM. The answers helped us decide how to handle the opportunity and who should take it. After that, there was still a long sales process ahead. Qualified leads needed an estimate of device requirements and a survey. We prepared a quote and sent it through WhatsApp using the CRM integration. Customers often needed internal approval, which meant months of follow-ups before we received a purchase order. The full cycle usually took three to six months. The voice agent handled the early qualification work and saved the team a lot of time there. The surveys, quotes, and later customer conversations remained part of the team’s work. I handled the product from design to deployment, including the integrations and the bug fixes. Once it had been stable for about a month, I couldn’t think of any new features to build. I started using the application as a salesperson, handling acquisition and lead verification myself. That changed how I saw the product around the voice agent. I found 30 features we didn’t need, removed them, and added features people actually wanted. Those changes came from using the application to do the work the team was doing. The team hadn’t been able to put all their needs into words. I had to get involved in the daily work to understand some of them. That’s something I’d keep in mind when building another agent into an existing workflow: spend time using the application that people rely on around it.
document fetched 2026-10-12 00:18 · canonical https://reddit.com/r/AI_Agents/comments/1x2zokk/with_my_experience_i_am_telling_you_you_can_make