I feel like a lot of people get caught up in building cool AI agents without really asking whether the business actually needs one. For example, imagine a business getting 50–100 enquiries every day. The problem might not be getting more leads at all. It could be that they're taking too long to reply, forgetting to follow up, or wasting time dealing with enquiries that aren't a good fit. Now, you could build a complex AI agent to handle the whole process, but honestly, I don't think everything needs AI. A simple workflow using n8n, webhooks, a CRM, and a little LLM logic could probably handle most of it. Something like: - Capture incoming leads automatically. - Use an LLM to understand the enquiry and qualify the lead. - Update the CRM and assign the lead to the right person. - Trigger follow-ups automatically. - Pass anything complicated to a human. The interesting part is figuring out what actually needs AI and what can be handled with simple rules. Using an LLM for everything just adds cost and more potential points of failure. I also think we should focus more on the results these automations produce rather than just how impressive they look. Things like response time, booked appointments, conversion rates, and hours saved are what really matter to a business. I'm curious about something: When you're building automations for clients, do you usually start by finding a problem in their existing workflow and then build a solution around it, or do you start with an automation idea and see whether the business needs it? Would love to hear how you guys approach this.