I’d look for tasks where people spend a lot of time collecting information, making decisions, or doing repetitive work. If AI can reduce that friction in a way that’s actually measurable, that feels like a strong starting point for an AI SaaS.

• This reply was modified 18 hours, 39 minutes ago by  Christopher.

I’d probably build around a problem that people already struggle with repeatedly, rather than trying to create something just because AI makes it possible. The AI should solve the problem, not be the whole reason for the product.

I’d look for a niche where people are still relying on spreadsheets, emails and repetitive manual work then build an AI product that connects those pieces into one workflow. The best opportunity might not be the most exciting problem on paper but the one where removing that daily friction creates obvious value for the customer.

If I launched an AI SaaS today, I would solve back office work for regulated small businesses like insurance brokerages, medical billing, and freight forwarding. These teams run on email, PDFs, and spreadsheets. AI can finally do that work end to end because the inputs are messy and the rules are implicit. The moat is context, compliance, and workflow depth, not the model itself.

Start with one capable agent as the user facing interface, but architect it so specialists can plug in behind it later. Users want a single assistant with shared memory and consistent personality. Behind that interface, specialized agents can handle routing, actions, retrieval, and domain reasoning. The mistake is building one agent in a way that makes many a rewrite, or building many before you know the jobs.

I’ll probably build something that solves a real problem people have every day. There are already many AI tools, so I’d focus on making something that saves time or makes work easier. The AI should be useful, not just added for the sake of it.

The first thing every sane producer builds when considering building a SaaS should be a consumer community where your customers can come together and discuss various features about your SaaS which would in turn give you insight about what works and what doesn’t. Its important to have genuine users in this community because only then would you truly understand what they want.

I’d avoid building another generic AI assistant and instead target a workflow where AI can own a measurable outcome. For example, build something that takes a messy business process support requests, invoices, lead qualification, internal approvals and manages it from intake to completion, involving humans only when necessary.