Probably a $50-150k MRR business hiding inside the problem of bespoke internal workflow automation for small to mid-sized companies.

I think this looks like an AI agent that understands natural language requests to connect multiple internal tools, possibly even non-standard APIs or legacy systems. Imagine telling it 'When a new support ticket is marked critical in Zendesk, check our internal knowledge base for similar issues, then notify the on-call engineer in Slack with a summary and a link to the relevant article, then create a task in Jira.' The agent would learn and adapt to unique API endpoints and data schemas over time. The customer isn't looking for a Zapier clone; they need something truly custom that can handle edge cases without constant developer intervention.

The part I am unsure about is how truly 'autonomous' these agents can be without significant setup and ongoing training. My concern is whether the initial configuration costs and the complexity of managing potentially flaky integrations might outweigh the benefits for the target SMB. Is the 'last mile' of truly bespoke integration still too human-intensive to be profitable with AI?