AI Roadmapping Proposal

This proposal lays out an eight-week assessment to turn six AI opportunity areas inside a mid-market NYC PR firm into concrete workflows, technology decisions, and an implementation roadmap. Rather than starting with tools, it starts with the work itself: map the current process, identify bottlenecks and decision points, design the target workflow, then evaluate models and vendors against those requirements. AI is reserved for tasks that require interpretation; rules-based systems handle calculations, checks, routing, and other deterministic work.

A worked invoice-processing example shows the level of detail behind that approach. It follows a vendor invoice from receipt through payment, shows how missing operational evidence can break automation, and separates what AI can assist with from what Finance still needs to approve. The goal is not to promise savings before the work is understood, but to determine where automation is actually viable, what should be bought versus built, and what implementation would require in terms of systems, owners, cost, and change.