Robotic Process Automation (RPA) vs AI Agents
RPA follows fixed scripts across GUIs; AI agents plan, adapt, and handle ambiguity using language models and tools.
In plain English
RPA — Robotic Process Automation — is software that drives the user interfaces of other software. A UiPath or Automation Anywhere robot opens the HR app, clicks through the new-hire form, and pastes data from a spreadsheet. It is brittle: rename a button, move a menu, and the robot breaks. But for legacy systems with no API, it is the only bridge that exists.
AI agents take a different approach. Instead of scripting UI clicks, they reason about the goal, use APIs when available, and fall back to browsing only when required. When something unexpected appears — a new CAPTCHA, a reworded error message — an agent can often figure it out; an RPA bot cannot. The two stacks increasingly coexist: RPA for the predictable stretches, agents for the judgment calls in between.
Why it matters for Black Box
Black Box's Browser specialist is agent-first: it reasons about what the page is doing, not what pixel to click. It uses APIs when available, the browser when not. The Action Feed shows what it tried and why, so owners never debug a brittle click path blind.
Examples
- RPA: a scripted bot that copies data from an old desktop app into a modern web app.
- AI agent: a specialist that reads a vendor portal and extracts the invoice, even if the layout changes.
- Hybrid: an agent that plans and delegates the deterministic bits to an RPA bot.