Agents

tl;drAn AI agent is software that can understand a goal stated in plain language, decide what steps are needed to reach it, and carry those steps out without a human doing each one by hand.

In accounting, an agent might read an incoming bill, code it to the right account and cost center, check it against a purchase order, and route it for approval, then post it once approved. The person who owns the process states what should happen. The agent does it.

This differs from older automation. Rules-based automation and RPA scripts fail the moment a case falls outside the rule: a new vendor format, a missing field, an exception that needs judgment. An agent handles the exception too. It investigates what's missing, asks the right person for it, and finishes the job, rather than stopping and creating a task for a human to pick up later.

For a finance team, the practical difference shows up in where time goes. Work that used to require someone to open a file, check three systems, and make a judgment call now runs in the background, inside the controls the team already has: roles, approval workflows, an audit trail. The team doesn't disappear. It moves up, from processing the transaction to setting the policy the agent runs by.

At Light, agents work through the Light Command Interface (LCI), in Slack, Teams, or the web app, with the same permissions and audit trail as the people they work alongside. Every action is logged, attributable, and reversible. Agents run the operational layer so the finance team can spend its time on the decisions that actually need a person.

Back to the glossary