Sobot has upgraded its AI Agents so they aim to finish customer requests rather than only answer questions. The company, which positions itself as an agentic customer contact platform, said the change is not just a better model but a different way of working. Earlier Agents followed fixed workflows triggered by detected intent and could not adapt well outside those paths. The new Agents run a continuous loop of reasoning, acting and observing, known as ReAct. In a return request, for example, the Agent can work out what information is still missing, check what the customer has already provided and ask only for the rest. A unified resource layer of knowledge, skills, tools, memory and variables is shared across Agents. Building an Agent is now conversational: users describe the goal and the Studio assembles the needed settings. Managers can also ask performance questions in natural language instead of digging through dashboards.
Sobot has upgraded its AI Agents so they aim to finish customer requests rather than only answer questions. The company, which positions itself as an agentic customer contact platform, said the change is not just a better model but a different way of working. Earlier Agents followed fixed workflows triggered by detected intent and could not adapt well outside those paths. The new Agents run a continuous loop of reasoning, acting and observing, known as ReAct. In a return request, for example, the Agent can work out what information is still missing, check what the customer has already provided and ask only for the rest. A unified resource layer of knowledge, skills, tools, memory and variables is shared across Agents. Building an Agent is now conversational: users describe the goal and the Studio assembles the needed settings. Managers can also ask performance questions in natural language instead of digging through dashboards.