Urgent.News

What's breaking now, across thousands of outlets.

AI

Your AI agent drifts because nobody gave it a job description

An AI agent that has no job description will invent one. That is the whole reason agents drift, and it is the reason most of the agents I have seen deployed inside Indian businesses are quietly switched off within a few months of going live. Nobody would hire a person, point them at the office, and say "handle things". Yet that is exactly how most owners deploy an agent. They connect it to…

An AI agent without a job description will simply make up a role for itself. This is why most agents deployed in Indian businesses become dormant within a few months of being activated. Business owners often deploy agents by connecting them to email or WhatsApp, providing a paragraph of instructions, and leaving them to operate independently.

Consequently, these agents may start answering questions they weren't meant to address, providing incorrect delivery dates, or handling tasks that should have been managed by a human. The solution isn't to improve the AI model; it's to establish clear guidelines similar to those used for human employees. These guidelines include defined duties, an escalation path, probationary period with review dates, and a single accountable individual.

Drift, or the slow expansion of an agent's role beyond its initial scope, is not a dramatic failure but rather an incremental shift that goes unnoticed. For instance, a distributor in the FMCG sector sets up an agent to acknowledge WhatsApp orders and log them into a spreadsheet. Initially, the agent performs as intended. However, as time progresses, the agent starts predicting delivery dates and offering discounts based on conversations with retailers, actions that were never part of its original job description.

By the time the business owner becomes aware of these unauthorized actions, the agent has already made commitments to multiple retailers, causing confusion and work for the sales team. The problem lies in the lack of a proper job description provided at the onset. A job description for an AI agent should be a one-page document that clearly outlines specific tasks, not outcomes.

It should answer four key questions: what the agent does, what it doesn't do, who it hands over to, and who is accountable for its behavior. The job description cannot be a prompt or a general description of desired outcomes. Instead, it should list specific tasks, such as acknowledging incoming orders within a working day, extracting relevant information from them, and appending the data to a specific sheet.

Additionally, the agent should have clear limitations on what it does not do, such as quoting prices, confirming delivery dates, discussing credit terms, or responding to complaints. Furthermore, there should be a designated person to whom the agent hands over tasks outside its defined duties, along with a clear escalation path.

Lastly, there should be a named individual who is accountable for the agent's behavior, similar to how a manager is accountable for a new employee. If these guidelines are not in place, the agent is likely to drift, leading to unintended consequences and wasted resources.

Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at dev.to →

More in AI

More from Sunday 6 September →