Why most businesses shouldn’t start their AI journey with a chatbot
The most visible application of AI is not always the most valuable place to begin.
Start one question earlier.
A chatbot is easy to imagine. It provides a familiar interface and a tangible answer to the question: what could we do with AI? But that question arrives too late in the sequence. Before deciding what to build, a business needs to decide what it is trying to improve.
The starting point might be a team repeatedly assembling the same information, a decision delayed by missing context, or a handover that regularly breaks down. None of these problems comes with a predetermined interface. Understanding the work should shape the solution.
Follow the work, not the demonstration.
A useful discovery exercise follows one task from beginning to end. Where does the information originate? Who checks it? Which decisions depend on it? What happens when the usual path fails? The answers reveal the surrounding work that a polished demonstration can leave out.
Sometimes a conversational interface is exactly right. A person may need a flexible way to explore a large body of information or ask an unfamiliar question. In other situations, the better experience is a prepared brief, a flagged exception or a completed draft inside an existing system.
Choose something you can learn from.
A first AI initiative should make an assumption testable. Define a specific task, a clear starting point and a useful measure of improvement. That might include preparation time, completeness, review effort or the number of handovers required. Choose the measure before the prototype changes the workflow.
Keep the first scope narrow enough to inspect. Give the people doing the work an active role in testing it. Their experience will expose whether the system helps, moves effort elsewhere or creates new work that the initial demonstration did not reveal.
From our work: capture in 2ndChannel.
In 2ndChannel, the starting point is a physical item. The phone workflow asks the operator for photographs and a voice note; the desktop workspace then brings the prepared listing together for review. That sequence shaped the interface around the operator’s contribution and the decision they need to make.
The useful question for a pilot is whether the whole journey works, including upload, preparation, corrections and approval. A fast first draft alone cannot establish a time saving.
The interface is a consequence.
The aim is to create a better way of working. Once the outcome, constraints and responsibilities are understood, the right interface becomes easier to choose. It may be a chatbot. It may be something quieter and more useful.
Start with the business problem. Let the technology earn its place in the answer.
Intelligence in practice.
