Chatbots: Are We Automating What Should be Eliminated?

A team in HR builds a chatbot to walk hiring managers through a 14-field hiring request form. It takes three months and a meaningful chunk of their budget. When they launch it, the percentage of correctly completed forms improves from 30% to 75%. There is much rejoicing. “AI Wizard” coffee mugs are awarded.

Then a hiring manager asks a question nobody had thought to ask: how many of those 14 fields does HR actually use—and how many are asking for data that HR already has?

The answer is: only four of the fourteen fields needed to be collected.

• Four fields collected information that went nowhere.
• Six could have been auto-populated from data the organization already had—department code, financial code, and the classification from the last time they filled the same position.

In other words: ten of the fourteen fields didn’t need to be asked of the hiring manager.

The form should have had (net) four fields: clearly labelled fields with well-written, plain-language instructions. In many organizations, that’s the kind of request a hiring manager could complete in a few minutes—without needing a chatbot, nor help from HR. And this simple form could have been developed quickly and easily.

Instead, the organization built an AI solution to compensate for a form that likely shouldn’t have been so complicated in the first place.

I see some version of this mistake a lot in knowledge work: automating waste instead of eliminating it.

Making the wrong thing easier for the service provider instead of making the right thing simple for the client.

Even if a chatbot helps the hiring manager fill in those extra fields, the manager still experiences the process as overly complex—and draws a quiet conclusion about dealing with HR in the process.

That time the hiring manager spends searching, interpreting, and explaining may be guided by AI, but it’s still time the client didn’t expect to spend.

That frustration ends up being aimed at HR, not the chatbot. Reducing the form to only what genuinely matters saves effort, speeds up the experience, and signals competence.

And once the chatbot exists, it can become a barrier to simplification.

Someone will say: "We can't redesign the form now—we just invested in a chatbot that uses those 14 fields." The technology that was supposed to improve the process now protects the broken process from being fixed.

The principle is straightforward: eliminate the unnecessary complexity first. Then, if the remaining process is still genuinely complex—regulatory requirements, multi-party coordination, legitimate technical detail—use AI to guide users through what's left.

You may say: a modern AI agent doesn't need a form at all - it can extract the information from a 30-second conversation with the hiring manager. True. And the question still applies. Are you sure you need to collect all 14 pieces of information? Are you sure HR uses what's collected? An AI agent that gathers irrelevant data faster is still gathering irrelevant data. The eliminate-first principle isn't about the interface. It's about the work.

A chatbot for four essential fields that require real guidance? That can be smart.

A well-designed process needs less AI. 

Key takeaways:

• Before building any AI solution, ask: should this step or field exist at all?
• Eliminate unnecessary complexity first. Then use AI for what's left.
• A chatbot for a bad form is still a bad form.
• Technology that compensates for poor design becomes a barrier to fixing the design.

The right question isn't "can AI help here?" It's "should this process be this complicated?"

Craig Szelestowski is a recovered executive, the founder of Lean Agility Inc. an instructor at the Telfer Centre for Executive Leadership, University of Ottawa, and a Subject Matter Expert at the New Jersey Institute of Technology.