Don’t raise the cost of bad writing

AI does not eliminate the need for good writing – it raises the cost of bad writing. 

Let’s look at every major player, Stripe, Intercom GitLab, Zendesk, Microsoft Copilot, Atlassian. Each one is heavily invested in AI, and each one can expect to see resource costs go up, not down, unless their AI agents are fed the right info. Humans remain the backbone of good agentic experiences. 

I realize that when we are all knowledge gatekeepers (whether tech writers, L&D specialists, or any instructional content creator), it is tempting to think that we are shouting into an echo chamber. That we think one way, but the market thinks another. When that thought creeps in, remind yourself of how you felt the first time you set up your AI chatbot and ran a pilot test on it. Was that a good experience? Even with all the access it had to existing content, was it providing the clearest answers? 

If you are smiling and shaking your head while reading that, it is because you know the answer. 

It is not a riddle. What does it take to provide a great AI chatbot experience to users?

Hours, hours, and hours of human labor. Writing AI-optimized docs, inputting guidelines, giving instructions, teaching the chatbot, setting up secure connectors. 

It is a constant effort (it was a job in itself at my previous company), and the minute you let up, the user experience gets negatively affected. 

And I know that work doesn’t always fall on the dedicated specialist. It might be a Support Agent put on dual maintenance, an Automation Specialist running the backline, a CSM helping to set up an agentic sales flow, or a Marketer building a discovery workflow. Whatever your role, if agentic related work lands on your plate – the same truth applies: without that near constant effort, AI intelligence goes down. 

This is what I know, this is where we are now. Organizations that invest in human-authored, well-structured content see dramatically better AI outcomes. End of sentence.