Lessons from projects and descriptions of solved problems
A company asked how much ranking in ChatGPT costs. I checked what its site sends to AI crawlers: five findings, what the research says and how to measure visibility.
The system organized calls to companies and used AI before and after the conversation. Only practice showed where technology really helps, and where it merely bypasses the hard part of sales.
More information does not always give a better answer. On privacy, control and giving artificial intelligence only as much context as the task at hand requires.
A list of technologies alone does not help a business owner judge a contractor. How I rebuilt the site around companies’ problems, case studies and two simple ways to get in touch.
Instead of a long form, the client describes the matter in their own words. The assistant organizes the information, and I personally decide on a quote, an audit or a further conversation.
Every fix was supposed to remove one error, but after a dozen or so rules the assistant became rigid and lost the thread of the conversation. What I changed in the way I design it.
One notification outage uncovered more serious problems with payouts, access to data and the charging of fees. How I fixed them without throwing the whole system away.