The Register a Bot Should Use for a Difficult Question
Pay, grievance and personal circumstances need a different voice from annual leave. What to specify, and where the bot should decline to be warm.
Operational playbooks, strategy frameworks, and AI-ready guidance for leaders building modern teams.
Explore a more focused slice of the newsroom, with each story designed to be useful in real teams and real decisions.
Pay, grievance and personal circumstances need a different voice from annual leave. What to specify, and where the bot should decline to be warm.
HR owns the content, IT owns the platform, and the answers belong to neither. The three jobs that need a name against them.
A bot trained on your documents can answer from all of them. What strict mode is for, what it still cannot stop, and the questions to ask a vendor.
One corpus, several audiences. Where document-level permissions break down, and the three populations most companies forget to separate.
Every unanswered question is a documentation gap somebody has already asked about. How to read the queue and what to write first.
Most HR questions need an answer. Some need an action. Buying one when you needed the other is the commonest mistake in this category.
Structure beats prose quality here. One fact per paragraph, headings that match questions, and the four formatting habits that break retrieval.
Only four of eight vendors publish a language count, two publish the same figure, and one contradicts itself on its own homepage. What to check before you buy.
The categories where machine translation is good enough, the ones where it is not, and why the failures are quiet rather than obvious.
The four question types that need a person regardless of how good the tool is, and why a vendor that admits this is safer than one that does not.
Tell us about your team and we'll match you with the right resources, tools, and guidance - at no cost.