Saltar al contenido
People & Play
People & Play
People Analytics & AI

5 min de lectura

Prompts without context: when AI answers well the wrong question

A prompt without context can sound perfect and solve the wrong problem. David Botella explains why your own context matters more than the model you pick.

Person typing on a laptop in front of an AI chat screen, a symbol of a prompt without context

David Botella

CTO · 6 Nov 2026

Índice de contenidos

1. The perfect answer to the wrong question2. When the prompt copies the question and loses the context3. AI answers what you write, not what you mean4. Three habits to give context before asking for an answer5. Questions before trusting an AI answer6. A good prompt starts by explaining the problem, not the question

The perfect answer to the wrong question

More and more people teams use generative AI to draft communications, analyse surveys or prepare a talent decision. The result is usually fluent, well written and, often, completely useless, because it answers a generic question instead of the organisation's actual problem.

  • A prompt without context gets an answer that is correct for any company and useful for none.
  • Fluent generated text doesn't guarantee it solves the real problem.
  • AI doesn't know your context unless someone gives it to them.

When the prompt copies the question and loses the context

We see teams that ask to "draft a talent retention plan" and get back a generic document that could serve any company in the sector. The real turnover figure is missing, what was already tried and failed is missing, the available budget is missing.

AI doesn't invent that context, it simply fills the gap with the most likely pattern in the language. It sounds good and steers badly.

  • Generic action plans nobody can execute without rewriting them entirely.
  • Survey analyses that repeat common places without touching the actual data.
  • Communications that sound corporate and don't sound like the team's real culture.

AI answers what you write, not what you mean

Our reading is that the problem is almost never the model, it's the prompt. A powerful model with a poor instruction produces a poor answer that looks confident.

“AI without context isn't wrong, it answers well the wrong question.”

— David Botella

CTO at People & Play

That difference matters because a fluent, wrong answer is more dangerous than a clearly incomplete one.

Three habits to give context before asking for an answer

You don't need to be an AI expert. You need the discipline to explain the problem before asking for the solution.

  1. Describe the context before the task
    • Include your own data, what was already tried and the real constraint at play.
    • A prompt with three lines of context beats one with perfect instructions and none.
  2. Ask for the reasoning, not only the result
    • Ask why the AI proposes that option and not another one.
    • If the reasoning doesn't hold up, neither does the answer.
  3. Review the answer with someone who knows the real context
    • A team member validates whether the imported context was applied correctly.
    • Human review is not a final formality, it is part of the process.

Questions before trusting an AI answer

Before taking an AI answer into a people decision, it is worth asking:

  • Did we give the AI our organisation's real data, or just the general question?
  • What constraint of ours doesn't appear in the prompt and would change the answer?
  • Did someone on the team validate the reasoning, not only the final text?
  • Would this answer serve equally well for any company in the sector?
  • What would happen if we added the missing context and asked again?

A good prompt starts by explaining the problem, not the question

Generative AI multiplies the judgement that already exists on the team. It doesn't replace judgement that was never made explicit.

If you want to build prompts that bring in your real context before asking for an answer, let's talk. The first change is usually more context, not a better model.