Let’s talk

FIELD NOTE / PRACTICAL GUIDE

Prompt engineering vs AI engineering

Understand the difference between writing and testing model instructions and building reliable software around AI.

The two areas overlap, but the evidence of competence differs. A strong prompt can improve a model response; a reliable application also needs data flows, integrations, access control and operations.

What prompt work covers

Define the task, organise source material, provide examples and compare outputs against acceptance criteria. For a support draft, check that the answer uses the approved policy and identifies missing information. Store the test cases so later changes can be compared.

What application engineering adds

Build and maintain the surrounding system: authentication, data storage, API calls, retries, input validation, logs and deployment. An assistant that produces a useful answer in a chat is not evidence that a customer-facing application is ready to operate.

Where the disciplines meet

A knowledge assistant needs both useful instructions and reliable retrieval of approved source material. The team must test wrong, incomplete and conflicting documents, decide when a person takes over and prevent unintended actions. These responsibilities should have named owners.

Choose your next learning step

If you can write a good draft but cannot evaluate it, practise source checking and test design. If you can evaluate outputs but need a production integration, learn the relevant engineering skills or collaborate with a qualified developer. Build a demonstrable project without claiming skills you have not tested.

Keep this checklist

  • Task quality separated from system reliability
  • Tests include incomplete inputs
  • Integration owner identified
  • Learning plan matches the missing skill

This is an original practical guide. Examples are illustrative; tool interfaces and available features may change.

Put it into practice.