The objective problem with AI work
Writing learning objectives has always required observable, measurable verbs. The principle does not change when the job involves AI. What changes is what "observable" means when the task is partly performed by a machine.
If a learner asks Copilot to draft a report and then edits it, what is the observable skill? Not "write a report" — but "judge the output, identify the gaps, and direct the revision." That is a harder objective to write. It is also the real one.
By the end of this unit- Write learning objectives for roles that use Copilot and agents in daily work.
- Identify ESCO-aligned competency language for AI-augmented tasks.
- Apply the delegation model as a framework for learning objectives.
The three-layer objective for AI work
When a task is AI-augmented, the observable skill sits in three places: before the AI does anything, while directing the AI, and after reviewing what it produced.
Before: framing the task
The learner must be able to define the task clearly enough to prompt the AI effectively. This is a pre-delegation skill and must be taught as an objective in its own right.
During: directing and adjusting
The learner must be able to assess whether the AI output is moving in the right direction and give corrective instructions. This is a real-time quality judgement skill.
After: reviewing and owning
The learner must be able to review the AI output against the original brief and take responsibility for the final product. This is where professional accountability lives.
Take a task that your target learner now completes with Copilot. Can you write one observable objective for each of the three layers above?
ESCO and AI-era competencies
ESCO — the European Skills, Competences, Qualifications and Occupations framework — provides a structured taxonomy of skills and competences across occupational categories. It is useful for ID because it gives you standardised, employer-recognised language for writing objectives.
Why ESCO matters for AI-focused learning design
As AI tools change job tasks, ESCO is being updated to reflect the competences that AI-augmented roles actually require. These fall into three categories that every ID practitioner designing for AI-using workforces needs to know.
Transversal competences — the foundation
Objective example: "Evaluate AI-generated outputs against defined quality criteria and identify errors that require correction." (Critical thinking, applied to AI context.)
Digital transversal competences
Objective example: "Apply appropriate prompting strategies to direct Copilot to produce a specific deliverable, and evaluate the output for accuracy and relevance."
Occupation-specific competences
Objective example (HR analyst): "Use Copilot to generate a first-draft job description from a role brief, and revise the output to meet the organisation's grading framework."
Do not write an objective that says "use Copilot to do X." Write an objective that says "produce X, using Copilot as a tool." The accountability sits with the learner, not the tool.
The ESCO verb set for AI-augmented work
These verbs appear in ESCO competence descriptions and translate well to observable ID objectives for AI-using roles:
For objectives about directing AI tools:
Instruct, specify, brief, configure, adapt, adjust, refine, iterate, redirect, constrain.
For objectives about evaluating AI output:
Evaluate, assess, critique, verify, validate, identify (errors), compare (against criteria), prioritise (revisions), flag.
For objectives about final work products:
Produce, draft, compile, synthesise, present, deliver, document, communicate, publish.
Delegation as a learning target
The Copilot Cowork model frames AI interaction as delegation — giving the AI a task with enough context and constraints for it to produce useful output. This is a learnable skill, and it can be written as a series of observable objectives.
The four-phase delegation model
Outcome — what should exist when the task is done
The learner defines the deliverable in precise terms before prompting. Objective: "Specify the format, audience, length, and purpose of a deliverable before initiating a Copilot prompt."
Context — what the AI needs to know
The learner provides the background, constraints, and relevant information. Objective: "Identify the context elements required for a Copilot task and include them in the prompt."
Constraints — what the AI should not do
The learner specifies scope limits, tone requirements, and exclusions. Objective: "Apply appropriate constraints to a Copilot prompt to reduce the need for post-generation editing."
Quality check — does the output meet the brief
The learner evaluates the output against the original outcome definition. Objective: "Evaluate a Copilot output against defined criteria and identify revisions required."
A learner is asked to "use Copilot to write a business case." Which objective best captures what they need to be able to do?
Measuring AI-augmented performance
Evaluation in AI-augmented learning is complicated by the same thing that makes objective-writing harder: the AI does part of the work. A learner who produces an excellent Copilot-assisted document may have excellent prompting and evaluation skills — or may have got lucky with the output.
What to measure and what not to
- The quality of the learner's prompt (specificity, context, constraints)
- The accuracy of the learner's evaluation of the AI output
- The decisions the learner made in revising the output
- Whether the final product meets the original brief — and who is responsible for ensuring that
- The quality of the AI output alone (the AI, not the learner, produced this)
- Speed of task completion (AI will always be faster; this is not a human skill)
- Whether the learner used AI at all (the tool is not the learning target)
Design an assessment task for a learner who uses Copilot to draft a project status report. What would you ask them to submit as evidence of their skill — not just the document?
End of Unit 2
You should now be able to:
- Write three-layer learning objectives for AI-augmented tasks (before, during, after).
- Use ESCO verb categories to build observable objectives for AI-using roles.
- Apply the delegation model as a framework for both objectives and assessment design.
Continue to the unit review.
Unit review
Which ESCO competence category is most directly relevant to evaluating AI outputs?
Why is "use Copilot to write a proposal" a weak learning objective?
The four phases of the delegation model are Outcome, Context, Constraints, and…?
What is the most useful piece of evidence to assess a learner's Copilot skill?
End of module
You have completed Course 02: Writing Outcomes for AI-Augmented Roles. Next: Prompting for Learning Design — using Copilot to generate, accelerate, and improve your own ID deliverables.