Module 1 · Lesson 5

Where AI Should Assist — and Where Humans Should Stay in Control

Good AI adoption is not about deciding whether humans or AI should do the work. It is about deciding which parts of the workflow each is best suited to handle.

✓

By the end of this lesson

You will know how to divide marketing work between AI and people, choose appropriate levels of human oversight, and recognise decisions that should remain firmly under human control.

The human-AI workflow

Delegate → Review → Decide → Approve → Learn

Delegate

Give AI the parts of the task it can perform efficiently.

Review

Check output against the brief, evidence and quality standard.

Decide

Apply commercial judgement to the available options.

Approve

Keep meaningful accountability with an appropriate person.

Learn

Use outcomes to improve prompts, processes and future decisions.

AI strengths

Where machines often have the advantage

Volume

Process more information

AI can review large amounts of text, data or creative material faster than a person can manually.

Speed

Create first passes quickly

Drafts, summaries, variations and classifications can often be produced in seconds.

Consistency

Repeat defined workflows

AI can apply the same instructions repeatedly without becoming bored or distracted.

Pattern Finding

Surface useful signals

AI can help identify recurring themes, anomalies and relationships across large datasets.

Variation

Explore more options

Generate multiple creative or strategic alternatives for comparison.

Availability

Work on demand

AI can support marketers whenever a task arises rather than waiting for specialist availability.

Human strengths

Where judgement still matters most

Commercial Judgement

Understand what matters

Humans decide which outcomes genuinely serve the business rather than merely optimise a metric.

Context

Understand nuance

Internal politics, customer relationships, operational constraints and strategic priorities may not exist in the AI's context.

Taste

Judge creative quality

Strong marketing still requires judgement about tone, originality, emotion and brand fit.

Ethics

Decide what should be done

A technically possible marketing tactic is not automatically appropriate.

Relationships

Handle people

Negotiation, trust, empathy and sensitive customer situations often need genuine human involvement.

Accountability

Own the outcome

Businesses and people remain responsible for important marketing decisions and published claims.

Think in levels of autonomy

AI does not have to be either manual or fully autonomous

Level	AI role	Human role
1. Assist	Generate ideas or analyse information	Human performs and approves the work
2. Recommend	Suggest actions or priorities	Human chooses what happens
3. Draft	Create near-finished outputs	Human reviews and approves
4. Execute with approval	Prepare or perform actions after authorisation	Human controls important execution points
5. Autonomous execution	Complete defined work independently	Human monitors outcomes and exceptions

Use the lowest level of autonomy that still creates the business benefit you need.

Low-risk work

AI can often take a larger role

Consider generating internal headline ideas.

If one suggestion is weak, very little harm occurs because a marketer can reject it before it reaches a customer.

AI generates 30 concepts → marketer selects 5 → marketer edits 2 → team tests the final versions.

This is an excellent human-AI workflow because AI provides scale while the marketer provides judgement.

Higher-risk work

Human control should increase with consequence

Now consider an AI system proposing that a company double a £50,000 monthly advertising budget.

The system may identify an opportunity, but the decision could affect:

Cash Flow

Can the business fund it?

Advertising economics do not exist separately from financial constraints.

Operations

Can demand be fulfilled?

More orders or leads may create pressure elsewhere in the business.

Strategy

Is growth desirable now?

The company may deliberately prioritise margin, capacity or another market instead.

In this case, AI should provide evidence and recommendations. A responsible human should make the final commercial decision.

Human-in-the-loop

Build approval into the workflow

A human-in-the-loop process deliberately includes a person at an important stage.

Workflow	Useful human checkpoint
AI drafts a paid ad	Approve claims, offer, tone and compliance before publishing
AI analyses campaign performance	Validate data and reasoning before changing budget
AI creates an image	Review brand fit, accuracy and rights issues before use
AI classifies leads	Audit accuracy and monitor high-value or unusual cases
AI drafts customer responses	Escalate complaints, sensitive issues or unusual requests to a person

Human-on-the-loop

Sometimes monitoring is enough

Once a low-risk workflow is reliable, a person may not need to approve every individual action.

Instead, the human monitors:

Exceptions

Unusual cases

The system escalates activity outside the normal rules.

Quality

Sample reviews

A person periodically checks output to make sure standards remain acceptable.

Performance

Outcome monitoring

The team checks whether the automated process continues to achieve the intended result.

Where humans should remain firmly in control

Keep accountability close to consequential decisions

Decision	Why human control matters
Major marketing budget changes	Requires commercial, financial and operational judgement
Legal or regulated claims	Incorrect output can create serious business risk
Brand positioning	Needs strategic understanding and organisational commitment
Sensitive customer situations	Empathy, context and judgement may matter more than efficiency
Use of personal or confidential data	Privacy, security and governance requirements must be respected
Ethically contentious activity	A business needs accountable human judgement about what it should do

Avoid automation bias

AI recommendations can feel more objective than they really are

Automation bias occurs when people give excessive weight to a recommendation because it came from a system.

For example:

“The AI says Campaign B should be paused, so we should pause it.”

A better response is:

“What evidence led to that recommendation? Is anything missing? What would happen commercially if we acted?”

AI should make human judgement better informed, not less active.

Interactive autonomy checker

How much control should AI have?

Impact if wrong

Easy for a human to review?

Task rules

Quick challenge

AI recommends a major budget increase based on improving campaign results. What should happen?

Knowledge check

Check your understanding

1. What is a strong role for AI in higher-impact marketing decisions?

Provide analysis, alternatives and recommendations while an accountable human makes the final decision Make the decision automatically because AI processes more data

2. What is a human-in-the-loop workflow?

A workflow with a deliberate human review or approval stage A workflow where AI cannot be used at all

3. When can greater AI autonomy make sense?

When the task is clear, repeatable, low-risk and outcomes can be monitored Whenever AI is technically capable of performing the task

4. What is automation bias?

Giving excessive weight to a system recommendation simply because it came from an automated or AI system A technical problem that makes automation run slowly

Key takeaway