New technology often attracts attention through its capabilities.
A system can generate text, produce images, summarize documents, analyze information, or automate a sequence of tasks. These capabilities can be impressive, but their significance depends on the problems they solve and the conditions under which they operate.
Marketing is a particularly useful context for examining this distinction because it combines creative judgment, structured information, repeated processes, and decisions that affect business outcomes.
The Long Run's Technology & AI section explores how intelligent systems can change marketing work, where automation may create value, where limitations remain, and why better technology does not automatically create better decisions.
AUTOMATION DOES NOT FIX A BROKEN PROCESS
Making an inefficient workflow faster can preserve the inefficiency.
Automation is often introduced to reduce repetitive work, improve speed, or remove manual steps.
Those are reasonable objectives. But automation works within the structure of the process it is given.
If responsibilities are unclear, inputs are unreliable, approval criteria are inconsistent, or the workflow contains unnecessary steps, automating the sequence may reproduce those problems at greater speed.
The first question should therefore be whether the process is well designed.
Understand the workflow before automating it
A marketing workflow may include research, briefing, creative production, review, approval, publishing, and performance evaluation.
Each stage has inputs, outputs, responsibilities, and possible failure points.
If the team repeatedly revises content because the brief is incomplete, automating content generation may increase the volume of work that requires revision.
If performance reporting is unreliable because metric definitions differ, automatically generating reports will not resolve the underlying inconsistency.
The process needs to be understood before the tool can be evaluated properly.
Identify the right role for automation
Some tasks are repetitive and governed by clear rules. Others require contextual judgment, interpretation, or approval.
The appropriate level of automation depends on the task's variability, risk, reversibility, and consequences of error.
A system may be suitable for preparing a draft or organizing information while a person remains responsible for verifying claims or approving an external communication.
The objective is not to automate everything. It is to determine where automation can reduce unnecessary effort without weakening reliability.
The importance of feedback
A workflow should be evaluated after implementation.
Does it reduce total effort? Does it improve consistency? Are errors becoming less frequent? Has the burden shifted to another stage? Can people understand and correct failures?
These questions matter more than whether the automation successfully executed its technical steps.
The larger point
Automation is valuable when it improves the complete process.
The goal is not simply to make a workflow run automatically. It is to make the workflow worth running.
AI CAN GENERATE THE OUTPUT. WHO DEFINES WHAT GOOD MEANS?
Generation is only one stage of creative and analytical work.
AI systems can produce material quickly, including drafts, visual concepts, summaries, and structured suggestions.
But a generated output does not automatically satisfy the requirements of the task.
It may be fluent but inaccurate, visually polished but strategically irrelevant, or technically complete while failing to communicate the intended idea.
The quality of the result depends partly on the system, but also on the quality of the inputs, the task definition, the evaluation criteria, and the review process.
The missing specification
A request to create a marketing campaign is broad.
A useful specification needs to define the audience, objective, offer, positioning, constraints, tone, available evidence, and intended outcome.
Without that context, a system may generate plausible material based on general patterns rather than the specific needs of the business.
The result may look finished while remaining poorly suited to its purpose.
Evaluation must be designed
If a team cannot explain what makes a good result, it will struggle to evaluate generated work consistently.
Creative quality may involve relevance, clarity, differentiation, brand fit, and the ability to communicate a specific idea. Research quality may involve source accuracy, coverage, reasoning, and the appropriate treatment of uncertainty.
These criteria differ by task.
AI-assisted work should therefore be evaluated against explicit requirements rather than accepted because it appears polished.
Human judgment becomes more specific
The role of human judgment does not have to mean manually producing every element.
It may involve defining the problem, selecting the evidence, establishing the criteria, resolving ambiguity, evaluating trade-offs, and accepting responsibility for the final decision.
The distribution of effort can change even when those responsibilities remain important.
The larger point
The ability to generate an answer does not eliminate the need to define the question or evaluate the result.
The value of AI depends on the quality of the system surrounding it.
THE MARKETING STACK IS NOT THE SAME AS THE MARKETING SYSTEM
A collection of tools does not automatically create a connected operation.
Marketing teams may use different products for research, analytics, advertising, content, publishing, collaboration, customer information, and reporting.
Each product may solve a real problem.
But the organization's ability to operate effectively depends on how information and decisions move between them.
A team may have excellent tools while still relying on manual copying, repeated data entry, fragmented approval processes, and separate interpretations of the same campaign.
Tools solve tasks. Systems connect decisions.
A tool can make one activity easier.
A system connects the activity to its context: the objective, the inputs, the people responsible, the next step, and the way the result will be evaluated.
This distinction becomes more important as the number of tools and workflows increases.
Without clear relationships between them, additional software may introduce more coordination work than it removes.
Integration is not only a technical question
Technical connections can transfer information between systems, but useful integration also depends on shared definitions, permissions, process design, and accountability.
If two systems interpret the same metric differently, moving the data between them does not make the definitions consistent.
If no one owns a workflow, connecting its stages may simply make the ambiguity more visible.
The larger point
A marketing system is defined by how the work fits together, not by how many tools it contains.
Technology should support a coherent operating model rather than become a substitute for one.
A RESPONSIBLE VIEW OF AI
AI capabilities and limitations vary by model, task, configuration, and implementation.
Marketing organizations should consider factual accuracy, privacy, data permissions, intellectual property, brand safety, review requirements, and the consequences of mistakes.
They should also distinguish between assistance, automation, and autonomous action.
A tool that drafts a campaign is not necessarily a tool that can safely approve, publish, and optimize it without supervision.
The degree of independence should match the reliability of the workflow and the risk of the action.
Claims about specific model performance, market adoption, or productivity improvements should be supported by identifiable evidence rather than broad assumptions.
Technology is most useful when it helps organizations think and operate more effectively.
The Journal should continue exploring how AI and automation affect creative production, workflow design, research, decision-making, and the responsibilities that remain with people.
Explore related topics: Perspectives · Strategy · Creativity · Research: Marketing Industry · Research: Research Methodology.