Why AI Works Best with Humans: What 3,600 Consumers Can Teach Businesses About AI Collaboration

The discussion concerning the role of AI in the creative industries is usually presented in terms of replacement – namely, how much of the work can AI take over? How fast can AI carry it out? What are the implications for the people who are currently carrying out the work? 

But what if those are the wrong questions?  

In fact, new consumer research we carried out on AI-generated packaging suggests that the most commercially successful approach is neither entirely human nor delivered by AI alone – it is a combination of both. 

Global research 

Ours is a global packaging and design agency and our research involved 3,600 consumers in both the United States and Europe. We examined packaging graphics created in three ways: by human designers, by AI on its own, and through a combined approach in which designers used AI as part of the creative process before refining and developing the results. 

The findings offer an interesting insight into a much larger question that businesses are currently facing – where does AI actually add value and where does human judgement still make a measurable difference? 

In almost every important commercial respect, the hybrid method performed better than either of the other two approaches across all six consumer categories studied, proving that combining AI with human judgement can make brands’ strategies more effective and reliable.  

Perhaps even more revealing was the extent to which AI was able to do well on its own and the point at which it started to show shortcomings. 

AI is very good at getting attention 

A major strength of AI is its ability to generate ideas quickly. When given the right inputs, generative tools can produce impressive images, interesting visual combinations, and creative approaches that would otherwise take a great deal of time to discover through conventional methods. And consumers noticed. 

In the study, packaging designed entirely by AI scored highest for visual impact (5.21 out of 7), while hybrid design scored 5.09 and human-only work 4.32.  

So why is that significant? In the field of packaging, just as in advertising or digital experience, the first challenge is gaining attention. AI is able to quickly increase the number and variety of ideas that a creative team can examine so it produces something surprising enough to break out of ordinary visual patterns. This makes it a key opportunity for businesses competing for attention in a crowded marketplace. 

AI can be a very effective tool for divergent thinking when it is carefully incorporated into workflows, as it enables teams to go beyond obvious ideas while at the same time keeping strategic control. Although attention and value are not the same, understanding this difference can help your organisation make effective use of AI without compromising trust or credibility. 

Standout did not translate into trust 

Even though the design that relied only on AI performed better in terms of visual appeal, consumers gave a more negative reaction when it was evaluated on other characteristics. The hybrid packaging received an average appeal score of 5.36 out of 7, compared with 4.85 for the human-only design and 4.42 for the AI-only design. 

Purchase intent also showed the same trend, with hybrid concepts receiving a score of 3.78, as compared to 3.41 for human designs, and 3.12 for AI.  

And the difference was even more interesting for perceived quality and trust, since these factors matter for long-term brand loyalty and consumer confidence. The hybrid design scored 5.11 for premium quality and 5.08 for trust, while the AI-only concepts scored 4.33 and 4.12, respectively. Although AI was good at creating the visual interruption, businesses would do well to take note of the fact that consumers once again seemed to value the outcome more when human judgement was involved. 

The hidden human value 

Generative AI is extremely capable at producing output. But output is only one part of a professional creative process. Someone must still decide whether an idea is appropriate, credible, distinctive, strategically useful, and properly resolved, reinforcing the importance of human judgement in creative processes. Human judgement is often a hidden part of the creative process, but its value comes into focus clearly if it disappears. 

A common problem in discussions about automation is that they often concentrate on the most obvious aspect of a person’s job. For a designer, that could be creating the image itself. However, much of the value a skilled professional adds happens before and after that moment. Designers interpret the brief and grasp what the client is really aiming to achieve, including subtle brand values that AI cannot fully understand without human insight. Indeed, the most important input in a creative project may never even reach the formal brief. 

A client might make an apparently throwaway comment during a meeting – a concern, ambition, or observation that reveals something important about what the project needs to achieve. A good designer recognises that signal and incorporates it into the work. It is difficult to reduce that capability to prompting. AI can generate hundreds of options, but people still provide much of the context required to know which option matters. 

Consumers may be detecting the absence of judgement 

A particularly interesting feature of the study was that consumers were not told which packages were made by humans, by AI, or by a combination of the two. Nevertheless, they did react differently. This does not always mean consumers can consciously identify work produced by AI. Rather, it indicates that they may be responding to the effects of a less refined creative process. 

AI-generated images may have subtle inconsistencies, excessively smooth surfaces, a reliance on common visual conventions, or features that just don’t feel resolved. Although each detail on its own might seem unimportant, together they can affect how credible or intentional the image seems. 

The clarity data supports that possibility. Consumers correctly identified the product type and variant on 92% of the human-designed packs and 89% of the hybrid packs. For AI-only graphics, that figure fell to 77%. This points towards a wider lesson for businesses deploying AI for creative and branding projects. The risk is not always that customers will see something and think, ‘This was created by AI’. The greater risk may be that they simply experience the output as slightly less useful, trustworthy, or coherent.  

Efficiency versus effectiveness 

Efficiency needs to be measured alongside effectiveness in decisions about the balance between AI and human inputs. Businesses may, understandably, be tempted to focus primarily on AI’s advantages as a cost-reduction tool. And the efficiency opportunity is real, of course. Under the right circumstances, incorporating AI into elements of the design process can reduce time and development costs, including internal resource and external agency expenditure.  

But our research also suggests there are pitfalls in a strategy that involves optimising for the sake of cost reduction alone. 

Customers said they would be willing to pay about 7% more than the normal category price for items using hybrid packaging. The human-designed option commanded a 2.4% premium, while the AI-designed option was 1.8% below the customers’ usual category price.  

This is not proof that companies can simply raise prices by 7%! However, it does indicate that the way a task is creatively delivered can influence perceived value. If the human element is reduced too much, cost savings may be achieved earlier but there is a danger that commercial performance could suffer later down the line. When leaders are considering where to introduce AI, the more useful question is not merely ‘How much time can we cut?’ but rather ‘How much efficiency can we achieve without eliminating the judgement that generates value?’ 

Think augmentation before automation 

Our study’s most significant finding was that there is a quantifiable benefit to combining people and technology rather than choosing between the two. AI broadened the range of possible outcomes, while humans provided strategy, refinement, context, and judgement. Only when they worked together did consumers react most strongly. This approach also applies to areas beyond packaging. 

Marketing teams can use AI to explore campaign routes before experienced people decide which ideas deserve investment. Researchers can use it to interrogate large datasets before applying domain knowledge to the conclusions. Product teams can use it to prototype quickly while keeping humans responsible for customer context and trade-offs. Professional services firms can accelerate routine elements of delivery while preserving the judgement clients are paying for. 

The key organisational question is therefore not merely where AI can carry out a task, but where it can do so within a system that still retains the most important human capabilities. Businesses must design the workflow, not just purchase the tool.  

During the initial stage of adopting enterprise AI, a great deal of attention has been paid to access – that is, which models employees should use, which licences to buy and which tasks can be accelerated. 

The next stage must be more advanced; companies should determine where AI provides real advantages and where human involvement improves the quality of the outcome. This involves asking: 

• In what situation is speed more important than nuance? 

• Where does trust matter? 

• What kinds of decisions call for experience or for an understanding of the context? 

• In what situation does a human make an improvement rather than just approve an AI output? 

• Which sections of the workflow can be safely automated when human participation adds little value? 

Bespoke balance 

The answers to those questions will differ from one industry, function and use case to another. In our study, the responses varied by category and region. Although hybrid models performed best as a general rule, the extent of their advantage depended on what the consumers were purchasing and where they lived.  

There is probably not going to be a one-size-fits-all formula for the correct human-AI balance. The source of a competitive edge will be in intelligently designing that balance and some successful AI strategies may be deliberately human.  

Generative AI will keep on becoming faster, cheaper and more capable. But that does not mean that the most successful organisations will be those that eliminate the largest number of people from their processes. They might instead be those that become a great deal more precise about the areas in which humans add disproportionate value. 

Our study provides an early glimpse into what that might look like. While AI on its own produced novelty, it was humans that provided clarity and trust. Yet the most favourable commercial reaction was observed when AI and humans worked in collaboration.  

For companies that are rushing towards adopting AI, this difference is important. The aim ought not to be to automate everything that can be automated. Instead, the sweet spot is to design workflows in which machines take on the tasks they do best, while giving people more chance to use the judgement, context and creativity that make the final result worthwhile.