a client walks into an agency… with ai slop
A CLIENT WALKS INTO AN AGENCY… WITH AI SLOP
27/07/2026
Clients are increasingly approaching agencies with their own ideas and visualisations for packaging designs created using AI. Ever since the first project of this kind came our way, we have been wondering how we should approach such materials as professionals, what would be best for the client in these situations, and how to convince them of it. After working through several such cases, we developed an approach in which we neither reduce our role to simply recreating the visualisation presented to us nor disregard the client’s involvement and intentions. In simple terms, we analyse the material, identify its strengths, point out its limitations, and propose changes that can transform the initial concept into a professional packaging design.
We still believe that the best results are achieved when the process begins with a strong brief and the agency is given the space to analyse the problem and develop the right direction. However, we do not want to present these two models of collaboration as opposites. We simply want to show that, regardless of the starting point, we can guide the client through the process in a collaborative way, drawing on our experience. We believe that the development of generative technologies and the ability to use them consciously is one of the great privileges of our time. Clients using these tools to communicate their vision more effectively can provide a valuable starting point for further work. The most important factor, however, is mutual trust. When the client remains open to our assessment and experience, we can develop the idea together and turn it into an engaging and effective creative direction.
Having already completed several projects based on materials generated by clients, we would like to share our most important observations.
Evaluating the Concept Through the Lens of the Brand
Sometimes, the brief includes designs generated quickly and intended primarily as inspiration. At other times, a client presents a concept that, as they emphasise, they have spent a considerable amount of time developing with AI.
In the first case, the material can be extremely useful because it helps us understand the client’s stylistic preferences: the mood, colour palette, personality, and visual direction they are drawn to. We do not treat it as a finished solution, but as a starting point. Based on it, we can propose our own interpretation, one that is better aligned with the brand strategy, tone of voice, target audience, product category, and market realities.
The second case is more complex. Even if a brand analysis and strategy prepared with the best available large language models are accurate and comprehensive, this does not automatically mean that they will be translated into the right visual design. When such an analysis is condensed into a prompt for an image-generation model, some important assumptions may be oversimplified, interpreted too literally, or omitted altogether. A side note: Of course, we understand that this depends on the quality of the tool, the user’s level of control over the output, and their technical expertise. That is a subject for another article. Here, we are considering the use of widely available tools by people without advanced programming or design knowledge. Along the way, important factors may be lost, including the hierarchy of information, differentiation from competitors, shelf visibility, the flexibility of the design system, the characteristics of the target audience, and the technological limitations of the packaging. And that is only the tip of the iceberg.
For this reason, even a multi-stage and seemingly well-developed AI-generated concept should be treated as material for evaluation rather than as a direct visual representation of the strategy.
Maintaining Distance from Your Own Concept
Working in an agency with a clearly defined structure means confronting a project with the opinions of many experienced industry professionals with different specialisms, perspectives, ages, and levels of seniority. As a result, even a designer who is strongly attached to their own concept can gain some distance and better understand which elements are working and which need improvement.
As generative tools have become more accessible, people without design experience have also gained the ability to create visual concepts. In some cases, this can lead to a strong emotional attachment to their “own” creation, making it more difficult to evaluate objectively. That does not mean such a concept is inherently misguided. It simply requires the same critical analysis as any other design. The agency’s role is to create a safe space for this evaluation, one that does not diminish the client’s effort but also does not compromise professional standards. In the age of generative artificial intelligence, the role of the critic — someone capable of evaluating an output objectively — has become more important than ever.
Creativity
Generative models are trained on vast datasets of existing images and descriptions. Their outputs are therefore based on patterns and relationships already present in that data. In the context of packaging, this often results in references to visual styles and compositions that are already well represented in the market. If certain solutions dominate a category, traces of them are likely to appear repeatedly in generated proposals. The model creates new combinations of learned patterns, but market originality, brand distinctiveness, and the potential to set new trends are not automatically built into the generation process. Addressing these factors requires conscious selection, competitor analysis, and strategic decisions made outside the model itself. The further a project needs to move away from established category conventions, the more important the designer’s experience, iterative experimentation, and ability to develop the generated material become. Using AI alone, it is easier to create an attractive interpretation of an existing trend than to genuinely establish a new one.
Consumers Can Recognise AI Slop
As the volume of AI-generated material grows, consumers are becoming tired of visuals that are idealised in a very specific and recognisable way. Unnatural lighting, overly smooth textures, and errors in small details can create a sense of distance and even reduce trust in a brand.
This is particularly important in packaging. If a generated image appears attractive at first glance but begins to feel artificial after closer inspection, the result may be the opposite of what was intended.
That is why we place great importance on controlling realism. It is not that we avoid generative models ourselves. What matters to us is that the final result resembles a carefully produced product photograph from Adobe Stock, rather than generic AI slop. Our designers ensure that every image meets our standards. We check proportions, textures, lighting, shadows, the presentation of the product, and the image’s consistency with its real appearance. Sometimes this means retouching generated material. In other cases, it means combining it with photography or a 3D render.
Information Hierarchy Cannot Be Added at the End
We sometimes receive concepts with a very interesting composition but no… logo because “it can be added later”. This is only one example, but the same issue applies to all important messages, USPs, icons, and legally required information for which there may no longer be space within the AI-generated composition. Once all the required elements have been squeezed into the layout, the design may no longer look as fresh or original as it did initially. This is why these elements need to be considered from the very beginning. A separate issue is the legally required minimum size of text and symbols. Verifying this requires access to the dimensions of the packaging dieline and the appropriate professional software.
Can the Design Actually Be Produced?
Creating an image without understanding printing technologies, packaging construction, and the limitations of a particular material can result in a design that is impossible to manufacture. Such concepts may include colour transitions that cannot be reproduced between CMYK and Pantone, graphics placed over seals, or text located within restricted areas. These solutions may look good on screen but cannot always be reproduced faithfully in print.
At FYNK, our DTP specialists assess whether a design can be produced using the planned technology and within the available budget. We evaluate the feasibility of using Pantone colours, white underprinting, varnishes, foils, embossing, and other finishing techniques.
In addition, a generated visualisation usually presents the packaging as a flat, perfectly displayed object. In reality, the design must be applied to a specific dieline, taking into account seals, folds, flaps, cuts, openings, valves, and technical areas.
An element that appears to be perfectly positioned in a visualisation may end up on a seal, fold, or edge once transferred to the dieline. An illustration may be cropped, while the most important communication element may be shifted away from the main display panel. We also check whether the system will work across different formats. A solution created for one type of packaging cannot always be transferred easily to another variant. A well-designed packaging system should be responsive: it should retain its personality, hierarchy, and recognisability regardless of scale or construction.
AI-generated materials can therefore provide a valuable starting point, but only professional analysis can reveal their true potential. The agency’s role is neither to recreate the client’s vision uncritically nor to reject it. Our role is to develop it with consideration for brand strategy, market realities, and production capabilities. The best results emerge when the client’s creativity meets the experience of designers, strategists, and DTP specialists. AI can accelerate communication and open up new creative directions, but it is conscious design decision-making that gives a concept coherence and effectiveness. At FYNK, we do not see technology as a competitor to the design process. We see it as a tool that, when used appropriately, can support its individual stages.
We still believe that the best results are achieved when the process begins with a strong brief and the agency is given the space to analyse the problem and develop the right direction. However, we do not want to present these two models of collaboration as opposites. We simply want to show that, regardless of the starting point, we can guide the client through the process in a collaborative way, drawing on our experience. We believe that the development of generative technologies and the ability to use them consciously is one of the great privileges of our time. Clients using these tools to communicate their vision more effectively can provide a valuable starting point for further work. The most important factor, however, is mutual trust. When the client remains open to our assessment and experience, we can develop the idea together and turn it into an engaging and effective creative direction.
Having already completed several projects based on materials generated by clients, we would like to share our most important observations.
Evaluating the Concept Through the Lens of the Brand
Sometimes, the brief includes designs generated quickly and intended primarily as inspiration. At other times, a client presents a concept that, as they emphasise, they have spent a considerable amount of time developing with AI.
In the first case, the material can be extremely useful because it helps us understand the client’s stylistic preferences: the mood, colour palette, personality, and visual direction they are drawn to. We do not treat it as a finished solution, but as a starting point. Based on it, we can propose our own interpretation, one that is better aligned with the brand strategy, tone of voice, target audience, product category, and market realities.
The second case is more complex. Even if a brand analysis and strategy prepared with the best available large language models are accurate and comprehensive, this does not automatically mean that they will be translated into the right visual design. When such an analysis is condensed into a prompt for an image-generation model, some important assumptions may be oversimplified, interpreted too literally, or omitted altogether. A side note: Of course, we understand that this depends on the quality of the tool, the user’s level of control over the output, and their technical expertise. That is a subject for another article. Here, we are considering the use of widely available tools by people without advanced programming or design knowledge. Along the way, important factors may be lost, including the hierarchy of information, differentiation from competitors, shelf visibility, the flexibility of the design system, the characteristics of the target audience, and the technological limitations of the packaging. And that is only the tip of the iceberg.
For this reason, even a multi-stage and seemingly well-developed AI-generated concept should be treated as material for evaluation rather than as a direct visual representation of the strategy.
Maintaining Distance from Your Own Concept
Working in an agency with a clearly defined structure means confronting a project with the opinions of many experienced industry professionals with different specialisms, perspectives, ages, and levels of seniority. As a result, even a designer who is strongly attached to their own concept can gain some distance and better understand which elements are working and which need improvement.
As generative tools have become more accessible, people without design experience have also gained the ability to create visual concepts. In some cases, this can lead to a strong emotional attachment to their “own” creation, making it more difficult to evaluate objectively. That does not mean such a concept is inherently misguided. It simply requires the same critical analysis as any other design. The agency’s role is to create a safe space for this evaluation, one that does not diminish the client’s effort but also does not compromise professional standards. In the age of generative artificial intelligence, the role of the critic — someone capable of evaluating an output objectively — has become more important than ever.
Creativity
Generative models are trained on vast datasets of existing images and descriptions. Their outputs are therefore based on patterns and relationships already present in that data. In the context of packaging, this often results in references to visual styles and compositions that are already well represented in the market. If certain solutions dominate a category, traces of them are likely to appear repeatedly in generated proposals. The model creates new combinations of learned patterns, but market originality, brand distinctiveness, and the potential to set new trends are not automatically built into the generation process. Addressing these factors requires conscious selection, competitor analysis, and strategic decisions made outside the model itself. The further a project needs to move away from established category conventions, the more important the designer’s experience, iterative experimentation, and ability to develop the generated material become. Using AI alone, it is easier to create an attractive interpretation of an existing trend than to genuinely establish a new one.
Consumers Can Recognise AI Slop
As the volume of AI-generated material grows, consumers are becoming tired of visuals that are idealised in a very specific and recognisable way. Unnatural lighting, overly smooth textures, and errors in small details can create a sense of distance and even reduce trust in a brand.
This is particularly important in packaging. If a generated image appears attractive at first glance but begins to feel artificial after closer inspection, the result may be the opposite of what was intended.
That is why we place great importance on controlling realism. It is not that we avoid generative models ourselves. What matters to us is that the final result resembles a carefully produced product photograph from Adobe Stock, rather than generic AI slop. Our designers ensure that every image meets our standards. We check proportions, textures, lighting, shadows, the presentation of the product, and the image’s consistency with its real appearance. Sometimes this means retouching generated material. In other cases, it means combining it with photography or a 3D render.
Information Hierarchy Cannot Be Added at the End
We sometimes receive concepts with a very interesting composition but no… logo because “it can be added later”. This is only one example, but the same issue applies to all important messages, USPs, icons, and legally required information for which there may no longer be space within the AI-generated composition. Once all the required elements have been squeezed into the layout, the design may no longer look as fresh or original as it did initially. This is why these elements need to be considered from the very beginning. A separate issue is the legally required minimum size of text and symbols. Verifying this requires access to the dimensions of the packaging dieline and the appropriate professional software.
Can the Design Actually Be Produced?
Creating an image without understanding printing technologies, packaging construction, and the limitations of a particular material can result in a design that is impossible to manufacture. Such concepts may include colour transitions that cannot be reproduced between CMYK and Pantone, graphics placed over seals, or text located within restricted areas. These solutions may look good on screen but cannot always be reproduced faithfully in print.
At FYNK, our DTP specialists assess whether a design can be produced using the planned technology and within the available budget. We evaluate the feasibility of using Pantone colours, white underprinting, varnishes, foils, embossing, and other finishing techniques.
In addition, a generated visualisation usually presents the packaging as a flat, perfectly displayed object. In reality, the design must be applied to a specific dieline, taking into account seals, folds, flaps, cuts, openings, valves, and technical areas.
An element that appears to be perfectly positioned in a visualisation may end up on a seal, fold, or edge once transferred to the dieline. An illustration may be cropped, while the most important communication element may be shifted away from the main display panel. We also check whether the system will work across different formats. A solution created for one type of packaging cannot always be transferred easily to another variant. A well-designed packaging system should be responsive: it should retain its personality, hierarchy, and recognisability regardless of scale or construction.
AI-generated materials can therefore provide a valuable starting point, but only professional analysis can reveal their true potential. The agency’s role is neither to recreate the client’s vision uncritically nor to reject it. Our role is to develop it with consideration for brand strategy, market realities, and production capabilities. The best results emerge when the client’s creativity meets the experience of designers, strategists, and DTP specialists. AI can accelerate communication and open up new creative directions, but it is conscious design decision-making that gives a concept coherence and effectiveness. At FYNK, we do not see technology as a competitor to the design process. We see it as a tool that, when used appropriately, can support its individual stages.