Driving Innovation by Activating a Company's History as "Cultural Data" with Generative Artificial Intelligence
Generative Artificial intelligence is here to stay. People are enthusiastically experimenting with it, exploring its creative potential. Now it is time to embrace a human-machine-interaction perspective that focuses on emerging practices of applied tooling. Rather than focusing on the limits and dangers of AI, we want to ask — So what? To what end should we actually employ this technology, and how could it augment human creativity? This shift in focus allows us to imagine what we could realize in collaboration between users and AI that neither a human or a machine could achieve by itself.
This article discusses the specific use case of harnessing a company's history as "cultural data" to proprietarily train a generative AI system for innovation purposes. We argue that this would enable a company to create novel designs that are coherent with a brand's identity and tradition. It would allow for storytelling beyond the established canon. We could analyze organizational culture at scale. And, potentially, such a system could be used to build organization-specific scenarios and anticipate the future. In other words, a company's history could be used via generative AI for novel product development, marketing and communications, organizational culture management, and strategy planning.
Altogether, this article gives a hint at what could emerge at the intersection of the past and the future when we start harnessing generative AI in meaningful ways. To get there, however, we must first understand how history is usually treated in a corporate setting, decode the workings of so-called "organizational cultures of remembrance," and unravel what the AI-assisted processing of cultural data requires in terms of work, expertise, and infrastructure.
An organization's history as an underused asset and cultural data
One unique and entirely underused asset that many future-focused companies tend to brush aside is their history. In order to understand how organizations could harness their history with generative AI, we need to discuss the phenomenon of so-called “organizational cultures of remembrance” first, as it poses the theoretical foundation for this paper (for an elaborate empirical account see my book, Mai, 2015).
History is a unique yet underused asset in companies
In an increasingly fast-paced economy, many business organizations idolize innovativeness and the ability to foresee changes in the marketplace as key qualities in the competitive struggle for garnering profit. While the present and the future appear to be forgeable and thus calculable factors, the past, in contrast, is commonly ascribed the image of an inalterable monolith no longer worthy of concern. Hence, cultural theorist Dirk Baecker (1987) once argued that the “social system of the economy” operates according to the “premise ‘bygones are bygones,’ systematically treating the past as irrelevant” (p. 519).
Despite the prevalence of future-orientation, the phenomenon of “corporate amnesia” Kransdorff (1998, p. 1) – i. e. the blatant disregard of past experiences – does not always take hold in a business environment. Some old brands, such as Mercedes-Benz, BMW, Audi, Leica, or Coca Cola have erected historical archives, corporate history museums, and “tradition departments” as keepers of corporate identity and heritage. The field of organizational memory studies has been exploring this phenomenon for several decades (see e.g. Decker, Hassard & Rowlinson, 2020). Aforementioned companies are “working the past” (Linde, 2009), thus bringing “various representations of various past events […] into the present to shape the future” (p. 222). After all, every company has a past, but only a few reconstruct themselves a history to make it usable for the present.
Carriers, purposes, and forms of organizational remembrance
These so-called “carriers of organizational remembrance” (Mai, 2015, p. 31) engage in history management not for the mere sake of retrospection. Instead, they pursue specific purposes that are of contemporary corporate interest. As my empirical exploration found, these “purposes of official organizational remembrance” include (p. 86):
Retaining physical proof of one’s existence
Legal protection of trademark rights
Historical accountability
Corporate identity construction in public relations
Brand identity construction in marketing
Direct economic utilization via spare parts, restoration, and merchandise
Satisfaction of entertainment demands
The breadth of these purposes demonstrates that organizational remembrance is, on an official level alone, much more than just knowledge management (e. g., Argote, 2013), the retention of organizational expertise and skills, or a marketing-oriented “dog and pony show,” as proclaimed by Keulen (2013, p. 115). History management and “tradition work” may fulfill several important functions which actively support a company and create additional value on several levels.
The mnemonic means through which a company pursues these business-oriented purposes are identified here as “cultural forms of organizational remembrance” (Mai, 2015, p. 30). They are constituted by the practices, media, physical sites, tools, and mental conceptions through which organizational remembrance takes place and becomes manifest in a corporate setting. The sheer variety of forms conveys that retrospection is not just a matter of archives and corporate history books offered on the periphery of business operations, but that it also occurs in various spheres of day-to-day corporate life.
Remembering the past on an informal level happens in every business organization
On an informal level, the workforce of a company may also embrace an elaborate set of cultural forms of remembrance. Employees tend to tell each other elaborate founding stories and nostalgic anecdotes about their past experiences and their conceptualization of a company’s history that transcend their personal tenure. They gather artifacts and memorabilia, write up work-related autobiographies, attend commemorative meetings of former colleagues, and so on.
In fact, the corporate past and its retrospection play an essential role in the construal of organizational reality and the understanding of corporate life among a workforce in most companies that have been around for several generations. References to distinct elements of the organization’s past help a workforce obtain a sense of how the company operates over time. As I found out in my research (Mai, 2015, p. 316), collective remembrance of the corporate past enables members of the workforce to:
Explain contemporary market conditions through select references to the corporate past
Trace the roots of supposedly common working practices in the past
Construe and legitimize the existence of contemporary conflicts inside the organization
Use reductionist images of the corporate past to evaluate the quality of contemporary corporate life and their own working conditions
Craft departmental and production site-specific collective identities through recollections of the corporate past
Achieve concreteness of identity as a greater corporate collective on a brand level
Anticipate how the organization may act in future terms of crisis based on previous experiences
In other words, the past is remembered in one form or another in virtually any business organization by a wide variety of actors for purposes of making sense of, and giving sense to, the present. This also means that the exact content of what is recalled – the individual elements of the past that are meaningful to the various internal stakeholders of an organization – may differ substantially, depending on the perspective and intended purpose of remembrance. For instance, the temporally distant elements a corporate history department decides to recall in their museum may not always matter to the workforce, while the stories floating around the workbenches may have not become canonized by corporate historians or marketers yet.
Due to its highly social and organically performative nature, I conceptualized the emergent system of meaning around carriers, forms, and purposes of retrospection as “organizational culture of remembrance” (Mai, 2015).
An organization’s history constitutes cultural data
Why does this theoretical excursion into organizational cultures of remembrance matter when talking about the creative potential of generative AI? The key lies in training an AI and customizing it to the liking of an organization to drive innovation in the domains of product development, communications, cultural management, and strategy planning. All the elements of the corporate past – i.e. the content of what is recalled – could be used as “cultural data.” This qualitative data is unique to every organization, and it creates the context necessary to catapult generative AI out of its generic creativity dilemma.
Cultural data, of course, is constituted by more than just an organization's history. The time frame does not necessarily need to be retrospective. Every contemporary trace of social interaction, nugget of communication, proof of group formation, working practices, media, work tools, organization charts, job descriptions, and onboarding handbooks, but also office design and corporate architecture could technically be rendered as "cultural data" if it enables us to paint a holistic picture of what is going on inside an organization from a human-centric perspective.
The reason why I am limiting this thought experiment to representations of an organization's past in the form of history is because temporal distance and a longitudinal view allows us to assess whether something is genuinely cultural or just social in nature. Building on aspects of Niklas Luhmann’s cultural theory (see Burkart & Runkel, 2004), phenomena are only considered cultural here if they are generated in social interaction, emerge in widespread patterns across organizational life, remain persistent over time, are relevant for collective identity construction (i.e. distinctions between "us" and "them"), assume a normative character that communicates values and behavioral expectations for organizational members, and if they are passed on across generations among the workforce.
What if one could harness that kind of cultural data, its carriers, purposes, and forms of organizational remembrance via generative AI to proactively shape its future in ways that have not been envisioned yet? What if a company was able to train their own generative AI based on their proprietary “cultural memory” of the organization, feeding it all its vintage designs, blueprints, photos, old marketing slogans, bygone creative campaigns, and reminiscent stories floating around the shop floor? The following sections discuss what exactly such an AI system could be good for and what it requires to turn organizational history and the elements of an organizational culture of remembrance into cultural data.
Potential outcomes of training generative AI with an organization's cultural data
The opportunities of training generative AI with the cultural data that is specific to an organization would be vast, ranging from the creation of non-obvious product ideas and designs, to marketing ideas that are creative and new, yet authentically grounded in corporate heritage.
In more abstract terms, generative AI systems have the potential to transform work in business organizations in five different ways. These ways build upon a human-tool-task interaction model developed by our own Paul Hartley (2017). This classification scheme allows for a more realistic assessment of the potential outcomes of training generative AI with an organization's cultural data.
How emerging technology transforms human labor
First of all, AI could be used as a novelty machine. It would allow its users to do something they could otherwise not do without the tool. While this is the dream and the hyped up grand idea that surrounds the popular AI discourse, this is a rare alteration, however. The vast majority of current applications of generative AI do not fall into this category.
Second, AI could be used as an extension machine. Users are able to expand their own abilities beyond what they could accomplish alone. Similar to a hammer, a calculator, or an automatic language translation engine, the tool serves as an adaptation of the user's body, knowledge, or information processing abilities. Tool and user collaborate to accomplish tasks, thus evolving the user-tool relationship and expanding the tooling possibilities. This is where the biggest potential of generative AI resides right now.
Third, AI assumes the role of a displacement machine. It displaces some of the user's capacities or responsibilities by assuming them itself in better, cheaper, more efficient ways. The user is either relieved of unwanted responsibilities, or must cede some of their own. While the latter case is the main worrying concern of many professionals who envision AI to have almighty powers, this kind of technological transformation impacts mostly unskilled labor. That's the human course of automation (and a curse for some) .
Fourth, AI can be seen as a transference machine. This means that the tool transfers capacities or responsibilities between users or other tools. The tool allows for the sharing or exchange of roles or responsibilities. It can often be a platform for the redistribution of user’s roles. AI thus works as a mediation platform.
And fifth, AI could become a combination machine. The tool combines, collapses, or eliminates roles or responsibilities. The tool allows users to accomplish tasks that might require multiple users, or to eliminate entire sub-tasks entirely, thus eliminating roles. The greatest potential of AI resides in this very role, but it is difficult to anticipate how exactly this will play out.
The potential outcomes of harnessing cultural data with generative AI that I will discuss in this paper mostly reside in the second category. AI systems may serve people inside a business organization as an extension machine. Given the current state of technological development, this is the most plausible role.
Novel designs coherent with a brand's identity
Every company is constantly on the lookout for imaginative designs and product ideas that are progressive but also on point with their brand. If a generative AI system was trained with a breadth of proprietary designs, blueprints, and vintage products owned by a company, its creative outputs would reverberate with these source materials.
To provide a practical example, when I conducted ethnographic research at AUDI AG (Mai, 2015), my fieldwork found that several senior Audi car designers frequently went into the archives, internal vintage car and materials depot, and corporate museum to gain inspiration for new car designs and material usage. They would then make references to historical models, some of which are more hidden than others. It was both an attempt to create something new that was in line with previous products and design languages, but it was also a respectful nod to the aesthetic expertise of previous generations of car designers. Paying homage is a creative cultural practice currently done by humans, using all sensory facilities. But it is also something that could be accelerated or turbocharged with generative AI.
While the generative AI system could combine, recycle, and reassemble the training data in novel ways, technically, the outputs would still be grounded in a company’s past due to its unique and proprietary origin of cultural data. Designers could then use these outputs for inspiration, employing generative AI as a tool that makes new connections no human has established yet. Groundedness in cultural data would also allow a company to authentically make claims of seamlessly combining tradition with innovation through technology, which would be in line with their identity.
A few people are currently experimenting with what emerges if one trained generative AI with representations of a company's past. For instance, Swedish design agency Space10 did so on behalf of IKEA (Roettgers, 2023, March 26). They fed a generative AI system a range of scanned clippings of vintage IKEA product catalogue images to generate retrospective mash-ups of future product ideas, such as chairs. Activating a company's cultural memory, however, needs to go deeper. It requires cultural data beyond product designs, and, most importantly, it requires continuous, curative work, as I will explain later.
Storytelling beyond the established canon
Most companies love telling stories about their founders, key inventors and decision-makers, milestone products and developments, and supposedly unique ways in which the organization works. Stories are fantastic vehicles for communicating meaning. They allow people to wrap a company's norms, working practices, with people, technology, place, and circumstance into a memorable, humanized package that bears potential for learning and identification. If several stories share a particular pattern, it's interpreted to be indicative of an organization's culture – i.e. what is desired and expected inside an organization across time and generations.
Generative AI could help both with identifying these patterns and establishing new ones. It could create meaningful links between previously unconnected stories, and it could discover and elevate stories from the fringes of the workforce to the corporate center. In other words, generative AI could retrace and map a network of stories and qualify the nodes by which they could be connected. Human actors could then take these newly identified narrative connections and employ them as sensemaking and sensegiving devices in marketing, external communications, internal communications, human resources, etc.
Analyzing organizational culture at scale
Many companies are already using AI to optimize business and production processes, make logistics more efficient, or find new compounds of materials. They are, first and foremost, feeding quantitative “business data”. Nobody is feeding qualitative “cultural data” yet, however. The analysis of organizational culture is a domain accessible to human experts in business anthropology, organizational sociology and psychology (see e.g. Martin, 2022, Schultz, 1995). Good exploratory studies – especially those embracing an ethnographic methodology – usually employ a mix of qualitative methods that involve participant observation, deep interviews, informal conversations, media and document analysis, etc. (see e.g. Brewer, 2006, Schwartzman, 1993). Quantitative analyses and "measurement" of organizational culture (e.g. Hofstede et al., 1990) are considered a positivistic joke in the organizational culture studies community, which is why they bear little potential for producing genuine cultural data.
We have absolutely no idea what a more evolved generative AI system could synthesize if it was fed a cornucopia of an organization's Slack chat protocols, intranet content, the entire email communication within a company, onboarding materials, or the offboarding protocols of retired employees. While this may sound like a dystopian corporate future in which an employee's every movement and utterance are monitored and analyzed, AI may uncover non-obvious behavioral patterns within the social system of an organization. For lack of a better description, this would resemble an automated version of qualitative data analysis software like ATLAS.ti or MAXQDA.
Naturally, these patterns will require subsequent interpretation and theoretical explanation by human experts to uncover the "why behind the what." The latter process is what turns a cluster of descriptive observations into a genuine human insight. Since so-called professionals in commercial research do not even get this right when they are untrained in the social sciences and humanities, we doubt that this skillfully interpretive and highly creative procedure will be machine-automated any time soon.
Scenario building and anticipating the future
Given the right kind of cultural data, paired with economic, political, and environmental forecasts, generative AI could assist in imagining scenarios how an organization might evolve and react as a social entity to external change factors in the future. This would allow a business to anticipate different futures how the world around them may change (e.g. what Peter Schwartz, 1996, has been promoting).
Within these worlds, generative AI could help play out different scenarios depending on variations in internal change factors as well. It would enable horizon scanning of the internal organization and not just the external world, the latter of which tends to be the case in foresight. Prompts could be used to direct the AI to find weak signals and combinations of elements in a signal that it would not have considered to look for itself.
World building and critical uncertainty combinations could potentially become coherent with the click of a button, yet reliably grounded in data. This would leave humans to narrate and imagine additional worlds for which the machine has no reference points. Generative AI could enable practitioners to make scenarios more tangible, visible, and potentially more impactful as a result.
AI could also advise on leverage points where the organization could proactively shape or influence the future, as outlined in the scenarios. And it could provide recommendations where to simply react. All of this would be highly speculative, but so is any kind of foresight.
What the AI-assisted processing of cultural data requires
The dictum of the digital economy holds true with generative AI: data can be a gold mine. But not all data automatically is. Cultural data involves work and cultivation. And an AI system needs to be set up in specific ways to be able to process it, which is why we have identified a range of human-centric design principles that should be embraced:
Proprietary, machine-readable training data
Cultural data, first and foremost, must come from inside the organization, which means that it is proprietary. It belongs to the organization. More importantly, data should manage to grasp and represent the company as a social entity in as many different facets as possible. This may include company records, development reports, communication protocols, blueprints, speeches, strategy papers, internal presentations, onboarding handbooks, internal memos, press releases, etc. – essentially anything a good business historian would consider a viable source.
Initial training data is not sufficient in the long run. Ideally, the AI system could be fed with a constant stream of cultural data in real time, since the organization itself will keep evolving. This would ensure that the system kept its digital fingers on the pulse of culture.
While some of these sources may already exist in digital format (e.g. Word documents, Slack protocols, PowerPoint presentations, Adobe Indesign files), others will need to be digitized first, especially when dealing with an organization that predates the digital age. This starts with an information asset register process, where an index of all the analog and digital data and databases available inside an organization is created. Established information asset register processes, however, usually only take into account traditional assets with business value (see e.g. Leming, 2015). A paradigm shift is required, broadening the organization's perspective to incorporate cultural data as a future source of value.
Next, cultural data will need to be brought into a machine learning system-readable format and language. It requires a different approach than what is currently used by archivists. People must anticipate how AI users may phrase a query, which requires linguistic expertise. And cultural data must become immediately retrievable.
Controllable external inputs
As in any creative process, inspiration should not only come from internal references. Innovation often emerges from the novel combination of disparate inputs stemming from different disciplines (Reckwitz, 2012). This is why any AI training also requires company-external input data. In order to determine the mix of company-internal and external inspiration, a variable slider could set the guardrails for the creative process.
Context and history-aware system
Any data is meaningless if it lacks a context in which it can be interpreted. By crafting a narrative thread in the form of corporate history, clusters of cultural data can be turned into a meaningful superstructure, which situates the organization in time, place, and community. It unravels the cultural context in which an organization operates as a social entity of human actors who collaborate and use tools to reach a common goal, embracing certain principles, values, taken-for-granted models of thought, and practices. That's why any good history is not just a chronological sequence of events, but it manages to interpret them and provide explanations of human action in light of the greater Zeitgeist.
This contextual understanding is essential for a generative AI system to create outputs that are neither generic nor random. More importantly, contextual understanding is an important first step to creating an artificial historical consciousness, which, on the other hand, is necessary to evaluate the nature of change over time, and anticipate the future in light of the past and the present. Equally important, corporate history renders an official version of the past that the organization holds "true" – however constructed this may be. The existence of a dominant version of the past is important because it may serve the generative AI system as a guiding structure and sensemaking device that allows for plausibility assessment of highly ambiguous cultural data.
Transparency of source & genuine accountability
To overcome the issue of opaqueness and blackboxing in current generative AI systems, any future system would need to be able to backtrack and reveal its original input sources that led to the output. This could be achieved in the form of inspiration backlogging. It would also allow users to manually fact-check, modify inputs, and fine-tune queries. Most importantly, it would enable corporations to claim genuine ownership of the newly generated ideas by tracing the creative curation process behind it, and become accountable for the outputs when exploiting them for commercialization.
Infrastructure and people professionally working the cultural data
For a company to be able to feed a gen AI system with “cultural data,” however, you don’t only need machine-readable content. You also need the infrastructure and organizational structures. While new tools, such as a generative AI system, augment the way people do their jobs, new formal structures allow them to actually do their job.
To start with, this requires a professional team that knows how to work the past in light of the future. They need to retrieve elements of the past, narrativize and activate them as cultural data. As mentioned above, they need to institutionalize certain forms of cultural remembrance in order to reconstruct the past in the form of a corporate history, which gives all that “cultural data” a context and establishes a binding truth as a superstructure.
Archivists, corporate historians, business anthropologists, and linguists then have to collaborate with data engineers, data scientists, and data architects to design cultural databases, and build predictive models. They also need to ensure data security, since this internal gold mine may render an organization extremely vulnerable.
A tooling perspective that reclaims human authorship & intellectual property
Contrary to the media hype, generative AI systems are neither autonomous agents nor persons at the moment. Technology still has not evolved that far. These systems do not have a sense of the consequences of their actions, and they do not follow any intent beyond producing new results that follow a query. They are neither proud when they succeed, nor do they get offended or self-conscious if they don't. In other words, generative AI is not sentient, however "self-learning" these machines may be proclaimed to be.
This is why it is important to remember that generative AI systems are tools built by humans according to human logic and capability that are used by humans to do human things. Hence, let's use them as such – as creative tools with the help of which people create something new they could not create on their own. This also means that human users and skillful practices of building, customizing, and employing these tools remain integral components of that equation. Generative AI outputs are not illusive phenomena that are magically conjured up somehow. They still require human interaction, which is why human authorship (e.g. wording queries) and intellectual property (based on e.g. cultural data used for training) will remain essential.
Conclusion
Contrary to popular belief, generative AI systems are neither world-ending nor revolutionary. They will not immediately replace hordes of creative professionals and knowledge workers across all industries. And they will not make human action, decision-making, and creative ingenuity superfluous.
Instead, generative AI systems offer a new toolset which can be used as an unexpected solution machine to existing and newly emerging human problems. More specifically, as I tried to demonstrate in this article, generative AI systems could be employed in the field of corporate innovation – for creative product development, communications and marketing, organizational culture management, and strategic planning, if one acknowledges and activates a company's history as "cultural data." The true value proposition would be the establishment of a machine learning system that adheres to the internal, cultural logics of an organization.
This endeavour not only requires a deep expert understanding of how organizational cultures of remembrance work and should be professionally managed in practice. It is also recommended that the AI-assisted processing of cultural data follows a set of human-centric design principles to ensure that the system acts as a tool serving the interests of the humans within an organization.
Cite this article:Mai, D. (2024, March 11). Driving Innovation by Activating a Company's History as "Cultural Data" with Generative Artificial Intelligence. Human Futures Studio. Retrieved from https://humanfutures.studio/articles/cultural-data-ai-innovation
References
Argote, L. (2013). Organizational memory. In Organizational learning: Creating, retaining and transferring knowledge (pp. 85–113). New York, NY: Springer Science + Business Media.
Baecker, D. (1987). Das Gedächtnis der Wirtschaft. In D. Baecker & N. Luhmann (Eds.), Theorie als Passion. Niklas Luhmann zum 60. Geburtstag (pp. 519–546). Frankfurt am Main: Suhrkamp.
Brewer, J. D. (2006). Ethnography. In C. Cassel & G. Symon (Eds.), Essential guide to qualitative methods in organizational research (pp. 312–322). London: Sage.
Burkart, G., & Runkel, G. (Eds.). (2004). Luhmann und die Kulturtheorie (1st ed.). Frankfurt am Main: Suhrkamp.
Decker, S., Hassard, J., Rowlinson, M. (2020). Rethinking history and memory in organization studies: the case for historiographical reflexivity. Human Relations, 74(8), 1123-1155
Hartley, P. (2017). Unfamiliar Sounds in Familiar Settings: On the Cosmopolitan Labour of Film Composers in Istanbul. Ph.D. Dissertation, University of Illinois Champaign-Urbana.
Hofstede, G., Neuijen, B., Daval Ohayv, D., & Sanders, G. (1990). Measuring organizational cultures: A qualitative and quantitative study across twenty cases. Administrative Science Quarterly, 35(2), 286-316.
Keulen, S. (2013). History as a dog and pony show? The (mis)uses of history for marketing by Wells Fargo & Company. In L. C. Neilson (Ed.), Varieties, Alternatives and Deviations in Marketing History: Proceedings of the 16th Biennial Conference on Historical Analysis and Research in Marketing (CHARM) (pp. 115–129). Copenhagen: CHARM Association.
Kransdorff, A. (1998). Corporate amnesia: Keeping know-how in the company. London: Butterworth-Heinemann.
Leming, R. (2015). Why is information the elephant asset? An answer to this question and a strategy for information asset management. Business Information Review, 32(4), 212–219.
Linde, C. (2009). Working the past: Narrative and institutional memory. New York, NY: Oxford University Press.
Mai, D. (2015). Organizational Cultures of Remembrance: Exploring the Relationships between Memory, Identity, and Image in an Automobile Company. Berlin/New York: De Gruyter.
Martin, J. (2002). Organizational culture: Mapping the terrain. Thousand Oaks, CA: Sage.
Reckwitz, A. (2012). Die Erfindung der Kreativität: Zum Prozess gesellschaftlicher Ästhetisierung. Berlin: Suhrkamp.
Roettgers, J. (2023, March 26). Ikea’s generative AI furniture designs are trippy, retro, and inspiring. Fast Company. Retrieved from https://www.fastcompany.com/90871133/ikea-generative-ai-furniture-design
Schultz, M. (1995). On studying organizational cultures: Diagnosis and understanding. Berlin: Walter de Gruyter.
Schwartzman, H. B. (1993). Ethnography in organizations. Newbury Park, CA: Sage.
