The last few years have made ad industry creatives very, very nervous about their jobs. You might think that it’s the contracting economy or the rising cost of living that has people concerned, or the swathes of firings and redundancies as agencies and their clients dramatically reduce the bloat as a decade of low-interest investment finally comes to an end, but you would be wrong,
There is one topic of conversation dominating the WhatsApp groups and member’s clubs populated by creatives at every agency and in every marketing department the world over: “Am I about to be replaced by a robot?”
The answer, probably, is no. ChatGPT or Stable Diffusion are not coming to replace advertising creatives, however much Sam Altman or your CEO would like them to.
But what about your boss?
Algorithmic management, The use of AI and ML implementations to partially or completely execute workforce management functions, is becoming more and more common within creative and cognitive workplace settings, accelerated by the hybrid working commonplace in the wake of Covid.
Hiding among these debates about worker productivity and technology lie fundamentally ontological questions about the nature of creative work in contemporary capitalism. When we understand the emergence of generative AI in the context of broader theories of post-Fordist cognitive labour and managerialism (as well as the implementation of other forms of AI in the post-COVID ‘hybrid-work’ landscape), we find that the logical endpoint of these technologies is not the replacement of ‘creative’ workers, but rather a continuation and deepening of the bureaucratic norms and practices that have come to define ‘creative’ work while we weren’t looking. In other words, Robots are not coming to take our jobs because we are operating under a false assumption about what our jobs actually are.
Algorithmic Management?
Whilst the term algorithmic management
Algorithmic management is an umbrella term that was first used to describe the managerial aspects of gig economy platforms such as Deliveroo or Uber; it's the sorting system that gets your driver going to the right place. It has since expanded to include any set of programmed algorithms, in any workplace setting, that autonomously makes decisions to partially or completely execute workforce management functions. This could be hiring and firing, coordinating work, or monitoring performance; for the moment, let's focus on the particular area of productivity analytics and optimisation.
Take a service like AktivTrak. The platform tracks the digital working lives of cognitive workers in an organisation via their work devices, using machine learning to analyse application usage and web habits, reporting to managers and to the users themselves how their time is spent and how to optimise.
Software that merely tracks this sort of productivity in the creative setting is trivially simple and has been used for decades. One can imagine a platform that measures time between a task being assigned , since computers are unable to automatically assess a creative task as finished or good, human input is necessary to let the system know that the task is completed, or for a manager to score or assess the output. In this system, users would be able to see how long they took to complete a task, relational information (eg, time vs quality) or identify the most and least productive work periods.
However, software like AktivTrak attempts to go beyond this to analyse precisely what happens between the task being assigned and the task being completed, and optimise that productivity directly through user interactions.
AktivTrak proclaims that “Modern work needs analytics, not surveillance”; However, the divide between productivity analytics and employee surveillance, especially in a cognitive labour setting, seems narrow.
AktivTrak intermittently captures screenshots of employee computers, as well as automatically building a list of apps and sites employees use and classifying them into productivity categories according to AI-given algorithms. This makes sense in a work setting with clearly defined, often menial tasks; using Facebook when you should be inputting data into Excel, for instance, is clearly a non-productive use of time. However, this is profoundly unhelpful in a creative setting, or indeed any setting where knowledge work might require research, innovation, or inspiration.
Excessive amounts of time spent watching YouTube videos or exploring the social web would be categorised as unproductive by AktivTrak’s algorithm, yet these are spaces that frequently yield inspiration and ideas for the advertising creative. Inspiration can come from anywhere; there is no meaningful way to categorise apps or sites as more or less productive in the ideation process, especially since creatives will typically work across multiple industries and categories. A Mumsnet discussion of boot space in cars might be useful on Monday, but has no bearing or relevance for Tuesday’s FMCG brief.
When it comes to the analytics element of the software, it stands to reason that, if AI cannot differentiate between productive or unproductive app or website usage in this context, it cannot meaningfully report on creative productivity, at least beyond unnecessary surveillance or the simplistic tracking software described earlier.
The personal ideation processes of individual creatives are incredibly varied; two similarly productive workers might favour entirely different personal approaches, from reading books offline to scouring the internet for inspiration to simply sitting and thinking. There is no one-size-fits-all approach to generating ideas, yet AM implementations by their nature attempt to standardise the un-standardisable.
Ultimately, creative ideation is a cognitive process that happens in the head of the creative, often away from a computer. Software isn’t psychic; it has no meaningful way to improve the productivity of a task that happens entirely inside a worker's head.
Clearly, this is not the whole story. The life of an agency creative is not wholly spent generating ideas. While AI management tools like AktivTrak are bad at tracking cognitive tasks, they are fantastic at tracking the scutwork that have come to define the creatives’ work day.
Performative Productivity
As any creative, and indeed any cognitive worker, can attest, it is not enough to just work; one must be seen to be working:
“One of the reasons that the concept of “cognitive work” is so unsatisfactory is that thinking is the last thing one is permitted to do at work now. Work only counts as work if you can be seen doing it, and if it is quantifiable: so answering emails feels like real work whereas “just” thinking doesn’t.”
Mark Fisher
Creative agency work, and cognitive work generally, is an endless stream of emails, Slack chats and Zoom meetings that rarely contribute to the fulfilment of creative tasks. With the workday taken up by busywork and the mind-numbing drudgery of menial tasks, cognitive labour is relegated to our time outside the office, blurring the lines between work and leisure. These boundaries are further destroyed by the ubiquity of what Fisher calls “management by iPhone.”
Our work, and therefore our productivity, is mediated through our devices; the office WhatsApp group buzzing with messages last thing at night, or the ‘don’t worry about this until Monday’ email you receive on a Saturday morning. To cope, you are obliged to play along, to perform productivity in an effort to keep your managers at bay. We’ve all intentionally left the sending of an email to later in the evening as a way to demonstrate just how hard and how late we are working, or loudly discussed with colleagues the work we did at the weekend in earshot of our managers. Cal Newport refers to this as “Productivity Theatre.”
The Hybrid working model that followed Covid initially offered respite from this pressure; workers spent less time on pointless tasks and therefore had more time to think during the workday. This resulted in a subjective sense of a better work/life balance, and cognitive workers overwhelmingly report feeling more productive when working from home. There is likely some truth to this; It has been argued that a recent 7.5% decrease in US worker productivity directly correlates with the return of the cognitive labour workforce to the office.
However, managers, now no longer able to observe employee performative directly, were afflicted with what Microsoft dubbed productivity paranoia. A recent study found that 87% of knowledge workers felt they were productive yet only 12% of managers have confidence their teams are performing productively; the boom in AM systems and their usage within cognitive work settings is a direct response to this paranoia.
Creative Work as Communicative Capitalist Realism.
Ultimately, AM applications in cognitive and creative work do little beyond bringing the performative productivity of the office into the home of the worker. Apps like AktivTrak cannot track creative productivity, but they can easily show the degree to which workers are using email or Slack channels, making phone calls, or any of the hundreds of different ways that productivity theatre is expressed. It is the production of signs, not of work, an expression of societies of control as identified by Deleuze. This is what Franco Beradi calls semiocapitalism.
Beradi maps the Fordist/post-Fordist paradigms to the physical body and what he refers to as soul; our thoughts, our feelings and emotions, our creativity, affect, and libidinal investments. If Fordism was the spatio-temporal confinement of the body to the soulless repetition of the assembly line, the semiocapitalism born from post-Fordist embrace of datafication and digital technology “takes the mind, language, and creativity as its tools for the production of value.” For Beradi, as with Baudrillard, Semiocapitalism functions as an economy of signs, minus the signified. Layers upon layers of symbols that pile on top of each other.
Fisher builds on the work of Beradi and Deleuze to identify communicative capitalist realism, the idea that the future is to be fundamentally shaped by digital communicative technology and that there is no alternative:
“The drive to assess the performance of workers and to measure forms of labour which, by their nature, are resistant to quantification, has inevitably required additional layers of management and bureaucracy. What we have is not a direct comparison of workers' performance or output, but a comparison between the audited representation of that performance and output. Inevitably, a short-circuiting occurs, and work becomes geared towards the generation and massaging of representations rather than to the social goals of the work itself."
Algorithmic management in agencies should therefore be understood as a continuation of these bureaucratic anti-production norms and practices.
Postscript or The Future Is Now, Old Man
The latest innovation within artificial intelligence is Generative AI. algorithms such as ChatGPT3 and Stable Diffusion can draw on vast stores of data to autonomously synthesise original and novel images, texts, audio, and other types of digital media.
The speed with which these technologies have found their way into creative agency workflows is astonishing; OpenAI’s beta playground has only been available for a matter of weeks, but I doubt there is an agency in the UK that hasn’t at least experimented with ideas generated by ChatGPT-3. In my experience, with the right prompt, ChatGPT3’s ideas are indistinguishable from the work of a human creative, but where the latter takes days, the former is finished in a matter of seconds.
The mid-20th century was a period of dramatic technological discovery; so much so that it lead Frederic Jameson to imagine that computers, robots, new energy sources, and new information technologies would replace industrial labour completely, heralding the “end of work”. Frederic Jameson’s utopian vision of the future designated this presumed technological phase of capitalism “postmodernism.” (Graeber, 2014)
“Where are the flying cars… and all the other technological wonders any child growing up in the mid-to-late twentieth century assumed would exist by now? “We haven’t moved even computing to the point of progress that people in the fifties expected we’d have reached by now. We don’t have computers we can have an interesting conversation with, or robots that can walk our dogs or take our clothes to the Laundromat.”
David Graeber, 2014
Ten years later, one need only gesture towards today’s easily accessible generative AI applications; those wonders are right here, and we are using them daily. The oft-reported story of Google Engineer Blake Lemoine losing his job after becoming convinced that the AI he helped train was sentient and in possession of a soul, if nothing else, shows the extent to which the subjective experience of using these generative AI models is unnerving and beguiling in equal measure.
And yet, at work, rather than wonder, we experience dread. The impact of these technologies on cognitive work, even that performed by advertising creatives, is not difficult to discern. One need only look at the impact of AI on the translation industry to get a sense of the future; where once the craft of translation was reliant on in-depth knowledge of language and meaning, services like Google Translate now perform the bulk of the work instantly and for free, and translators are left to tidy up the output. They have moved from the role of knowledge producers to what Ben Thompson calls knowledge interrogators, editors whose primary role is to ensure accuracy in the final product, with very little of their hard-learned skills being used at all.
AI was supposed to be emancipatory; instead we have become prisoners of the algorithm. Rather than freeing us from the drudgery of menial tasks and allowing us the time and space to explore our creativity, AM intensifies the performative productivity that defines contemporary cognitive labour, while generative AI takes over the creative elements of our jobs that make the endless bureaucracy worth enduring.
Despite our fears, advertising creatives are not likely to lose our jobs to robots; What would be the point? Instead, our precariousness will ensure that we remain in our jobs, even as we spend our days checking and editing the recycled writing of a predictive model and touching up derivative art that we ourselves no longer get to produce.
