The Struggle of AI Marketing
Everyone wants a piece of the AI gravy train, and the marketing industry is no exception. Studies have shown that 80% of enterprise-level organisations have integrated AI in some form into their business, with 32% of these being marketing companies.
The concept of AI is an appealing one for marketers — identifying customers using their previous purchases to create a smart profile of an individual is something that has already been portrayed in science fiction. In Minority Report, Tom Cruise’s character has his iris scanned upon entering a Gap store, and then receives a personalised holographic sales pitch based on his sales history, where he is offered tailored suggestions of other products in stock.
The whole scene may have seemed very intrusive back in 2002, but today, is this concept so far fetched? Smart assistants such as Alexa are giving marketers a glimpse of the potential for consumer demand in this field, but unfortunately, there are currently key barriers in place that prevent the wider marketing industry in fully exploiting AI.
Insufficient Infrastructure
Any AI-powered marketing strategy requires a comprehensive technology infrastructure to support and power it. This is something many marketers are unprepared for, and often unable to produce. Machine learning algorithms are only able to derive useful insights through the processing of vast data streams, which requires substantial hardware and processing power. Securing the necessary budget for all of this can be challenging for SMEs. They also require the data to have been collected in the first place, which turns it into a chicken and egg situation.
As AI development continues, and the established AI platforms develop cloud-based offerings and AI-as-a-service, the need for physical infrastructure for marketers may decrease, with the rise in becoming a more tenable possibility.
AI Skills Shortage
The expense of buying and installing the IT infrastructure is only the beginning — what follows is frequent updates and maintenance, which also goes hand-in-hand with the requirement of well-trained support technicians.
Demand for technical roles in AI development is skyrocketing, resulting in a noticeable AI skills gap. AI experts are now in a strong position to negotiate large salaries, and often have the luxury of companies approaching them and becoming embroiled in bidding wars to secure them. Marketing departments’ budgets may not stretch to the large salaries being negotiated, or have the manpower to dedicate to participating in employee bidding wars between hiring firms.
Even if a company has opted for a bought AI solution over an in-house build requiring these expensive AI experts, there is still a need for training to use and deploy this technology, which must be factored in to the marketers’ budget.
The Data Issue
AI is built upon access to high quality, copious data streams. Despite being in the era of Big Data, there is still a large amount of confusion and disorganisation amongst marketers in how to effectively utilise the data streams open to them. In a 2016 study from IBM, 54% of marketers interviewed declared that their ability to act on insights derived from customer data was ‘poor’ or ‘very poor’.
Marketers may have access to data, but it’s not necessarily of the right sort. Machine learning models require data sets that are comprehensively curated and ‘cleaned’, which is a skillset that the average marketer may not have, and another indication of why data scientists are in such demand, with their role being named ‘Best Job in America’ in the 2019 Glassdoor roundup).
Acquiring data is a process that is subjected to a great deal more scrutiny now too, both legislatively and in the broader public consciousness. GDPR places larger restrictions and the prospect of severe penalties on companies who fall foul of it, while recent stories such as the Cambridge Analytica scandal in early 2018 have injected a new sense of data awareness into the general public. All of this makes the concept of casual data collection that is easily up to code for machine learning algorithmic processes far less attainable to casual participants.
The Evolutionary Nature of the Marketing Industry
Marketing is a profession that has rapidly evolved over the past twenty years. The rapid progression of the digital age, with the birth of online commerce and social media among other factors, has necessitated the formation of digital marketing, SEO and SEM. In an age where data collection is possible to construct AI platforms, there is no guarantee that the models implemented to train them will remain viable and functioning forever.
When IBM began to utilise its AI platform Watson to manage its programmatic marketing campaigns, they reported an average cost-per-click reduction of 35%, with some instances seeing drops of 71%. Watson’s ability to achieve this was through its analysis of customer data, including their browsing habits by time online, and device used.
However, as GDPR and related changes to data collection and utilisation have illustrated, consumer behaviour and access to data sets changes over time, which makes the prospect of AI platforms having access to reliable and consistently high quality, specific data streams difficult to guarantee.
Marketers need to factor in how their AI usage can evolve with their industry, to ensure ongoing efficacy.
