By the end of 2019, worldwide spending on AI systems is forecast to reach $35.8bn – a whopping increase of 44% on the previous year. And marketing and sales teams are the most likely to benefit from the budding technology.
But despite the promise of AI and machine learning, there’s still significant barriers for entry. Demand for AI talent has more than doubled in the last two years, with two roles available for every AI professional today, while more than half (51%) of large enterprises admit to not fully understanding the data infrastructure needed to implement AI at scale.
Caitlin: What does your role at Growth Tribe entail?
Bernardo: I build content and deliver training in analytics and machine learning.
Caitlin: Can you tell us a bit about Growth Tribe and the work you’re doing to enhance AI literacy?
Bernardo: Growth Tribe is an educational company that‘s trying to enhance digital skills and literacy in the workplace, and we have two key audiences.
First we have the young talent. This is a group who are looking to acquire skills faster, so their training lasts for 6-12 months. From this pool, companies can hire new talent or build their existing workforce’s skills in areas such as growth and AI. In these courses we teach digital experimentation and machine learning tips to improve customer journeys, for example helping them understand patterns in their users’ behaviour to improve personas.
“We teach digital experimentation and machine learning tips to improve customer journeys, for example helping them understand patterns in their users’ behaviour to improve personas.”
We also have a data science training system for the young talent, where bigger companies can hire from the talent pool or enroll their current workforce to gain skills to become what we call a “Type A Data Scientist”.
Caitlin: What’s a “Type A Data Scientist”?
Bernardo: This is someone who understands business proposition and can effectively communicate the results of the analysis. They are different from “Type B Data Scientists” who are normally the builders and put the results of the analysis into production, but don’t necessarily know how to communicate the results to the rest of the company. This lack of communication very frequently leads to frustration and what we call ‘the last mile’ problem, which is when many of the projects are never deployed or utilised again once they are produced.
Caitlin: Tell us about the second audience?
Bernardo: Growth Tribe mostly deals with upscaling experienced professionals, mainly from marketing backgrounds. This takes the form of crash courses that last for two days – or six evening classes. These educate them how to optimise metrics for growth and adopt a marketing funnel approach.
Caitlin: So, what advice would you give to companies just starting out using AI?
Bernardo: Educate yourself on how to evaluate the effectiveness and value of AI.
We always coach decision-makers to first understand what the tech can do and what business problems it can currently solve. Very frequently, people mistakenly think machine learning can be used for every task, but we have to demonstrate to them that most of the time AI is for automating the predictable, repetitive tasks.
“People mistakenly think machine learning can be used for every task, but most of the time AI is for automating the predictable, repetitive tasks.”
Decision-makers need to understand this intersection between what’s possible and what’s necessary to be solved – what we call the ‘feasible projects’ – and this decision-maker needs to find which are profitable and ethical, making that intersection even smaller.
Caitlin: ‘Fear of AI’ is obviously common, with many people – including business leaders – finding the technology intimidating. What sorts of concerns do clients have when they come to you hoping to utilise AI?
Bernardo: The ethical issues surrounding AI is a big one. To tackle this we try to demystify some of the aspects of the technology. For example, we discuss the interpretability of different algorithms: what a white box/black box algorithm is and what level of information each of them provides the decision-maker as well as the consumer. We outline which algorithms comply with the idea of both user and customer having the right to an explanation of how the tech they are using works.
Fairness is another topic that comes up. People are often concerned that the algorithms they want to use, or are using, are impartial because they’ve been trained with data previously labelled by humans, and has picked up their bias. Companies often come to us after they’ve gathered data that they realise is biased now that they are building models.
Many businesses and individuals also come to us because they feel their knowledge and skills are limited and they want to be more data-informed – to have more insights to understand personas better so they can improve customer journeys. They are looking to continue the optimisation process that started with digital experimentation, and machine learning is the top concern for them now.
Caitlin: How does becoming more data-informed improve customer relationships?
Bernardo: Becoming more data-informed means you spend resources more efficiently and you take less time to achieve the performance of an expert.
With the help of an AI tool like Datasine Connect, a novice marketer can become a mini-expert in channels like Facebook or Instagram much faster because they don’t have to test as many variations since the tool already suggests what will be engaging.
The same goes for any predictive model tool in marketing – using machine learning we can predict things like retention and user referrals, which can then be analysed to understand the causal effects. For example, if I observe in my statistical model that desktop users are converting more than mobile users then I can try to look at why this is happening and optimise the mobile customer experience.
Usually you get higher uplift if you have data-informed presumptions before you start experimenting.
Caitlin: What are you most excited for in the near-future of AI?
Bernardo: The ability to use and expand on computer vision and voice assistants.
In the last 10 years, the introduction of products such as mobile phones means that we’ve begun to process huge amounts of unstructured data in forms such as photos, video and audio files. This kind of data is the fuel for building, training and testing algorithms that can act as substitutes for a few tasks humans can do – for example the semantic content analysis that is behind Datasine Connect’s image scoring.
Voice assistants are just at the beginning of their evolution and will become far more advanced. In the near-future we will see products like this that will be far more based on vision and speech or text summaries while still profiting from the previous advancement of AI and algorithms.