Why People Are The Core Of An AI Team

AI systems are machines. There’s no debate around the central proposition that Artificial Intelligence (AI) and its Machine Learning (ML) engine power stems from software-driven computing devices which are essentially machine brains driven by algorithmic logic.

But really, it’s the people in an AI team that make the difference. If an intelligence team isn’t made up of the right mix of data scientists, business logic specialists, lower function systems administrators & higher tier systems architects… and even commercial function representatives such as salespeople and project managers, plus of course software application developers, then it’s not going to be a good team.

It’s a people thing

At present, we’re seeing a two-speed approach to the enterprise adoption of AI, with some organizations racing to implement solutions in an effort to generate rapid Return on Investment (ROI). At the same time, others are taking a longer-term view, hoping to reap future benefits based on investments made now. Regardless of where an organization maybe on its AI journey, it’s vital that the most important foundation of any AI strategy is not overlooked – that is people.

“Today we know that AI implementations face many of the same challenges as their technology predecessors and at present, most organizations aren’t thinking enough about the human element of their planned strategies,” proposed Alex McMullan, CTO International at data storage platform company Pure. “Technology's not always the problem, it can be come down to people - and when it comes to AI, this is very apparent. Part of the people problem is apparent in the AI skills shortage, where a lack of skilled data scientists, data engineers, platform engineers and other relevant professionals is hindering progress and adding to its cost. Salaries for AI professionals are at a premium right now and can account for up to 35% of project costs.”

Because another element of the people problem is managing AI data professionals in the right way so they are able to produce useful outcomes and drive efficiencies, McMullan thinks that the problem is compounded. We know that an AI model is only as good as the data that goes into it, which is in turn a people problem. If we also remember that biases can be introduced into algorithms resulting from the use of poor data bias, it’s easy to see why we need to think of professional people first… and then, professional AI second.

“As we move to the next wave of broader AI adoption, it’s an ideal time for organizations to conduct a detailed analysis of what is needed in terms of people, compute, storage infrastructure and data to deliver scalable projects that generate real ROI. It’s vital that any organisation considering an AI implementation invest in the right resources and technologies designed for the future to guarantee success,” said Pure’s McMullan.

Arguably well placed to comment on this issue is Human Resources (HR), financials and planning software specialist company Workday.

Tasks first, jobs second

“Building any AI function starts with developing an understanding of the workplace tasks and individual use cases that you are trying to achieve,” said Workday chief technology officer (CTO) Jim Stratton. “We use copilots internally written in our own proprietary software language [Workday XpressO is essentially an object-oriented wrapper for SQL database queries combined with a page builder] and external technologies in this space too. The challenge is, AI-centric copilots are good at coding, but they’re not so good at software engineering and being able to appreciate the human need behind any given implementation of AI.”

For Stratton and team, the development and implementation of AI is not just to accelerate job functions, it is about automation in a broader sense to enable software engineering to deliver applications that work across more world regions, to scale and grow better, to work with improved data safety (and eradicate AI bias), to present better integration capabilities and, of course, to offer more functions.

But why would AI help developers to integrate better? Stratton explains this point succinctly. Say one application needs to connect to another, a function often executed via the use of an Application Programming Interface (API), technology elements that act as a ‘glueing’ bond between apps and data services; here we see a developer able to use AI to shoulder the data mapping (to understand the format, shape, schema structure & syntax of the data) involved and make the whole process happen more accurately and more rapidly.

“Because our implementation of AI at Workday across HR, financials and planning solutions is often a question of working with other workplace systems, putting the human element into our approach to AI is fundamental to what we do,” said Stratton. “As we now work to elevate individuals to more strategic work tasks, we will always start AI projects with a people-centric mindset from the developers that create it to the users that use it. You can’t build highly distributed, highly available, high-performance systems (like ours) automatically, no software engineer in this space is going to be replaced by a copilot anytime soon.”

The point made here is clear, if we think about the implementation of a finance and HR system, it is connected to hundreds of other enterprise software systems covering revenue, employee benefits and onward to sales and marketing and so on. In many organizations, Stratton reminds us that that’s a very fragmented ecosystem with hundreds (or thousands) of applications to connect to. As we stand today, he is adamant that keeping the human factor at the fore in any AI implementation is tablestakes, first base and ground zero.

Chief AI officer (CAIO) at IFS is Bob De Caux. As a cloud enterprise software company with a key platform focus on ERP and related disciplines across Field Service Management & Enterprise Service Management, IFS has a vested interest and skill set aligned to AI team building and development. Agreeing with many of the sentiments and insights expressed here so far, De Caux says that from his viewpoint, a key part of building AI solutions is building the user experience to put around AI and people are crucial for getting that right.

“It's not such an issue when you're automating processes, but for a number of our industry use cases a human is still in the loop making decisions, with AI providing decision support,” said De Caux. “Therefore it's critical that we deliver AI outputs in a way that builds trust and confidence in what is often a non-technical user, which means finding a way to explain why AI models are making decisions, providing transparency into the data being used in the process and translating highly quantitative results into something with business meaning. Building that explainability and transparency framework to meet user needs is a very human venture.”

People first, data second, then AI

If we explain that people, processes, data and information provenance, work systems and task mining all come first, that doesn’t stop us still having the data use case argument front and center of this discussion. Because one of the biggest fears surrounding AI and automation is that machines are equipped to replace the workforce en masse, and we’re eventually all going to lose our jobs to robots and an array of AI systems manning assembly lines and customer inquiries, it’s important to remember that the reality is a lot more nuanced - and not nearly as intimidating.

This is the opinion of Richard Jones, VP for Northern Europe at data streaming platform company Confluent.

“In automating routine tasks, AI both creates new opportunities and enhances existing roles – I see it helping make people more rounded, skilled employees. Taking on the burden of tedious admin frees employees to focus on more complex, creative and strategic tasks that add value to the business,” said Jones, speaking to press &v analysts in London this spring. “I see job roles gradually evolving and placing more emphasis on informed decision-making and human interaction – allowing employees to upskill as they grow into these evolving responsibilities. But AI does more than just enhance existing roles; we’re seeing it create entirely new roles, such as prompt engineers, machine learning developers and data scientists.”

Talking about how smart use of data can elevate AI to higher levels, Jones reminds us that AI can sift through extensive datasets to deliver highly personalized experiences, meeting the rising expectations of consumers.

“Through leveraging data streaming technologies, this personalization can even occur in real-time, resulting in customer service applications that respond accurately, in a human way and in the moment,” said Jones. “AI, with the assistance of humans, holds the promise of significantly elevating customer experiences. AI-driven chatbots and virtual assistants offer immediate assistance around the clock, efficiently handling common queries and issues without human intervention. This instantaneity has the potential to greatly heighten the customer experience, while simultaneously freeing up people to address the more intricate and sensitive matters that do require human intervention.”

We want human AI

We know that businesses have reported improved customer satisfaction metrics following the integration of AI into their customer service workflows, underscoring technology's ability to augment rather than diminish the human touch – that’s if employees are willing to work alongside AI.

The people first (and indeed data first) AI message may continue to resonate for a significant period this decade as we work to engineer AI services into our enterprise operations layers, into our workflows and into our lives. Perhaps the question will soon be - when someone tells you there’s a new AI service bubbling up - great, how human is it?