Uses of Artificial Intelligence in HR

Many companies are already using AI across their business. Retailers are providing product recommendations to customers and improving their stock management. Insurers are predicting fraud risks in real time. Manufacturers are monitoring events in the real world and reacting to anomalies.

Many of our colleagues are already using AI in their everyday work.

Marketing teams are using AI to help them write copy for their website. Sales teams are guided as to when to contact their prospects and have their outcomes predicted. Customer service teams can use AI to help guide users to support documents. IT teams are using AI to help them respond to previously unseen security threats.

And people with little artistic talent (ahem!) can create images for their articles! Here's something that Bing created for me from the input "Heading image for an article about how artificial intelligence can be used by HR teams. Include something technical - maybe a person talking to a computer? Something inspiring in the style of Picasso's early work."

So is HR missing out? What can we do today, what will we be able to do soon, and what do we need to worry about?

While some HR teams have been successfully using tools such as Robotic Process Automation for some time, many organisations have found working with data a challenge. ^{1. CIPD}

Text processing and understanding

Our everyday tools such as Word, Outlook and Google are already finishing our sentences for us. The tools we work with often know what we're trying to say before we know ourselves. Outlook suggests replies to emails. Google understands the meaning of what we're searching for and proposes alternatives.

Businesses are starting to embrace the opportunities. For example, Microsoft Graph allows us to bring the content our teams create into the world of AI.

AI can help us understand the overall sentiment of everyday language documents without the effort of reading them ourselves. Search tools can suggest content from our teams that might be most relevant to us. Custom connectors can add AI capabilities on top of information stored in separate systems and unlock huge potential. Connectors are available for many existing platforms such as Salesforce, Confluence, Oracle and Workday (in preview). Custom connectors can be created for line of business applications or for databases where an off-the-shelf connector is not yet available.

AI can help us cope with the vast amount of plain text documents and unstructured data. These tools are available today - many of them within the Azure cloud.

Automatic conversation summaries

Dynamics CRM can convert the conversations in Teams meetings with sales prospects into plain text. It can summarise the conversation and highlight key points from the meeting. More generally, Teams has premium features such as intelligent recap with personalised highlights. Automatic chapters help find parts of meetings most relevant to us.

Imagine the possibilities for conducting job interviews that are automatically summarised and searchable.

Content generation

The images on this page were all generated from short text samples given to the DALL-E tool. They showcase an ability to parse and understand natural language and are grounded in massive amounts of pre-existing content. This allows the creation of images in styles similar to famous artists. It's a fun tool for generating images.

Image generation can be great for research and inspiration before a human artist takes over. They can also be useful for enhancing the appearance of presentations (and web articles). They are subjectively pleasant enough, and often a useful alternative to standard (are sometimes overused) stock photos.

Text generation is perhaps more relevant to our world. ChatGPT has given us almost magical content generation, that has made us question our own unique abilities. This article was written without AI-generated text, but it's quite possible that ChatGPT could have done a better job given the right inputs.

ChatGPT and similar tools work from Large Language Models (LLMs) as their starting point. LLMs are derived from vast amounts of text taken from across the Internet. This includes a staggering amount of published fiction, scientific literature and even forum discussions. It's easy to see why these tools perform so successfully in exams. And these are exams taken without any specific training. ^{2. OpenAI}

Content generation can be a great starting point for a variety of tasks that we often struggle with. Inspiration can come slowly when creating presentations and documents. AI can give us a head start with ideas we might not have thought us. Essentially the AI knows how other humans have attempted to solve the problem before us.

Whether the solutions are correct, valid, legal and ethical is for us to decide. Like with all powerful tools, humans must take care.

Data Analytics

Working with large and complex data is an increasingly important part of our role. Artificial intelligence can help us with these challenges.

Tools such as Power BI have AI features built in. Text analytics can extract key phrases and analyse sentiment from plain text. Anomaly detection points us to unusual data points. Machine learning generates insights and predictions from data sets.

When we need to understand the reason for changes in our data, AI can help us investigate the information and detect patterns that we would be unlikely to spot.

Image created by AI - "Show data flowing from one data source through a processor function and splitting to multiple outputs, in the Renaissance style of da Vinci."

Algorithms affect us all. YouTube and Netflix use our history to suggest what we should watch next. LinkedIn suggests new roles it thinks we might prefer.

The potential for AI-based algorithms to make bad decisions about people is obvious.

The responsibilities of HR teams remain the same – to treat people fairly, avoid direct and indirect discrimination, promote diversity and lead our teams' ethical conversations. The challenge is how to continue to do so when we rely on algorithms that we don't, and can't, fully understand.

AI is something of a black box – a function that generates an output from a given set of inputs, but the exact inner workings are unknown to us. We can't see inside. Even when we objectively know the process, the data sets can be so massive that we simply can't reason about them. Some algorithms are non-deterministic (or stochastic), so won't always give the same results to the same input.

We must be careful to avoid discrimination. For example, facial recognition can fail for some people because of inadequate training data.

AI can make incorrect decisions. So too can humans, of course. But AI can make incorrect decisions at unprecedented speed and scale. Serious consequences can occur before we even realise them. We should always be ready to justify our decisions. This can be challenging when we rely on AI, as we are unlikely to fully understand the process.

Ensuring data security is a crucial element of AI. Security can be more challenging for AI systems. Incorrect configuration can expose data to the wrong people. AI can inadvertently bypass existing security protocols.

AI is at such an early stage that society doesn't yet have the answers. We don't yet know the questions. There are problems we haven't imagined yet. But we remain in charge of the decisions recommended, and humans must guide the ethical use of AI.

Image created by AI - "Risk and reward of technology, in the style of Salvador Dali"

Artificial intelligence is incredibly exciting. Magical? Yes. Unnerving? Definitely. Boring? Quite possibly! The reality of transformational technologies is that the journey often starts with some basic housekeeping.

• Data engineering

  The process of transforming your data and making it available is referred to as data engineering. It's not glamorous and is often unseen. But data engineering is essential – both for traditional management reporting, modern analytics, and future AI decision-making.

  The more data you have, the more opportunities you will be able to find. The better your data quality, the better your outcomes are likely to be.

  Good data engineering can help ensure the accuracy and integrity of your data.

• Automate

  If you're still creating management reports and dashboards manually, automation will be useful. Reports depend on a data pipeline that combines and transforms data from different sources. Manual pipelines are expensive to run, unreliable and hard to scale.

  Once a pipeline has been automated, it becomes much easier to extend the processes. Data can be easily pushed to a data lake for future machine learning. AI routines can be added to the process to influence the outputs.

• Grow your technology stack

  The cloud is synonymous with AI. While it's true that the cloud can speed up the delivery of AI projects, legacy on-premises systems can still benefit from AI. Machine learning services and AI processes can be hosted separately to simplify the implementation.

  If you don't yet have an HR data lake, it is worth considering the benefits. Data lakes can make securing your data easier without compromising existing business systems. They can also help speed up project timescales and remove the dependency on internal IT and application development teams.

  Improving your reporting tool set can make adding AI easier. Modern tools such as Power BI have built-in AI features and can be connected to machine learning models.

• Grow your people's skills

  Many HR departments have focussed on soft skills and are naturally oriented towards people rather than data. HR teams often struggle to compete for highly skilled talent with departments that are better funded.

  A partial solution is for HR Business Partners to focus on a data-centric approach. Many HR BPs express a desire to become “more strategic”. Data skills give people the confidence to lead conversations with objective data and the information necessary to justify their opinions. Understanding the data they work with every day is an important step to identifying the possible benefits of machine learning and AI.

  HR people don't often aspire to be data scientists, but their business knowledge and instincts can guide the use of AI for the decision-makers that they work with.

  Improved data presentation and communication skills can be particularly useful for HR people. Managers and senior leaders can find data intimidating. People are often suspicious of statistics. Well-presented charts go a long way to conveying stories.

• Find a simple use case

  The hardest part of a journey is taking the first step. Technology in general, and AI in particular, can be intimidating.

  For clients beginning their AI journey, we recommend finding a simple problem that demonstrates some business value. The returns don't have to be huge. A proof of concept can demonstrate the technology and be useful in discussing opportunities.

  It can sound surprising, but failing fast is good. It's certainly better than failing slowly. Failure can be a useful guide towards success.

  And if your first use case can demonstrate a little bit of magic (perhaps like some of the images on this page), you'll definitely catch people's attention.

We love helping people see the value in their data and discovering the potential of AI.

Get in touch to chat with us about the opportunities of AI for your business.

Hurry up though - we may not have much time left!

Image created by AI - "Concern, but ultimately hope, in the style of Andy Warhol" / "...Banksy"