Upskilling in the AI Era

Meg Donovan, the Chief People Officer at Nexthink, provides some commentary on what upskilling looks like in the AI era, how companies can develop an upskilling plan for their workers, and more. This article originally appeared in Insight Jam, an enterprise IT community that enables human conversation on AI.

Digital upskilling has never been a straightforward task, but in the era of agentic AI, it has also never been more important. The UK government, for example, has invested millions in providing free training for every adult in the UK, aiming to help 10 million people improve their AI capabilities by 2030.

Such initiatives are laudable, creating huge economic benefits for all concerned, and should be emulated by more countries. But they also show that basic familiarity and aptitude with AI tools is swiftly becoming the bare minimum required to remain globally competitive. If businesses want to retain a competitive advantage, they will have to support employees in going further and developing deep expertise. Skills such as iterative prompt engineering or output verification will need to be improved rapidly in a short period of time.

However, doing so requires a substantial resource investment in a short period, leaving very little scope for error. Worse, because every organization faces its own unique challenges, there is no ‘right’ way to go about prioritizing AI upskilling initiatives. Dozens of factors, such as industry, company size, and existing processes, need to be considered when developing a robust upskilling strategy.

Finding the Good, the Bad, and the Ugly

There are three types of use cases businesses need to identify when it comes to AI. The good (i.e., use cases that are resulting in staff becoming faster and more productive), the bad (those that might be saving time initially but are creating additional work down the line), and the ugly (use cases that could create significant financial, reputational, or other risks). What makes it more complicated is that sometimes the same use case can fall into multiple categories.

Let’s take email drafting as an example. There are so many elements that can impact which category it falls into. Is the recipient internal or external? Is there any sensitive information being included? Are there legal or regulatory questions involved? Even though the action is the same, wildly different levels of scrutiny and verification may apply depending on the context.

There are some general best practice guidelines around monitoring and transparency that apply in almost all instances. But there are also plenty of use-case-specific issues where users will need more targeted support because they aren’t sure how to proceed safely.

Defining the Top 10

At Nexthink, we’ve put significant resources into identifying the top ten AI use cases that are adding value to the business:

• Research
• Content Writing
• Problem Solving
• Process Automation
• Data Analysis
• Email Drafting
• Creative Ideation
• Quality Assurance
• Translation Tasks
• Learning & Training

Each of these use cases can also be broken down by department and tool, giving us much more granular detail into how adoption is happening and enabling us to be more targeted in our interventions. For example, while Product and Marketing are mostly using ChatGPT, Pre-Sales leans more heavily on Copilot. And these are not just technical metrics—we also collect user sentiment feedback to understand, in real-time, the frustrations they are having and what support they feel is needed.

Using these insights, we have been able to start building a holistic AI use case library, and we’re also refining it with better role-specific prompts based on differences between favored tools. Even more importantly, this breakdown allows us to be more precise when creating and enforcing output verification guidelines.

Cutting Across Siloes

One of the biggest benefits of this level of insight is enabling better cross-functional collaboration through targeted interventions. For instance, if one of our marketing teams is experiencing learning barriers to using Claude for data analysis, we can identify Claude champions in other departments who have had success in that area and arrange a workshop or Lunch & Learn to share their knowledge.

Without these knowledge-sharing systems, people inevitably waste time and money attempting to replicate processes that have already been tried and discarded, or that have been iterated on to produce better results. Moreover, the less cohesion in AI adoption there is within the organization, the harder it is to maintain and scale successful solutions and use cases.

Putting People First

Ultimately, there is no one-size-fits-all approach to upskilling employees in the AI era. Nor can it be left to individuals to figure out which of the dozens of models and tools works best for them. To succeed, businesses will need to develop far more personalized training programs tailored to how their employees use these tools and the use cases they find most valuable. Having that foundation in place not only enables employees to be more productive and collaborative, but it’s also crucial to fast, effective scaling of AI across the enterprise.