Staffing technology organizations solidified the concept of the “knowledge worker,” an employee primarily focused on using their brain to solve problems in an individual capacity, versus using their physical or managerial talents. Software engineers, technology architects, and digital strategists are all different versions of the knowledge worker, each using their skills and abilities to gather and synthesize information to produce an outcome.
With the emergence of natural language AI tools, ChatGPT being one of the most visible, it seems that these types of tools are squarely aimed at performing knowledge worker-like roles. While many companies are contemplating changes to their products, customers, and overall industry, too few are considering how to change their hiring and staffing based on these tools.
From Sole Contributor to Conductor
One of the greatest promises of AI tools that mimic individual knowledge workers is that dozens of virtual “workers” can be added to your teams for very little cost. This has given rise to speculation that bots will replace entire classes of jobs.
These types of articles make for great attention-grabbing headlines, but they miss the larger point that technology changes often engender shifts in how we work, versus creating a zero-sum game in which workers disappear forever. In the case of AI tools, the worker of the future may be taking pages from workers of the past, versus being eliminated altogether.
One of the intriguing aspects of the rise of the knowledge worker was the reduction in middle management roles, whereby a dedicated class of workers existed primarily to direct and manage the output of others. Everything from technology to cost-cutting efforts has reduced the need for, and impact of, these middle managers, to the point that the role has become regarded with suspicion.
However, the current crop of AI tools more strongly resembles a bright new hire or intern than a self-managing knowledge worker. AI tools will authoritatively make factually wrong assertions and lack the human ability to self-assess their competence and use that information to appropriately caveat areas where they lack experience.
Just as humans who are earlier in their careers benefit from guidance, mentoring, and oversight, so too bots will need experienced humans to oversee their performance. Bots are also supremely task-oriented, performing their assigned tasks immediately and rapidly. However, they lack any ability to manage their workload beyond the current tasks and are unable to break a complex or poorly defined objective into component tasks.
For example, a relatively straightforward assignment like “Get me around 8 slides for the executive presentation next week,” a task that any mid-level knowledge worker could perform with ease, would completely confuse the current generation of bots. A knowledge worker has the context needed to complete this request and the understanding that there are multiple discrete tasks required to fill this request, most of which could be successfully performed by bots.
Those tasks might range from performing independent research on several industry topics, speaking to various coworkers about their projects, creating a storyline and writing the associated content, and finally assembling that content into slides and sharing for review.
Intriguingly, the skills of middle management that many organizations have deemed redundant will be extremely relevant in these types of scenarios. A “conductor” of sorts will be required to break these types of tasks into their component parts, assign them to the relevant bots, and then revise and integrate their discrete outputs into a cohesive whole.
The other critical role required as companies begin to deploy bots will be technicians who can effectively “train” these tools. Most current bots are generalists and are trained on commonly available data like web content. While a fleet of low-cost generalists can certainly benefit your organization, exposure to internal data will ultimately produce the most benefit. Abilities like being able to pull last year’s financials and do a quick year-over-year analysis are perfect for bots, assuming they are trained on your internal data.
This requires specialized technical resources and an understanding of which data are relevant. These roles are in increasingly high demand and are a great domain for a partner that can provide on-demand resources to fill. Technical partners can provide expertise in training and enhancing AI-based tools; however, you’ll still require individuals who know which data and systems are the right sources for training data.
Finding your Conductors
Unfortunately, the trend away from middle management has deemphasized the skills that would help create a successful bot “conductor.” The ability to take complex tasks, break them into component parts, assign them to the right resources, and then reassemble and revise the result is often lacking with most individual contributors.
In addition to general management skills, your conductors will also require some degree of subject matter expertise. Bots lack the human “tells” when providing incorrect information and will confidently share information that’s downright wrong and will fail to develop the human trait of humility when corrected.
Ideally, your conductors either have enough content knowledge to vet the results your bots produce or have access to internal or external experts who can vet the results of your team of virtual workers.
This creates a bit of an odd dichotomy, whereby routine tasks and information can be generated by bots, but deep expertise is required to validate the work of the bots, as well as respond to requests that are too complex for the bots to complete.
Your conductors will need to manage both emotionless machines, as well as highly specialized experts, managing the collaboration and integration between each.
These types of individuals may not currently be on your tech teams, presenting a bit of a challenge. Leaders will be faced with the decision of whether to develop knowledge workers into conductors, or seek partners that can provide conductors, teachers, and the bots themselves. Alternatively, competent middle managers that exist outside traditional tech roles may be well-equipped to become AI conductors.
Look for individuals that have been successful at managing teams of individual contributors and integrating the results of their work, even if that individual is not a deep expert themselves.
For tech leaders in an AI development company, what made your teams successful in the past—deep expertise, ability to work as an individual contributor, and minimal oversight or management requirements—is not necessarily the recipe for success in an AI-enabled team in the future.
Whilst many consider AI advances as ‘scary’, leaders will need to determine whether to invest in upskilled knowledge workers so they can develop expertise that exceeds the abilities of current AIs or shift them toward a conductor-type role. In all cases, the tech shop of the future, and the composition of the staff in those roles, is poised for significant change.
If you enjoyed this, be sure to check out our other AI articles.
