This third post in our series explores the evolving role of AI in enterprise, focusing on the rise of digital workers. At Artium, we observe firsthand how AI transforms business operations, and this post delves into these changes, offering practical insights and strategies.

Many discussions about the future of work are too abstract. As AI redefines employees' roles, let's focus on practical changes in companies. In this post, I'll outline how to prepare for and benefit from this change.

For now I’ll leave the discussions of job displacement, data privacy/security, AI bias, and ethics to other posts and focus on how we transition. 

Adopting AI and digital workers has become a strategic imperative for enterprises seeking to maintain a competitive edge. Early adopters gain significant advantages by streamlining processes, improving efficiency, and driving innovation. Digital workers offer businesses the ability to differentiate themselves in the market, optimise various functions, scale with ease and flexibility, and unlock valuable data-driven insights. Investing in these technologies is crucial for future-proofing your business, as those who fail to adapt risk falling behind their more proactive competitors and becoming obsolete in the AI age.

Make way for exponential disruption 

AI models improve 5-10x per year, fundamentally changing how we work. Harnessing AI's full potential requires rethinking business operations and mental models.

My prediction is that AI will bring about the rise of hybrid talent, which I call “polymath professionals.” Individuals who can leverage AI to augment their skills across multiple domains will be highly valued. Generalists and jacks-of-all-trades will become desired rather than avoided. Specialists will still be needed, but only in specific roles. 

First, Individuals who can adopt AI to augment their skills across domains have an opportunity to drive exponential outcomes while focusing on more creative strategic work. This shift will lead to the emergence of small, agile teams that can achieve massive impact and productivity by combining their unique skills and expertise with the power of AI. I believe the first one-person, billion-dollar company is imminent. 

Second, it will become increasingly important for businesses to focus on encoding knowledge and building digital brains. By capturing and organising your company’s collective knowledge and expertise in a way that AI can easily access and leverage, you can create a powerful competitive advantage. This will require a new approach to knowledge management and a willingness to invest in the tools and technologies needed to make it happen.

Finally, it's important to recognise that simply layering AI on top of existing processes won't be enough to realise its potential fully. To truly transform your business, you need to be willing to reinvent your processes from the ground up with AI at the core. This may require significant time and resources upfront, but the ROI will be worth it. Understanding the different types of digital employees that will emerge in the new workforce is crucial.

Digital workers decoded - 3 types of digital employees

I expect a shift in the workforce which will introduce the concept of digital employees.  These “employees” will exist on a spectrum from digital assistants to fully autonomous digital team members, leveraging proprietary data to perform tasks effectively.

In the same way that HR has shifted to assessing a candidate's specific abilities rather than their background alone,  digital employees will be designed to augment and replace certain skills rather than entire roles. Like any new employee absorbing a company's culture and processes, these digital employees require the "memory" of the company’s proprietary data to truly excel in a specific environment. This data imbues AI systems with the contextual knowledge necessary to apply their capabilities meaningfully.  The process mirrors the learning journey of onboarding and upskilling a human employee but takes minutes instead of months and years. 

Let's examine three snapshots along the spectrum of new digital employee types, one at a time: digital assistants, digital doppelgangers, and full digital employees.

Empowering digital assistants

From simplifying tasks to enhancing customer engagement, digital assistants, also called “co-pilots”, epitomise the initial steps towards integrating AI into our daily work lives. Their evolution from basic support to sophisticated advisory roles showcases AI's potential to complement human efforts.

Digital assistants exist in different forms. The first is the swarm of new SaaS platforms with a narrow focus on solving specific problems like summarising meetings. Next, we have AI integrated into OSs and workspace platforms, such as Google Workspace and Microsoft Teams, Windows and, soon, Mac OS. Lastly, we have a swarm of bots called “agents” that will have different specialities working together. These agents have some limited autonomy in that they can hand tasks off to each other to achieve a goal set for them. A key aspect is that, like human staff, they have access to the SaaS platforms we already use, making integration easy and powerful. 

This type of assistant already exists and is being used by entrepreneurs and influencers worldwide, but it has started to leap into the enterprise. For example, many new YouTube influencers do not hire marketing agencies - they have been adapting to use a digital assistant backed by a swarm of agents to design everything. It looks like this:

1. Prompt the agent to design the next 5 videos they need to make

2. The agent asks a research agent who has been primed on the domain of the influencer to go out on the web and look at what others in the space are writing and talking about

3. It comes back with a list of topics

4. Another agent, familiar with the target audience and specialised in analysing metrics, looks at past video performance and industry trends and selects the likeliest top-performing posts

5. Another agent, specialised in research, takes each topic and breaks it down into an outline of the script, paying attention to creating a hook and a flow that leads to a big point

6. It passes each outline to another agent, trained on the influencer’s tone and speech patterns, who writes a script

7. It passes the script to another agent, who generates visuals, such as a thumbnail and marketing messages, to accompany the points in the script.

8. These 5 scripts with visuals are passed back to the influencer, ready to record 

It’s easy to imagine this type of flow completely transforming how an enterprise marketing department creates campaigns. A similar flow can handle creating any number of different business artefacts. And this is the lowest level of AI enablement in our spectrum of digital employees. 

These AI systems, represented by chatbots, SaaS-based digital assistants, and agents, automate routine tasks and provide personalised interaction, setting the stage for advanced AI integration in daily operations.

Training digital doppelgängers 

Digital assistants discussed above are defined by taking the worldly knowledge of Large Language Models and reducing the domain space. For example, creating a researcher agent is as simple as saying, “You are an expert senior researcher” - that statement alone is enough for the LLM to reduce focus on all of the other skills it has access to and focus on the skills of research. It can then apply these to whatever problem you give it. 

These agents can be further focussed by including extra data. For example, with the influencer assistant above, the initial researcher had been given the transcripts of the previous videos created by the influencer to understand the subject matter and correctly identify subject matter that would suit that influencer. 

Digital doppelgangers go a few steps further. These digital replicas of employees are trained on all of the content of a specific employee. This specialised agent is given, for example, all of the following from that employee:

• documents they wrote or to which they contributed 

• slack message  

• emails

• meeting/ call transcripts

• Blog posts

• Linkedin posts

• work-related social posts

• notes

• and everything else digital

Just as LLMs are trained on all the digital data accumulated by the human race, digital doppelgangers will be trained on all the digital data accumulated by a single person.

These digital employees will open new avenues for training, simulation, and capacity augmentation. Instead of attending endless meetings and conversations to facilitate knowledge transfer from one part of an organisation or team to another, digital doppelgangers will handle these interactions in a fraction of the time. 

Working with digital assistants, digital doppelgangers can complete many of the necessary tasks without the intervention of their human counterparts, which can benefit from the effort's outcome. They embody the potential for AI to enhance human capabilities, offering a glimpse into a future where learning and experience transfer are seamless.

By creating digital versions of existing employees, organisations can extend their operational capabilities, blending the virtual with the real in innovative ways.

Enabling digital employees

A full digital employee is at the extreme end of the spectrum. It uses the general human knowledge provided by the LLMs in combination with the knowledge of all employees in the role within a specific organisation or, if accessible, across the entire industry. 

Representing the pinnacle of AI's integration into the workforce, these fully digital employees can undertake complex decision-making and operational tasks. Their development speaks to a future where AI's role transcends assistance, becoming a cornerstone of enterprise strategy and innovation.

These autonomous agents mirror the polymath professional in digital form. They mark the frontier of AI's integration into the workforce. Capable of independent decision-making and task execution, they can manage complex workflows, analyse vast datasets, and drive strategic decisions in a way humans cannot, all without human oversight.

Reduce meetings with a streamlined knowledge flow

Most corporate meetings are centred on the transfer of information. This can be from one person to another, one department to another, or one team to another. This highly inefficient process consists of creating slide decks and presentations and then walking others through them while they browse on their phones and dream about weekend plans. 

In the past, I taught teams to replace meetings with workshops, which meant eliminating “agendas” and creating a list of deliverables instead. This focused everyone on the outcome and turned meetings into working sessions. 

Digital assistants and doppelgangers can be sent to meetings and even workshops. They can then return with summaries of information itemised, prioritised and presented simply to their human counterpart for the highest effectiveness. 

This does not mean that humans never meet, but, like the workshops, it allows us to focus on effective productivity instead of endless information transfer. By leveraging digital employees to streamline information transfer and reduce meetings, we pave the way for a new era of augmented intelligence in the workplace.

Augmented intelligence rising

Augmented intelligence is the integration of human and AI into a single force. It is the underlying principle of this new culture of work. 

The integration of AI into the workforce isn't an overnight shift but a gradual journey. You may not know it, but it has already begun. Employees are already using LLMs to augment their work. Even in environments where certain platforms have been excluded, employees turn to their smartphones and tablets, like a teen sneaking a calculator into a maths exam. 

Rather than allowing this unstructured approach to integration, embracing the transition allows an organisation to leverage the swell of this bottom-up innovation. 

Organisations can iteratively develop more complex AI entities starting with basic digital assistants. This process mirrors the natural evolution of human roles within the workplace, emphasising learning, adaptation, and innovation. By beginning with what's achievable today, businesses set the stage for a future where AI's full potential can be realised, creating a dynamic, collaborative workforce.

How to start creating digital employees

In March 2024, Cognition Labs introduced Devin, promising a full digital software engineer. While Devin is not ready to replace human programmers yet, its capabilities are promising. Ambition is exciting, but starting simple is better. 

In eXtreme programming, which we practise at Artium, the idea of starting simple and then iterating to add complexity seems apropos. As we are all at the beginning of this journey, and no proven roadmap yet exists, using this related framework is the best way to ensure success. Here is an outline of steps that can be adapted to different use cases:

1. Identify a particular department or area with a clear set of proprietary data. Having data that no one else has makes the advantage of a digital employee much clearer. 

2. Create an agentic version of this data set. This means using a mixture of prompting, LLMs, small models, fine-tuning and RAG to create a chatbot version of an employee who is an expert in your particular domain. 

3. Give this digital team member access to related repositories of new data, such as related Slack channels, Google Doc repositories, data warehouse dumps, etc. This ensures the AI stays up to date with the latest. 

4. Let related staff use the chatbot to augment their workload, checking their work against the AI, using it for research, writing, presentation content and thinking through ideas. 

This is the most lightweight version of a digital employee. We evolve it by giving it more context. Even this simple version includes some hidden complexities, such as managing short-term and long-term memory, avoiding unwanted hallucinations, and embedding bias. However, these are all solvable problems. As we build this first generation, we address data preparation, model training, integration with existing systems and processes, governance, and more. 

As technology progresses, we can see a future where, using a combination of intelligent engineering and training human collaborators, your new digital employee can operate with the same or lesser risk than a human employee. As the cost of implementing AI continues to drop in the coming years, we may see the cost of building these new employees as no more than an annual executive salary, making them very accessible.

We work with a number of companies, building their first digital employees. Some of the skills we have come across include:

• research 

• subject matter expertise

• product management

• engineering plan assessment 

• marketing metric analysis

• code quality assessment

There are a million more to explore in your industry. 

The path to a new workforce

Taking proactive steps to integrate digital employees into your organisation is essential for long-term success. Embracing digital workers will position your company to remain competitive, agile, and future-ready. Understanding the different types of digital employees is the first step to preparing your organisation for change.

As we build this first generation, we need to address data preparation, model training, integration with existing systems and processes, governance, and more. A critical factor that cannot be overlooked is the importance of leadership buy-in. The successful adoption of AI and digital employees requires a strong commitment from business leaders to drive the necessary cultural and technological changes.

As tech leaders, it's our responsibility to guide our organisations through this transition. We must approach it with a spirit of experimentation and continuous learning, starting small and iterating as we go. We must also prioritise the human element, ensuring our people are supported, engaged, and empowered to thrive alongside their digital counterparts.

While the journey to company-wide augmented intelligence presents challenges, the rewards are immense. Those who navigate this shift successfully will lead their industries in the coming years.