Learn how Complexio is building Human-Machine Teaming
Achieving Autonomy Through Human-Machine Teaming:
From Task Automation to AI Agents and Reasoning
Augmenting Human Intelligence with Machine Performance
Human-Machine Teaming integrates human creativity and empathy with machine precision and data processing to enhance productivity and efficiency. This synergy is defined by Brookings Institution as a relationship involving humans, machines, and their interactions. The concept of Hybrid Intelligence, introduced by Van der Aalst, underscores how AI can augment rather than replace human capabilities. Today’s workplace faces productivity challenges such as excessive administrative tasks, context switching, and workflow inefficiencies. Human-Machine Teaming addresses these by automating repetitive tasks, improving data handling, and reducing errors.
PWC defines automation evolution in three waves: Algorithm Wave (structured data tasks), Augmentation Wave (repeatable tasks and workflow enhancements), and Autonomy Wave (AI-driven decision-making with minimal human intervention). Effective Human-Machine Teaming requires strategic task allocation: computers excel at high-precision tasks, while humans handle uncertainty and creative problem-solving. Current implementations include manufacturing robots, AI in healthcare diagnostics, and chatbots in customer service. Task automation tools like HubSpot, Salesforce, and UiPath streamline repetitive processes but face limitations in flexibility and integration. AI Agents, unlike traditional tools, autonomously interact with their environment to achieve goals, with applications in personal assistance, autonomous robots, and fraud detection.
Complexio’s technology combines Foundational AI, Prediction and Reasoning, and a Natural User Interface to enhance human-machine collaboration. This approach improves workflows, adapts to user behaviour, and drives operational excellence, making it a key driver of sustainable growth and efficiency in modern business environments.
Introduction
Defining Human-Machine Teaming
Brookings Institution defines Human-Machine Teaming as a relationship – “one made up of at least three equally important elements: the human, the machine, and the interactions and interdependencies between them.” It enables humans and machines to work together, allowing perfect synergy as each party completes the tasks that best suits them (Building Trust in Human-Machine Teams, n.d.).
Van der Aalst presents a different perspective on Human-Machine Teaming, referring to it as “Hybrid Intelligence.” Hybrid Intelligence (HI) focuses on the supportive role of machine learning, like deep neural networks, and other data-driven techniques in enhancing, rather than replacing, human intelligence. The concept of Hybrid Intelligence combines two types of intelligence:
human intelligence:
Characterised by flexibility, creativity, empathy, instinct, and common sense, it centres around people and their experiences.
machine intelligence:
Characterised by speed, efficiency, cost-effectiveness, scalability, and consistency, it focuses on data and algorithms.
“We aspire to create a human-centric, ethical, and trustworthy AI platform aligned with our corporate CARE values. We aim to foster innovation and advance our efficiency as one team, harnessing AI to simplify and improve every aspect of our business, guiding us towards operational excellence.”
Charles Maltby, 2024
Current Challenges in Workplace Productivity
Employees spending excessive time on administrative tasks
Workers being the bridge between systems, transferring data from one system to the next by navigating between different platforms
Lost productivity due to context switching between multiple applications, systems, and tools to gather and combine information
Diminished motivation from productivity blockers
Reduced capacity for strategic thinking because of imperfect and incomplete data
Workflow inefficiencies and human errors often require going back to fix small mistakes, disrupting productivity and adding unnecessary delays
Human-Machine Teaming aims to mitigate the costs of inefficiencies and human errors while managing the complexities of expanding tech stacks and enhancing our ability to act quickly and innovatively on the vast and unexplored whole company data landscape.
Human-Machine Teaming – Three Waves of Automation
PwC highlights three waves of automation in their news release “Education and retraining critical to help workers adjust to future waves of automation.”
1. Algorithm wave:
This wave is already well established and involves automating structured data analysis and simple digital tasks.
This can be seen in existing work automation tools like automated sales pipeline management and financial tasks like credit scoring. It is currently in its mature stage.
2. Augmentation Wave:
This wave is underway and is expected to fully mature later in the 2020s. It focuses on automating repeatable tasks and information exchange, with advancements in aerial drones, warehouse robots, and semi-autonomous vehicles.
This can be seen in the current batch of copilots. They are already demonstrating effectiveness in areas like code completion and email drafting, but their adoption is not yet widely spread. In the context of Complexio, we are bringing Augmentation wave benefits to general users doing complex recurring tasks. For instance, Complexio can enhance tasks such as drafting emails for charterers during the clearance process, automatically locating lost invoices and forwarding them to the accounting department and comparing terms in charter party agreements when changes have been made.
3. Autonomy wave:
Expected to mature by the mid-2030s, this wave will see AI increasingly capable of analysing data from various sources, making decisions, and taking actions with minimal human intervention. For example, fully autonomous driverless vehicles could become widely deployed during this phase.
At Complexio, we believe that this wave represents the next generation of Human-Machine Teaming. Our technology will incorporate Agents and Reasoning as integral components. While current iterations of Agents and Reasoning face challenges and limitations, we will be at the forefront of the Autonomy wave once the technology matures because we are leading the way in today’s Augmentation wave and will be ready to deploy newer technologies as they mature. Source: (Education and Retraining Critical to Help Workers Adjust to Future Waves of Automation, n.d.)
“Anything that is routine, the machines will be doing.”
Gerd Leonhard, 2024
Determining Task Allocation
Human-Machine Teaming begins by identifying which tasks are best suited for humans and which are best suited for computers, using the strengths of each to optimise performance and productivity.
Computers excel at complex and recurring tasks requiring high precision and accuracy, such as calculations, data analysis, and specific mechanical operations. When combined with AI, they become invaluable tools for problem-solving, processing large amounts of data from various sources. Their data acuity and AI algorithms enable them to discern hidden patterns and trends not obvious to humans, identify innovative solutions, scale personal experiences like shopping and customer service, and provide data to support gut-based decisions (Walker, 2023).
Humans are best suited to deal with uncertainty and unusual situations, to think creatively, solve new and complex problems, and adapt quickly to changing environments. This allows us to make well-informed decisions in complicated and changing situations where computers struggle. Our unique capacity for intuition allows us to excel in tasks requiring empathy, such as communication, reading body language, and understanding social cues. We can grasp context in situations, processing multiple sensory cues at once – sight, sound, and information – and excel at complex, multi-staged tasks that require flexible motor skills, adaptability, and on-the-fly judgement (Walker, 2023).
In summary, computers are best at complex recurring tasks, and humans thrive at innovation, handling uncertainty, abstract thinking, empathy and adapting to changing environments.
Present Day Human-Machine Teaming
Today, Human-Machine Teaming is common across many industries, combining human skills with machine accuracy to improve performance. Machines handle repetitive tasks, while humans focus on more complex and strategic work. For example, in manufacturing, robots take care of routine assembly tasks, allowing humans to handle quality control and problem-solving. In healthcare, AI helps diagnose conditions by analysing medical images, while doctors make treatment decisions. In customer service, chatbots answer standard queries, with human Agents handling more complex issues. These examples show how Human-Machine Teaming uses AI and automation to streamline processes, increase efficiency, and allow humans to focus on tasks that need creativity, empathy, and critical thinking.
Examples of Human-Machine Teaming
MANUFACTURING
Robots and automated systems work alongside human workers in factories, enhancing productivity and precision.
HEalthcare
AI assists doctors in diagnosing diseases and interpreting medical images, while robotic systems aid in surgeries.
Transportation
Semi-autonomous vehicles combine human decision-making into machine operation for improved navigation and safety.
Customer Service
Chatbots handle initial inquiries, with human Agents stepping in for more complex issues.
DATA Analysis
AI processes vast amounts of data, with humans interpreting results and making strategic decisions.
Creative Industries
AI tools assist in generating ideas or initial drafts, while humans refine and add creative input.
EDUCATION
Adaptive learning systems personalise content, with teachers providing guidance and emotional support.
SCIENTIFIC RESEARCH
Machines process complex calculations and simulations, while researchers design experiments and interpret results.
AGRICULTURE
Automated systems handle routine tasks, with farmers making high-level decisions about crop management.
FINANCE
Algorithms assist in trading and risk assessment, while human analysts make final investment decisions.
Task Automation Tools
The first step in Human-Machine Teaming is task automation.
“60 percent of occupations have at least 30% of constituent work activities that could be automated.”
McKinsey & Company, 2017
Defining Task Automation
Currently, Human-Machine Teaming is predominantly centred around task automation. Task automation uses computer systems to perform actions without human intervention. The actions can range from simple but time-consuming tasks to complex processes requiring thorough analysis. Task automation can be achieved through technologies like SaaS (Software as a Service), IoT (Internet of Things), machine learning, or artificial intelligence. The goal of task automation is to minimize direct human involvement to gain efficiency (Andrieu, 2024).
Examples of Task Automation Tools in Various Industries
SALES AUTOMATION
Software like Salesforce and Zoho CRM automate sales processes, including lead tracking and follow-ups.
Trends in Task Automation
Recent trends in task automation are significantly influenced by advancements in technology, particularly AI and machine learning (ML), and the growing demand for efficiency across various sectors. Here are key trends shaping the landscape of task automation:
TREND
01
Integration of AI and Machine Learning
AI and ML are at the forefront of automation trends, enabling businesses to automate repetitive tasks, enhance decision-making, and deliver personalised customer experiences. These technologies help streamline processes and improve overall productivity by analysing data and predicting outcomes.
TREND
02
Robotic Process Automation (RPA)
RPA is increasingly utilized to handle high-volume, repetitive tasks that were traditionally performed by humans. This technology enhances efficiency and reduces operational costs, making it a popular choice for organisations looking to optimise their workflows.
TREND
03
Low-Code/No-Code Platforms
The rise of low-code and no-code platforms democratises automation by allowing non-technical users to create and manage automated workflows without extensive coding knowledge. This trend empowers more employees to participate in automation initiatives, speeding up implementation and reducing reliance on IT departments.
TREND
04
Hyper-Automation
Hyper-automation combines multiple automation technologies, including RPA, AI, and process mining, to automate entire end-to-end processes rather than just individual tasks. This approach aims to maximise efficiency and minimise errors across complex workflows.
TREND
05
Generative AI
Generative AI is emerging as a critical component in task automation, enhancing capabilities in areas such as content creation, customer service interactions, and data analysis. Its integration into automation processes is expected to grow as organisations seek innovative ways to leverage this technology.
TREND
06
Cloud-Based Solutions
The shift towards cloud-native platforms is reshaping how organisations implement automation. Cloud solutions provide scalable resources that can adapt to changing business needs, facilitating remote access and collaboration among teams.
TREND
07
Focus on Security and Compliance
With increasing cyber threats, there is a heightened emphasis on automating data security measures and ensuring compliance with regulations. Organisations are investing in automated systems that monitor security protocols in real-time to protect sensitive information.
TREND
08
Sustainability Initiatives
As environmental concerns gain prominence, businesses are adopting “green automation” practices that focus on optimising processes to reduce energy consumption and waste. Automation technologies are being leveraged to monitor resource usage and enhance sustainability efforts within organizations.These trends illustrate a broader movement towards more intelligent, flexible, and user-friendly automation solutions that not only improve operational efficiency but also adapt to the evolving needs of businesses in a digital economy.
AI Agents
AI Agents differ fundamentally from AI tools in their level of Autonomy.
Defining AI Agents
Technology trends are increasingly steering us towards a future where human-machine collaboration is driven by AI Agents. An artificial intelligence (AI) agent is a software program designed to interact with its environment, gather data, and utilise that data to autonomously execute tasks aimed at achieving specific goals. While humans set the goals, the AI agent independently selects the best actions needed to fulfill those goals. AI Agents differ from traditional software tools, which perform tasks predefined by developers, by making rational, reason-based decisions based on perceptions and data to achieve optimal performance and results. They perceive their environment through physical or software interfaces, enabling them to sense and respond effectively (What Are AI Agents? Agents in Artificial Intelligence Explained – AWS, n.d.).
But while AI adoption in business is experiencing rapid growth, it has still yet to unlock the full value of AI-driven efficiency.
“There is a very small subset of business problems that are best solved by machine learning; most of them just need good data and an understanding of what it means.”
Lorang, 2016
Attempts to Build AI Agents
While Reasoning in Agents is yet to fully be realised, efforts have been made in application-specific examples to augment the work tasks of human through assistants, helpers and analysts. These examples are more Prediction-based, while we believe that ‘Reasoning’ is the core part of AI Agents being fully agentic.
Intelligent personal assistants:
AI-powered personal assistants like Siri, Alexa, and Google Assistant are now essential in daily life, understanding our needs, managing tasks, and personalising their assistance through learning from our interactions.
Autonomous robots:
Robotic helpers are transforming physical tasks, from household chores like smart vacuuming to industrial work in Amazon warehouses, using advanced sensors and AI to operate independently with minimal human assistance.
Gaming Agents:
AI Agents in gaming have evolved from simple opponents to advanced entities that can rival or beat human players, like Deep Blue in chess and AlphaGo in Go, using deep learning and strategic analysis.
Fraud detection Agents:
AI Agents help prevent financial fraud by analysing transactions for unusual patterns, allowing banks and credit card companies to detect and address suspicious activities more effectively.
(Source: AI Agents: Types, Benefits and Examples, n.d.)
Recent Developments
OpenAI has recently introduced a new five-tier system to track its progress toward artificial general intelligence (AGI) – when AI is considered to have capabilities that rival humans. OpenAI currently places its technology at Level 1, but nearing Level 2. Level 1 is conversational AI, where computers can interact in conversational language with humans. Their progress towards Level 2 has been demonstrated an internal GPT-4 research project that shows human-like Reasoning and with the recent addition of OpenAI o1, which “thinks” a long internal chain of “thought” before it answers. Level 2 AI can perform basic problem-solving tasks on par with a PhD-level of human without tools, with Level 3 rising to Agents that can take autonomous actions for users (OpenAI Says There Are 5 “Levels” for AI to Reach Human Intelligence — It’s Already Almost at Level 2, 2024).
Challenges and Considerations
Current iterations of Agents, such as LLM Agents, aim to address some limitations of large language models (LLMs) by connecting them to tools, systems, workflows, and user contexts that mitigate limitations in planning, memory and action. However, these Agents cannot think independently and are limited to predicting the next token in a string. They are prone to hallucinations and may provide inaccurate information when they lack the answer. LLM Agents cannot remember previous exchanges or build upon past conversations, and their capabilities are confined to their pre-trained data (Cognitive Architectures Explained for Non-Developers, n.d.). The hallucinations and other challenges of LLMs and LLM Agents have resulted in some high-profile failures because all LLM outputs are treated the same. Complexio avoids these issues with Foundational AI that augments and grounds LLM outputs in whole company data. Furthermore, Complexio aims to enhance the Reasoning abilities of current Agents to enable them to work more effectively and efficiently.
Complexio’s Technological Foundation for Automation & Augmentation
Complexio accomplishes Human-Machine Teaming through a combination of Foundational AI, the development of Predicting and Reasoning, and a Natural User Interface. Foundational AI provides a central AI-driven business intelligence layer for automation, orchestration, interaction and insights into Whole Company Data (WCD) that leads to more informed and cohesive decision-making.
“…autonomous systems have a built-in learning model and feedback loop and are told what to look at.”
Matthew, n.d.
Foundational AI (FAI) is a unified, integrated system that enhances the entire organisation’s performance. It takes WCD and makes sense of it by mapping the relationships in the data to build a model of each unique enterprise that deeply understands what business activities they carry out and how it currently does them.
The Natural User Interface accesses Complexio’s unique and holistic enterprise knowledge and insights with a discrete integration into the existing user workflows. Its aim is to be as non-disruptive and efficient as possible: providing automatic, contextual, multi-step assistance that automates complex recurring tasks, with minimal need for user interaction.
Prediction is key to Complexio’s approach to automation and augmentation – predicting the next action or task to be done to offer stronger and stronger productivity. This proactive strategy leverages Human-Machine Teaming by continually comparing Predictions to actual observed user behaviour in a continuous learning loop that enables Complexio to further refine its Predictions over time to achieve even greater efficiency and accuracy.
Reasoning further deepens this automation capacity to understand what is useful across all Predictions. Over time and application it builds an understanding of how the system can begin to automate across intents, and at heightened scale.
With the technological foundation Complexio creates, all business activities and decision-making are unified and grounded in AI. By building a connection between real-time data, AI technologies and a foundational model of how each unique enterprise operates, Complexio improves workflows and efficiency, continually adapting as user behaviour and business requirements change.
Complexio’s Roadmap for Automation & Augmentation
“With Complexio we can significantly reduce the manual load our employees carry by automating the processes around routine tasks.”
Mikael Skov, 2024
Complexio’s roadmap consists of two phases. The first phase involves Building the Foundational AI (FAI) Model by Ingesting and Mapping WCD, which enables the immediate ability to automate various existing and pre-defined use cases. Integrating this data into the existing company infrastructure creates a central hub that facilitates collaborative interaction and decision-making across the organization, using tools like Neo4j and Kineviz to gain insights. This phase not only accelerates the adaptation time for Human-Machine Teaming, offering a competitive edge in the market, but also focuses on augmenting knowledge workers by establishing a data foundation for Prediction capabilities.
Phase two is Leveraging the Model. This is focused on diving deeper into Prediction capabilities, with an end goal of automated Reasoning across current and future business outcomes. During this phase, LLMs and other AI models will analyse the data they have ingested and mapped to determine the next steps. The approach provides compounding automation gains across business activities, offering more and more assistance and augmentation as it learns from user interactions. The facilitation of automated Reasoning for potential future business outcomes also unlocks deeper value extraction from both existing and new AI tools and systems connected to a unified company-wide data source. Additionally, it offers the capability to automatically assess and integrate new tools seamlessly.
Building for an AI-First World
By building a foundation that enables enterprises to continually work with and adapt to AI technologies, we are creating a future where humans and machines can collaborate seamlessly. Humans can concentrate on the creative aspects of work, freed from the roles of data mover and paper pusher. Meanwhile, computers handle complex, recurring tasks, streamlining processes and eliminating inefficiencies. This teamwork combines the accuracy and speed of the machines with the creativity and intelligence of humans, making work more efficient and valuable. The unique combination of Foundational AI, a Natural User Interface, Whole Company Data, and predicting and Reasoning capabilities will enable us to achieve remarkable results with Human-Machine Teaming.
“Autonomy is about automating workflows while also removing the human element from the workflow. Humans aren’t particularly good at doing repetitive tasks… the future is automated in most of our industry verticals.
AI is not a technology that you can buy in a box and then just deploy, as it is inherently stupid. Where AI plays a role is in automation where humans take decisions based on data input. If that human tells the algorithm ‘this is right, this is wrong’ enough times, then that algorithm will eventually work as well as, or better, than a human.
Autonomy is the only way we will achieve sustainability at the scale needed to reverse current trends. And it won’t happen unless businesses can make it happen profitably. Sustainability and profitability must thrive together.”Ola Rollen, 2022 [10]
Conclusion
Human-Machine Teaming is a powerful way to improve productivity by combining the strengths of both humans and machines. By automating complex recurring tasks and improving decision-making with AI, businesses can become more efficient, innovative, and strategic. Complexio’s advanced technology, which includes Foundational AI, Prediction and Reasoning, and a Natural User Interface provides a strong framework for this collaboration. Side-stepping the limitations of existing automation tools and LLMs, the technology integrates seamlessly into a company’s existing systems, ingesting Whole Company Data and understanding how the company operates – from the inside-out. It grounds AI-driven analysis and decision-making in unique, holistic data that does not discriminate across department or team, rather, provides a unified system that boosts the entire organisation’s performance at every step. As we advance these technologies, the partnership between humans and machines can only get stronger from this base – driving even greater business success and sustainable, profitable and innovative growth.
References
15. Tank, Aytekin (Aug 2024). “Top 8 business automation trends for 2024.” Jotform. Accessible at: https://www.jotform.com/blog/automation-trends/
16. van der Aalst, Wil M.P. (2021) “Hybrid Intelligence: to automate or not to automate, that is the question,” International Journal of Information Systems and Project Management: Vol. 9: No. 2, Article 2. Accessible at: http://ijispm.sciencesphere.org/archive/ijispm-0902.pdf