In the rapidly evolving landscape of business, the collaboration between humans and artificial intelligence (AI) has emerged as a powerful force that drives the formulation of the most effective business strategies. The fusion of human creativity, intuition, and critical thinking with the analytical capabilities and computational power of AI presents a unique opportunity to unlock new realms of success and innovation. Gone are the days of viewing AI as a replacement for human involvement; instead, it has become a catalyst for empowering human potential and decision-making.
In this blog, we will delve into the dynamic relationship between humans and AI, exploring how they collaborate harmoniously to shape the best business strategies. From harnessing AI-driven insights and predictive analytics to leveraging human expertise in interpreting results and making informed decisions, we will uncover the synergy that occurs when these two forces converge.
Human AI Collaboration For Smooth Business Processes
This article discusses the collaboration between humans and machines in order to achieve the full potential of artificial intelligence (AI) and enhance business processes. It outlines three key roles for humans in assisting machines: training, explaining, and sustaining. Humans train machine-learning algorithms to perform specific tasks and teach AI systems how to interact with humans. They also explain AI outcomes, especially in fields where transparency is crucial, such as law and medicine. Additionally, humans play a role in sustaining AI systems by ensuring their proper functioning, safety, and ethical use.
Humans Aiding Machines
Humans have three vital responsibilities. Firstly, they must educate machines to carry out specific tasks. Secondly, they need to elucidate the results of those tasks, particularly when they are unexpected or contentious. Lastly, humans must ensure the responsible utilization of machines, such as by averting any harm to humans caused by robots.
Training
Training is crucial for machine learning algorithms to perform their designated tasks. This involves accumulating extensive datasets to educate apps like machine translation, medical diagnosis, and recommendation engines. Effective interaction with humans is another aspect requiring training for AI systems. Leading tech companies and research groups have already established proficient training teams and expertise.
Take Microsoft’s AI assistant, Cortana, for example. Developing the right personality for Cortana required extensive training, with a team comprising a poet, novelist, and playwright investing numerous hours. Apple’s Siri and Amazon’s Alexa also underwent similar training to align with their respective brand images. Siri, in particular, was imbued with a touch of sassiness, consistent with Apple’s consumer expectations.
Recent advancements aim to train AI assistants to display more complex human traits, such as empathy. Start-up Koko, a spin-off from MIT Media Lab, has developed technology to enable AI assistants to express empathy. Rather than offering generic responses, Koko’s system engages users by seeking more information and providing advice to reframe their issues. For instance, Koko may suggest viewing stress as a positive emotion that can be channeled into action.
Explaining
With the increasing use of opaque processes in AI systems (known as the black-box problem), it has become essential to have human experts who can explain the behavior of these systems to non-expert users. These “explainers” play a crucial role in industries that rely on evidence-based practices, like law and medicine, where understanding how an AI algorithm weighed inputs for decisions such as sentencing or medical recommendations is crucial. Explainers are also important in fields such as insurance and law enforcement, helping them comprehend why an autonomous car caused or failed to prevent an accident. In regulated industries and any consumer-facing sector where machine-generated outcomes could be challenged as unfair, illegal, or incorrect, explainers have become integral. For instance, the European Union’s General Data Protection Regulation (GDPR) grants consumers the right to receive explanations for algorithm-based decisions, such as credit card or mortgage rate offers. This aspect of AI will contribute to increased employment, with an estimated 75,000 new jobs required to meet GDPR requirements.
Sustainers
Companies not only require individuals capable of explaining AI outcomes but also need “sustainers” who work continuously to ensure that AI systems operate effectively, safely, and responsibly.
AI has the potential to enhance our analytical and decision-making capabilities while fostering creativity. For instance, a group of experts known as safety engineers focus on anticipating and preventing harm caused by AI systems. Developers of industrial robots, for example, take great care to ensure that these robots can identify nearby humans and avoid endangering them. In cases where AI systems do cause harm, such as a self-driving car being involved in a fatal accident, these experts may also assess the analysis provided by explainers.
Another category of sustainers focuses on upholding ethical standards in AI systems. Ethics managers investigate and address instances where an AI system, like one used for credit approval, exhibits discriminatory behavior towards specific groups. Similarly, data compliance officers ensure that the data used by AI systems complies with regulations like the GDPR, safeguarding consumer protection. Additionally, there are roles dedicated to responsible data management, such as Apple’s “differential privacy team.” Their objective is to balance the AI’s quest for insights about a group of users while preserving the privacy of individuals, as unrestricted data gathering can compromise privacy, upset customers, and violate the law.
Machines Helping Humans
Smart machines are playing a significant role in augmenting human capabilities in three distinct ways. Firstly, they can enhance our cognitive strengths, effectively amplifying our mental abilities. Secondly, they have the capability to interact with customers and employees, thereby freeing humans to engage in more complex and advanced tasks. Lastly, these machines possess the ability to embody human skills, enabling us to extend our physical capabilities beyond what is naturally possible.
Amplifying
By providing relevant information in a timely manner, artificial intelligence (AI) has the potential to enhance our analytical and decision-making capabilities. However, AI can also elevate creativity. An example of this is Autodesk’s Dreamcatcher AI, which stimulates the imagination of designers, even those who are exceptionally skilled. The designer provides Dreamcatcher with specific criteria for the desired product, such as a chair capable of supporting 300 pounds, with an 18-inch seat height, and using materials costing less than $75, among other specifications. Additionally, the designer can share examples of chairs she finds appealing. Dreamcatcher then generates thousands of designs that align with the given criteria, often inspiring ideas that the designer may not have initially considered. The designer can guide the software by expressing preferences or dislikes, leading to a new round of designs.
Throughout this iterative process, Dreamcatcher handles the complex calculations necessary to ensure that each proposed design meets the specified criteria. This frees the designer to focus on utilizing their distinct human strengths: professional judgment and aesthetic sensibilities.
Interacting
Collaboration between humans and machines opens up new and more efficient avenues for companies to interact with employees and customers. AI agents like Cortana, for instance, can facilitate communication between individuals or on behalf of individuals. For example, they can transcribe meetings and distribute searchable versions to those unable to attend. Such applications have inherent scalability, as a single chatbot can simultaneously provide routine customer service to a large number of people, regardless of their location.
SEB, a prominent Swedish bank, utilizes a virtual assistant named Aida to engage with millions of customers. Aida, capable of conducting natural-language conversations, has access to extensive data repositories and can address common queries, such as account opening or cross-border payment instructions. Aida can also ask follow-up questions to resolve customer issues and analyze the caller’s tone of voice to provide better service in the future based on their emotional state (frustrated or appreciative, for example). In cases where the system cannot resolve an issue (about 30% of cases), the caller is transferred to a human customer service representative, and the interaction is monitored to learn how to address similar problems in the future. By having Aida handle routine requests, human representatives can focus on tackling more complex issues, particularly those from dissatisfied callers who may require additional support.
Embodying
AI systems like Aida and Cortana primarily exist as digital entities, but in other instances, intelligence is embodied in physical robots that enhance human workers. These AI-enabled machines are equipped with advanced sensors, motors, and actuators, enabling them to recognize people and objects and operate safely alongside humans in various settings like factories, warehouses, and laboratories.
In the realm of manufacturing, robots are evolving from potentially hazardous and unintelligent industrial machines into smart “cobots” that are aware of their surroundings. For instance, a cobot arm may handle repetitive tasks involving heavy lifting, while a human worker focuses on complementary responsibilities that require dexterity and human judgment, such as assembling a gear motor.
Hyundai is taking the cobot concept further by incorporating exoskeletons. These wearable robotic devices adapt to the user and their environment in real time, granting industrial workers superhuman endurance and strength, thereby empowering them to perform their jobs more effectively.
| Humans Help AI | AI Helps Humans |
| Train AI models and algorithms | Enhance decision-making through data analysis |
| Provide labeled training data for AI learning | Improve efficiency and productivity |
| Interpret and explain AI outcomes | Automate repetitive tasks |
| Collaborate with AI systems to improve processes | Augment human capabilities and extend physical abilities |
| Define ethical guidelines for AI systems | Increase innovation and creativity |
| Ensure responsible use of AI technology | Identify patterns and trends in large datasets |
| Adapt AI systems to changing business needs | Enhance customer experiences through personalization |
| Apply domain expertise to fine-tune AI algorithms | Streamline decision-making with real-time insights |
| Design AI applications to align with business objectives | Enable predictive capabilities and proactive problem-solving |
| Identify new opportunities for AI implementation | Optimize resource allocation and cost management |
Reimagining Your Business
To fully harness the value of AI, companies must undertake a redesign of their operations. This involves identifying areas within the business that can be enhanced, whether it’s improving internal processes or tackling previously unsolvable problems through the application of AI. Advanced AI and analytic techniques can help uncover hidden issues that were previously invisible but can now be addressed with AI solutions.
In order to reveal these unseen problems, companies are using AI to discover unknown unknowns within their operations. For instance, GNS Healthcare employs machine learning software to identify overlooked relationships within patient health records and other data sources. By generating hypotheses and determining the most likely explanations for these relationships, GNS was able to uncover a previously unidentified drug interaction hidden in unstructured patient notes. This approach goes beyond traditional data mining and focuses on discovering causal links.
Next, companies must engage in co-creation to develop solutions. This involves collaborating with stakeholders to envision how they can work together with AI systems to improve processes. For example, a large agricultural company initially aimed to use AI to predict crop yields more accurately but discovered, through conversations with farmers, that real-time recommendations for increasing productivity were a more pressing need. An AI system was developed to provide such advice, and initial results were used to refine the algorithms. New AI and analytic techniques can assist in the co-creation process by suggesting innovative approaches to process improvement.
Finally, once a solution is developed, companies need to scale and sustain it. For example, SEB initially deployed their chatbot Aida internally to assist employees and later extended its use to serve one million customers.
When reimagining a business process, it is important to consider the desired transformation in terms of characteristics such as flexibility, speed, scale, decision-making, and personalization. Companies must determine which characteristic is central to the transformation, how intelligent collaboration with AI can address it, and what trade-offs and alignments with other process characteristics may be required.
Flexibility
Mercedes-Benz faced a growing challenge due to inflexible processes that couldn’t meet the increasing demand for individualized S-class sedans from their most profitable customers. The traditional approach to car manufacturing relied on rigid processes executed by “dumb” robots. To enhance flexibility, Mercedes-Benz introduced AI-enabled cobots, replacing some of the robots, and redesigned their processes to foster collaboration between humans and machines. In their plant near Stuttgart, Germany, cobot arms, guided by human workers, handle the lifting and placement of heavy parts, effectively becoming an extension of the worker’s capabilities. This system allows the worker to have greater control over the car’s assembly, shifting from manual labor to a more supervisory role alongside the robot.
The human-machine teams at Mercedes-Benz demonstrate remarkable adaptability. The cobots can be easily reprogrammed using a tablet, enabling them to handle different tasks based on changes in the workflow. This agility has enabled Mercedes-Benz to achieve unprecedented levels of customization. They can personalize vehicle production in real-time, aligning with the choices made by consumers at dealerships. Everything from the dashboard components to the seat leather to the tire valve caps can be customized. As a result, each car that rolls off the assembly line at the Stuttgart plant is unique, ensuring a high degree of individualization.
Speed
In certain business operations, speed is of paramount importance, such as in the detection of credit card fraud. Companies have only seconds to determine whether a transaction should be approved, as a fraudulent transaction could result in financial losses. HSBC, like many other major banks, has implemented an AI-based solution to enhance the speed and accuracy of fraud detection. The AI system analyzes millions of transactions daily, utilizing data on purchase location, customer behavior, IP addresses, and other factors to identify subtle patterns that may indicate potential fraud. Initially implemented in the United States, this system significantly reduced undetected fraud and false positives. HSBC then expanded its deployment to the UK and Asia. Danske Bank also achieved notable improvements in fraud detection rates and false positives using a different AI system, enhancing efficiency by enabling investigators to focus on flagged transactions requiring human judgment.
The fight against financial fraud is akin to an ongoing arms race, with better detection prompting more sophisticated tactics from criminals, leading to the need for further improvements in detection methods. Consequently, the algorithms and scoring models used in fraud prevention have a short lifespan and require continuous updating. Furthermore, different countries and regions employ distinct models. Therefore, a substantial workforce of data analysts, IT professionals, and financial fraud experts is necessary at the intersection of humans and machines to ensure that the software stays one step ahead of criminals.
Scale
In many business processes, poor scalability poses a significant challenge to improvement, especially when the process heavily relies on labor-intensive human efforts with minimal machine support. Unilever’s employee recruitment process serves as an example. The consumer goods company aimed to diversify its 170,000-person workforce, with a focus on entry-level hires who could be fast-tracked into management positions. However, the existing processes were unable to evaluate a sufficient number of applicants while providing individual attention, making it difficult to ensure a diverse pool of exceptional talent.
To address this, Unilever combined human and AI capabilities to scale personalized hiring. In the initial round of the application process, candidates engage in online games designed to assess traits like risk aversion. These games do not have right or wrong answers but help Unilever’s AI system identify individuals who may be well-suited for specific positions. In the subsequent round, applicants submit videos where they respond to position-specific questions. An AI system analyzes their responses, considering not only the content but also body language and tone. The AI identifies the top candidates from this round, who are then invited for in-person interviews, where humans make the final hiring decisions.
While it is too early to evaluate the impact of the new recruiting process on employee quality, Unilever has closely monitored the success of these hires and continues to gather more data. However, it is evident that the new system has significantly expanded the scale of Unilever’s recruiting efforts. Accessible via smartphones, the number of applicants doubled to 30,000 within a year, the representation of universities increased from 840 to 2,600, and the socioeconomic diversity of new hires improved. Additionally, the average time from application to hiring decision decreased from four months to just four weeks, and recruiters’ application review time decreased by 75%.
Decision Making
AI has the potential to enhance decision-making by providing employees with personalized information and guidance. This is particularly valuable for workers in critical operational roles where making the right decisions can significantly impact the company’s bottom line.
An example of this is the improvement in equipment maintenance through the use of “digital twins” – virtual models of physical equipment. General Electric (GE) creates software models of its turbines and other industrial products, continuously updating them with real-time operating data. By collecting data from a large number of machines in the field, GE has acquired a wealth of information on normal and abnormal performance. Their Predix application, powered by machine-learning algorithms, can now predict potential failures of specific parts in individual machines.
This technology has transformed the decision-intensive process of maintaining industrial equipment. For instance, Predix can identify unexpected wear and tear in a turbine’s rotor, analyze its operational history, report the increased damage over time, and provide a warning that without action, the rotor may lose a significant portion of its remaining useful life. The system can then suggest appropriate actions, taking into account the machine’s current condition, the operating environment, and aggregated data on similar damage and repairs. Predix also generates information on the costs, financial benefits, and includes a confidence level (e.g., 95%) for the assumptions made in its analysis.
Without Predix, workers would likely only detect rotor damage during routine maintenance checks, possibly leading to a costly shutdown if the damage goes unnoticed. With Predix, maintenance workers are alerted to potential issues before they escalate, equipped with the necessary information to make informed decisions that can save GE millions of dollars in some cases.
Wrapping Up with Conclusion
The majority of activities at the interface between humans and machines require individuals to engage in new and different tasks (e.g., training a chatbot) and to approach familiar tasks in a different manner (e.g., utilizing the chatbot to deliver improved customer service). However, in our surveys, we have observed that only a small fraction of companies have begun to envision and optimize their business processes for collaborative intelligence. It is evident that organizations focusing solely on automating tasks to replace workers will fail to harness the full potential of AI. Such a strategy is fundamentally flawed. The future’s successful leaders will be those who embrace collaborative intelligence, transforming not only their operations, markets, and industries but also their workforces. This paradigm shift is crucial for unlocking the true benefits of AI and capitalizing on its transformative power.