Artificial Intelligence (AI) has rapidly evolved over the years, transforming virtually every industry and revolutionizing the way we live and work. However, despite the immense capabilities of AI, experiential or “tacit” human knowledge remains an indispensable aspect of its successful implementation. In the concept of "human in the loop," the collaboration between humans and AI systems creates a synergy that maximizes efficiency and minimizes risks. This article explores the evolution of AI technology, the critical role “humans in the loop” will play in the future of AI, and the need for seamless integration of artificial intelligence and human intelligence (AI + HI) to maximize innovation without compromising safety or human flourishing.

The Evolution of AI

AI has made significant progress since its introduction. From rule-based and expert systems to machine learning and deep learning, AI has experienced remarkable advancements. Initially, AI systems relied solely on predefined instructions to carry out specific tasks. However, these systems had a limitation—they couldn't adapt or learn from new data effectively, which made them less efficient in solving complex problems.

Fortunately, the introduction of machine learning techniques brought about a significant change in the capabilities of AI. By being exposed to vast amounts of data, algorithms were able to recognize patterns and make predictions or decisions without explicit instructions. This breakthrough opened the doors for the development of neural networks and deep learning, which enabled AI to process unstructured data and accomplish tasks like image recognition, natural language processing, and autonomous driving. 

Today, companies are caught in the “race to automate” to achieve higher levels of productivity using generative AI tools emphasizing the crucial need to combine human intelligence, which includes tacit knowledge, experience, values, and wisdom, with artificial intelligence.

The "Human in the Loop" Approach

While the advancements in AI are impressive, there are limitations to purely autonomous AI systems. Safety, ethical considerations, and the need for human oversight are significant factors that drive the "humans in the loop" approach. This approach emphasizes collaboration between humans and AI, leveraging the strengths of each to achieve optimal outcomes.

Humans provide the contextual understanding, intuition, and decision-making abilities that AI systems lack. They play a crucial role in training, fine-tuning, and monitoring AI algorithms to ensure they align with human values and adhere to ethical standards. Humans also possess essential tacit knowledge, which includes intuition, expertise, and insights gained from experience—characteristics that are crucial in complex decision-making processes.

As Michael Polanyi once noted, human knowledge often goes beyond what we can express in words. This is one reason why AI, especially deep learning, is so fascinating. Its ability to identify patterns in complex phenomena that defy straightforward descriptions enables it to understand them on our behalf. However, this can also lead us to see AI as a kind of "magic button" that reliably provides the right answers without revealing the underlying process.

Many aspects of our lives contain valuable elements that we must preserve, making it impossible to categorize automation simply as "manual" or "automatic." We should aim for a balance, automating tasks where it proves beneficial, while preserving those where human participation is valuable. It's tempting to lean toward extremes—either automating everything or nothing at all. Yet, the best solutions usually find a middle ground, finding a harmonious coexistence between automated technology and the tools we use, as well as between automation and human interaction.

Adopting a human-in-the-loop approach transforms an automation challenge into a problem of designing effective Human-Computer Interaction (HCI). Our focus must shift from "how do we create smarter systems?" to "how do we integrate meaningful and practical human interaction into the systems?"

The Critical Role of Human Tacit Knowledge

Human tacit knowledge refers to knowledge that can be difficult to articulate explicitly but is crucial for decision-making and problem-solving. Tacit knowledge can be defined as skills, ideas and experiences that are possessed by people but are not codified and may not necessarily be easily expressed. With tacit knowledge, people are not often aware of the knowledge they possess or how it can be valuable to others. In the context of AI, human tacit knowledge helps bridge the gap between the limitations of AI algorithms and the real-world complexity of tasks at hand. Studies show that in our organizations, tacit knowledge represents up to 90% of knowledge. The same study (and a myriad of others) asserts that knowledge, and notably tacit knowledge, is an organization’s most strategically significant resource. This knowledge is much too critical not to be intentional about capturing and sharing it effectively.

As AI continues to evolve, the collaboration between AI systems and human tacit knowledge will remain integral to achieving optimal outcomes. The "humans in the loop" approach harnesses the strengths of both, enabling AI to augment human intelligence while humans provide the crucial contextual understanding, ethical considerations, adaptability, and complex problem-solving skills. As we move forward, striking the right balance between AI automation and human expertise will be key to unlocking the full potential of AI while ensuring its responsible and ethical use.

Incorporating human tacit knowledge into AI system interactions brings several critical value propositions to the forefront. These value propositions are pivotal in enhancing the functionality, adoption, and overall impact of AI technologies. 

Benefits of human tacit knowledge “in the loop” of AI system interaction:

Enhanced Contextual Understanding: Humans bring a deep contextual awareness that AI systems often lack. By incorporating human insights, AI can interpret data within the context of its cultural, social, and situational relevance, leading to more accurate and meaningful outcomes.

Improved Decision-Making: Human judgement plays a crucial role in decision-making processes, especially in complex scenarios that AI might oversimplify. Humans can weigh moral implications and consider intangible factors, leading to decisions that are not only data-driven but also ethically sound and contextually appropriate.

Increased Creativity and Innovation: While AI can optimize existing processes and generate solutions based on patterns, human creativity introduces new ideas and perspectives that AI cannot replicate. This collaboration can spark innovation, leading to breakthrough solutions.

Adaptability and Flexibility: Humans can adapt to unexpected changes and anomalies in data or situations, which rigid AI algorithms may not handle well. This adaptability is crucial in dynamic environments where conditions and requirements frequently change.

Ethical Oversight and Responsibility: Embedding human tacit knowledge into AI systems ensures that ethical considerations are integrated into automated processes. Humans can oversee AI operations to prevent biases, ensure fairness, and uphold privacy and ethical standards, making AI systems more trustworthy. (it’s important to acknowledge that humans themselves are not always consistently ethically aligned. Individual biases, influenced by personal beliefs, cultural backgrounds, or societal conditioning, can inadvertently impact the decision-making process, potentially leading to ethical dilemmas or unfair outcomes - more strategies to mitigate this risk in a future article.)

Personalization: Human insights can help tailor AI outputs to individual needs and preferences, enhancing the user experience. Personalization is particularly important in sectors like healthcare, education, and customer service, where individual-specific outcomes are crucial.

Cultural Sensitivity: Humans can guide AI systems to be culturally sensitive, which is important for global applications. This includes understanding and respecting linguistic nuances, cultural norms, and regional differences, which enhances the acceptance and effectiveness of AI solutions worldwide.

Training and Feedback Loops: Humans can train AI systems more effectively by providing nuanced feedback that helps refine AI models. This iterative process, where AI learns from human corrections and suggestions, continually improves the system's accuracy and reliability.

Risk Management: Human involvement in AI systems can help identify and mitigate risks that may not be apparent through data analysis alone. This includes foreseeing potential negative impacts, addressing security concerns, and ensuring that AI applications do not cause unintended harm.

Bridging Knowledge Gaps: Human tacit knowledge helps bridge the gap between theoretical data-driven models and practical, real-world applications. This linkage is critical in translating technological advances into actionable, beneficial tools and practices.

Through thoughtful human-centric design of AI systems, organizations can harness the full potential of artificial intelligence while mitigating many of the risks associated with full automation. This balanced approach ensures that AI technologies are both effective and aligned with human values and societal needs.

Nothing illustrates what this kind of interaction looks like better than an example. At inqli, we’re designing AI-powered systems that provide everyone with access to trusted knowledge at work. I have yet to meet a leader at any organization that isn’t trying to figure out how to help their people save time looking for information. The average knowledge worker is spending over 30 percent of their time looking for information or people who can help them, and in large organizations, this costs billions annually. Generative AI systems are excellent for accessing explicit knowledge (knowledge documented in spreadsheets, reports and databases). What’s missing from these systems is the consistent “human in the loop.” inqli is designed to support the behaviours associated with effective tacit knowledge-sharing - that is the experiential, contextual and wisdom-based knowledge that’s often not recorded and can be difficult to access without being asked the “right question” AND knowing who to ask. Here’s a step-by-step example of how a scenario like this might look for an employee seeking knowledge at work.

Scenario: Enhancing Sales Strategies with both (explicit) AI Knowledge and (tacit) Human Insights

Alana is a sales leader at a large software company venturing into a new market vertical. She is preparing for a crucial meeting with a potential client who is currently considering their product alongside a competitor's offering. The client has raised several specific objections that do not align with the standard customer objection FAQs available to Alana.

Step-by-Step Process:

• Initial Research and Information Gathering: Alana begins by using the company’s generative AI system to search for any existing information on similar situations within the company’s database. This includes sales pitches, customer feedback, and previous objections encountered in related verticals. The AI compiles documents that provide some insights but nothing directly applicable to the new objections raised by the potential client.
• Detailed Question Formation: After reviewing the generalized information, Alana realizes that she needs more specific advice tailored to the unique aspects of the new vertical and the nuanced objections raised by the client. These objections are particularly challenging because they involve comparisons with a competitor’s features that the company hasn't frequently countered previously.
• Engagement with inqli for Direct Peer Advice: Since Alana doesn’t directly know who to ask for insight, she turns to inqli and inputs a detailed question about handling these specific objections. Her question focuses on competitive feature comparisons, benefits in unfamiliar industry contexts, and strategies for convincing clients who may see more apparent short-term benefits with a competitor's product. inqli's AI-powered algorithm matches Alana’s questions to colleagues who have experience in competitive sales scenarios, those who recently entered new verticals, or have exceptional problem-solving skills in sales.
• Receiving Targeted Advice and Strategy Suggestions: Several experienced colleagues respond to Alana’s inquiry. One colleague shares insights from a sales campaign where they successfully shifted the client’s focus from feature comparison to long-term ROI and support services. Another provides a detailed account of how they overcame similar objections by demonstrating the adaptability of their product to the client’s specific needs through a customized demo - and shares the link with Alana.
• Strategy Development and Role-Playing: Armed with these insights, Alana organizes a virtual meeting with her helpful colleagues to role-play the upcoming client meeting. This session allows Alana to practice handling the objections in real-time, refine her responses, and gain confidence in the strategies discussed. Alana’s colleagues learned a few things in the process too!
• Implementation and Continuous Adaptation: Confident with the new strategies and insights, Alana meets with the potential client, equipped to address each objection effectively. Following the meeting, she revisits inqli to provide updates on what worked and what could be improved, offering valuable feedback for future reference both for herself and her colleagues. 
• Feedback Loop and System Learning: The feedback provided by Alana helps enhance both the generative AI's understanding of document relevance in new verticals and inqli’s algorithm for better matching questions to the most appropriate colleagues in future scenarios.

This example highlights how a sales leader can leverage both generative AI for foundational research and a generative AI knowledge-sharing platform like inqli for actionable, experience-based advice when confronting unique challenges in new market verticals. Through this integrated approach, Alana not only addresses specific customer objections more effectively but also contributes to the collective knowledge and adaptability of the sales team.

This is one example of how and why the “human in the loop” approach to integrating AI into your workflow can better leverage the power of AI to incorporate critical tacit human knowledge and democratize access to knowledge and resources across organizations. Digital Transformation must shift focus from tech-centric solutions to the heart of what really drives change: people. To effectively navigate rapid digital transformation and maximize human flourishing, the real key is in the development of dynamic organizational cultures and mindsets that foster agility, experimentation and collaboration.

More on that in my next article...

"In this AI age where Generative AI reduces the access of explicit knowledge to a fleeting moment, mastering the art of accessing tacit knowledge becomes the pivotal challenge for enterprises. It's not just about gathering data; it's about understanding the unarticulated wisdom and insights that drive true innovation. Harnessing tools like inqli, we can now begin to cultivate and disseminate this subtle, yet invaluable, form of knowledge across the enterprise at scale, transforming tacit understanding into a tangible asset for groundbreaking progress.” - Jean-Claude Monney, Microsoft’s former Chief Knowledge Officer

To learn more about leveraging AI-powered knowledge-sharing for both human and organizational growth, download our white paper “Questions Are The Answer: Improving Business Impact and Employee Well-Being by Democratizing Access to Knowledge