Harnessing AI with Confidence

A rapid, risk-minimized approach - 

Artificial Intelligence (AI) is revolutionizing sectors across the globe. For industrial executives, the promise of AI lies in enhancing efficiency, supporting decision-making, and facilitating innovative solutions. However, the challenge often lies in how to begin integrating AI into existing processes with minimal risk. Based on our experience deploying AI tools in consulting work, I outline key principles to implement AI, specifically large language models (LLM’s), against your capabilities balancing innovation with caution.

What AI Can and Can’t Do

AI, as sophisticated as it can be, is not a substitute for human thinking but rather a supplement to it. Understanding the strengths and limitations of AI is crucial. While AI excels in language processing and structured information analysis, it doesn't replicate human-like thinking or innate understanding.

1. Language Processing vs. Thinking: AI systems can process and analyze vast amounts of text efficiently, identifying patterns and insights that might be missed by human analysts. However, they lack the human ability to comprehend context and think abstractly. For instance, AI can summarize documents swiftly but often doesn't comprehend the nuanced meaning behind the text like a human would.
2. Conventional Thinking vs. Unique Thinking: AI can replicate conventional thinking through its training on large datasets. It can anticipate typical outcomes based on past data but struggles with developing unique methodologies or strategies that require creative and critical thinking — a domain where human ingenuity remains necessary. This often results in “MBA-speak”, as MBA curriculums involve the broadest most general and validated knowledge, but doesn't delve into detailed, contextual and advanced methodologies.

For instance, we recently observed a client who experimented with AI-driven “market strategy.” While the output provided plausible recommendations, analysis of the underlying analyses showed that the AI used a segmentation approach compatible with B2C markets and failed to account of key differences in an industrial B2B setting, (which are often not addressed in conventional marketing methods).

Roles AI Performs Well:

A helpful part of finding a role for AI lies conceptualizing the human roles that AI has shown to perform well.

• Research Assistant: AI can sift through large datasets, identify relevant information, and provide a distilled summary, significantly accelerating research tasks.
• Analyst: AI is efficient in organizing and structuring unstructured data, transforming it into analyzable data sets, but but struggles with complex or nuanced analyses.
• Ideation Partner: In brainstorming sessions, AI can generate ideas based on available data, acting as a catalyst for human creativity.

Building AI Capabilities - Grass Roots

A grassroots approach is ideal for integrating AI into your operations. This involves iterative exploration and building capabilities progressively.

1. Iterative Exploration: Start with small, manageable projects that allow for testing and learning. This reduces risk and provides valuable insights into how AI can be most effectively applied within your organization.
2. Self-Funding Through Productivity Improvements: Use the productivity gains from initial AI implementations to fund further development. For example, if AI improves the efficiency of data processing, the time saved can be invested in further AI research and deployment.
3. Replacing Single Activities:
   • Identify specific, discrete cognitive tasks that AI can perform.
     Examples include converting unstructured data to structured data or transforming lengthy documents into concise summaries.
   • Daisy-chain AI Capabilities: Connect smaller, validated AI steps to create a comprehensive, efficient workflow. This method minimizes risks such as large-scale errors or "hallucinations," where AI generates incorrect or nonsensical information.

This approach allows a granular assessment of work to identify tasks for AI and those where human thinking is still required, and an environment where AI performance can be analyzed by the people most knowledgeable in the standard of work. Rather than threatening replacement of individuals, the effort provides productivity gains immediately to the practicioners involved. The approach also avoids tasking AI with “grand analyses,” where output emerges as "plausible," but often hallucinated, and is difficult to quality check.

Example: Voice of Customer (VOC)

We often engage clients to do Voice of Customer work, generating insights about markets from interviews of market participants. This is an area that because of it’s unstructured language processing historically was done by experts trained in methodologies for analyzing customer data. We currently run VOC studies where AI supplements significant portions of the work.

• Past practice: VOC tasks were labor-intensive, involving manual transcription, data extraction, and analysis, given the unstructured nature of interview data.
• Today: AI can now record and transcribe interviews, draft tailored summaries, convert responses into structured data, propose ways to structure data, and highlight insightful behaviors.
• Work remaining for staff: Holding interviews - due to it’s social nature, and developing insights - given the need for contextual understanding and strategic thinking, both still require human activity.

Actioning the Insight

Implementing AI is not just about technology but also about cultivating a culture that embraces change and innovation.

1. Start Small! We've found great progress can be made by beginning with projects that target specific tasks at the individual level. This minimizes risk while allowing for performance evaluation and fine-tuning. It also allows AI enthusiasts to being to demonstrate performance benefits from AI assistance.
2. Validate Output Quality Before Automating: Before scaling, ensure that AI outputs meet the quality and reliability standards necessary for your operations. Repetitive testing is key to successful AI integration.
3. Build Craft-Sharing Capabilities in the Organization: Foster an environment where knowledge gained from AI projects is shared across teams. This promotes a continuous learning culture and ensures that the organization evolves collectively, not just in silos.

The VOC tasks listed above are comprehensive, yet were developed rapidly and incrementally with AI-proficient practitioners simply asking themselves if the AI could do this part of the work and then experimenting.

The integration of AI into industrial company capabilities offers substantial benefits, but it is essential to approach it thoughtfully. By understanding what AI can and cannot do, initiating a grassroots implementation strategy, and validating processes before full-scale automation, executives can leverage AI with low risk while maximizing its potential.

Until next week,

Kendall -

Find me on LinkedIn or Book a 1:1 call
Not a subscriber yet? Subscribe here