Why AI Needs Human Mentors? or Evolution of a new AI role

Aug 8

5 min read

“LLMs are fast learners, but context engineering is how we teach them what matters.”

INTRODUCTION: The Shift from Prompting to Context Engineering

We’ve all had that moment—the first time we used ChatGPT or a GenAI tool and got a surprisingly insightful or creative result. It felt magical! I was also fascinated by this tool that is so multifaceted, versatile, and agile in thinking, processing and responding.

But the novelty wears off fast, real work begins. As we learnt to work effectively with these new tools limitations, scope, and relevance became key concerns. The AI tools had to learn and understand our goals, our tone, and our desired outcomes. Often, we had to rewire our approach so the tool could adapt its responses. If we reflect on the activities that we take up to train a recruit for a job, we will be struck by the same actions we have taken with GenAI tools. To get positive outcomes from training a recruit, we had to ensure that good foundational information was provided to them. Similarly, for AI tools to achieve our expectations, we had to guide the tools with information and intent as we submitted our prompts.

That’s when prompt engineering leads to something deeper—context engineering.

#RealChallenge

AI isn’t the easy button—it just hides complexity under a simple interface.” — AI Leadership Compass, Priya Sarathy the easy button—it just hides complexity under a simple interface.” — AI Leadership Compass, Priya Sarathy

AI Literacy vs. AI Fluency: Why Prompting Isn't Enough?

Many have reached literacy. Few have crossed over into fluency. The key? Understanding context—and how to shape it. AI literacy is knowing what tools like Co-pilot and M365 can do. AI fluency is about understanding how GenAI tools can create unique, context-driven outcomes for your team or business.

#RealSolutions

Morgan Stanley Wealth Management team developed a domain specific GPT tools AI@MorganStanley. They used this tool to extract, aggregate, and customize the information to support their wealth advisor team. The wealth advisors would drive the context of the information to be pulled, so that they can correctly interpret the data for their client. It can take one small error to ruin client- relationships worth millions. Managing the scope of how AI applications are developed and used required the “human mentor” and "human advisers".

In AI Leadership Compass, I present how AI literacy and AI fluency are distinctive. AI fluency is about the ability to define the use of the tool and decide how and where it can be best used. Summarized below are examples across 5 dimensions.

Conversations with AI leaders in Healthcare, Fintech, Banking, Insurance, Automotives, Pharma tech and EdTech, lead to a consensus about the slow progression of AI fluency within the organization. As MIT researcher Irving Wladawsky-Berger observed based on an IBM’s Institute for Business Value survey (2025) - "The future of work is being rewritten with AI. But many employees are unprepared for what comes next — and progress will stall if too many are left behind."

Lessons from My Journey: Prompt Engineering as a Craft

Over the past year, I have become a skilled prompt engineer, working with GenAI applications. I used GenAI tools like ChatGPT, Copilot, Canva, Claude, and Gemini through Notes LLM to supercharge my productivity, test ideas, create visuals, and even write email drafts on unfamiliar topics.

But success wasn't just about giving the right command—it was about the right context. Here’s what I learned:

Prompting is like forensics—search, deduce, refine, repeat.
LLMs adapt quickly—but they need your steering.
Prompting is sequential, but our work is not. Crossed wires (a.k.a. hallucinations) happen.
Even with a perfect prompt, AI’s outputs vary if the underlying context isn’t aligned.
While I was training the GPT, I was getting trained by GPT as well - Symptomatic of Neural cognitive adaptive learning! 😊

The Rise of Context Engineering

As I write this, I’m learning and testing Agent GPT powered by Model Control Protocol (MCP). While it reduces the need for multi-step prompting, providing the right context is still the most important factor.

Context engineering is about designing the environment in which AI operates. It's about:

Feeding history, intent, tone, examples, tools.
Aligning AI’s “mental model” to your actual goals.
Systematically managing memory and structure.

“Context engineering is the invisible foundation that makes AI feel smart, helpful, and human. It turns a chatbot into a teammate.” — Inspired by Andrej Karpathy

You already do it. When you brief a colleague. When you craft a report. When you choose what to include in a pitch. That’s context engineering in action.

The same applies to GenAI. Telling a developer to ‘write code’ isn’t enough—you need to explain the purpose, the user, and the constraints. With GenAI, the more descriptive your ask, the better the result.

As career professionals, we have often heard the advice to ‘keep up with technology’, from successful leaders. In 2025, we are adopting AI tools with greater frequency. Learning has become more formalized as organizations seek to real ROI from their AI investments.

Khan Academy’s latest Khanmigo, is shaping a more reflective and interactive learning model to inculcate AI fluency among students. The goal is to develop nurture a student’s decision-making skills, expand intellectual curiosity and bring experiential learning not passive learning.

📦 Why It Matters: Personalization, Power, and Trust

GenAI’s superpower is personalization—but only if we set it up for success. Whether in healthcare, legal reviews, or strategic planning, personalization depends on:

Relevant data
Structured memory
Domain context

What’s “fair” in a contract? It depends on the business goals, who’s involved, and what history shaped it. Following up with another observations from the introduction to Agentic Artificial Intelligence: Harnessing AI Agents to Reinvent Business, Work and Life (2025), where Tom Davenport was on of the 25 collaborators, Wladawsky-Berger calls out "(that the use of agentic AI) success depends on providing clear and precise goals and instructions".

That’s why humans must shape the context, not just the prompt.

From Prompting to Engineering: A Job Title Evolution?

If you’re crafting context to guide GenAI tools—congratulations, you’re a Context Engineer.

It’s not a job title (yet), but it is a critical skill. You’ve moved beyond prompts. You’re building systems of understanding. While in the past this would have been called subject matter expert (SME) or a domain specialist, in today’s AI using organization you are using the same domain knowledge to engineer the context of the GenAI queries.

📣 Reflections Prompt

If your organization is struggling with GenAI effectiveness, ask:

Are your teams fluent in how to shape context?
Are you teaching AI what matters—or just testing what sticks?
Are you ready to shift from prompting to engineering?

“If your people aren’t ready for AI, your business isn’t either.” – AI Leadership Compass

Let’s discuss how to make your organization more AI fluent. AI for Business Leaders is a workshop that will help your teams build AI fluency skills along with the playbook – AI Leadership Compass. Connect with me info@wheeldatastrategies.com