Directing AI Starts With Finding the Root Cause

Everyone uses AI now, so the advantage goes to the people who direct it well. Vishwadeep Khatri, founder of Benchmark X360, says that skill looks a lot like Lean Six Sigma: define the real problem, prove the cause with data, and fix it so it stays fixed. Ask an AI tool to fix a bug and it will often hand you a patch. Someone trained to find root causes keeps pushing until the real problem is solved.

In this episode, I talked with Vishwadeep Khatri, the founder of Benchmark Six Sigma, now called Benchmark X360. I've worked with Vishwadeep and his team on a few projects, including their new website, and I've always wanted to have him explain this world to people like me who never came up through the corporate quality space. His company has trained thousands of individuals, plus teams at Google, Boeing, EY, Tata Motors, Coca-Cola, and Accenture.

His path there is a good story. After his engineering MBA, he did consulting, then noticed that companies respected auditors more than consultants. So he became a third-party auditor and ran more than 400 audits across manufacturing, pharma, hospitals, software companies, and schools. As an auditor, you can ask anyone anything and they have to answer, so he learned fast. Eventually he got bored, because an auditor only says whether something conforms. He wanted to coach. Six Sigma appealed to him because it speaks the language leadership listens to, which is profit.

I asked him to explain Lean Six Sigma like we're 11 years old. The core is structured problem solving. People tend to jump to a solution before they've confirmed the problem or checked whether their data is reliable, so they treat symptoms and the issue comes back. Six Sigma follows a sequence called define, measure, analyze, improve, control. It's data driven, and projects need a senior sponsor who picks the right problem and clears the way. What I liked most is that companies don't get certified. People and projects do, and in many companies you can't get certified until a project has delivered real results.

For anyone thinking about turning their expertise into training, his business model is worth noticing. Benchmark started with open enrollment programs for individuals. Those people applied what they learned, and then their companies asked Benchmark to train whole teams. The two sides feed each other. During COVID, corporate training dried up for a while, but individual enrollment grew because people at home wanted to build their skills.

Then we got into AI, and this is where it hit home for me. Vishwadeep noticed that when he asks ChatGPT to fix something, it tends to offer a patch, and he has to tell it to find the root cause. That's my experience too. AI was trained on human work, and humans take shortcuts. Software engineers often get better results from these tools because they've been burned before and know what to ask for. My advice is to give the AI your standard operating procedures and checklists, and keep track of the work outside the chat, in files or GitHub issues, so nothing gets lost when the conversation gets long.

I also shared something I tell the AI directly: stop planning in human time. It will look at a request, decide it's a two-week project, and offer you a smaller version, when the full thing would take it a few minutes. Vishwadeep has taken the lean idea of the eight wastes, like waiting and overprocessing, and built a version for AI, because he believes we either learn to eliminate that waste or we end up working for the tool.

His newest AI program follows the same project-based approach as Six Sigma. You learn, you apply it at work, and you defend how you used AI. Then you turn what you built into something reusable, like a skill or a tool someone else can pick up. His point is that knowledge is easy to get now, and judgment is what's scarce.

My takeaway is that problem solving is one of the safest skills you can build right now, and directing AI is the next layer on top of it. If you want to dig deeper, you can find Vishwadeep's programs at benchmarkx360.com.