Interview with Nav Deol MBA, Advisor, Massachusetts Institute Of Technology / Westgate

1. For TechBullion readers, how does your work in investing, shape your perspective on embedding AI into corporate strategy?

From an investor’s point of view, the question of how to integrate AI into a company’s strategy comes down to impact, and not hype.

I think about three things: unit economics, defensibility, and execution. Some of the questions I focus on are, does AI drive revenue, cost, or risk in a clear way? And does it drive those things in a way that’s hard to replicate, because it’s built on proprietary data and embedded in core business processes? These are the criteria I focus on when embedding AI into corporate strategy.

Ultimately, great companies start with a crystal-clear use case that drives ROI, and then scale. AI is not the business; it’s a multiplier on a business that’s already excellent.

2. What experiences across private equity, early-stage startups, and corporate development most influenced how you guide companies on AI today?

Private equity has helped to create a strict ROI filter: AI has to translate into real value, whether that’s top-line growth, cost savings, or better risk profiles.

Startups, particularly early-stage, have helped to create a culture of speed and iteration. The goal isn’t to make something perfect; it’s to find a small, high-leverage use case, prove it out, and scale up on the basis of real-world evidence rather than assumptions.

Corporate development has helped to reinforce the importance of integration. Even the best AI projects can fail if they don’t integrate into the business.

So, combining these influences, I work with companies to remind them to stay humble: to focus on high-leverage use cases, to prove out these cases quickly, and to integrate the AI into the business at a level that actually drives business outcomes, rather than just creating incremental efficiencies.

3. When you first assess a company or portfolio asset, what diagnostic do you use to separate high-signal AI opportunities from distractions?

My simple framework to distinguish between signal and noise is:

1. Economic impact. Look at where AI has the potential to impact the P&L: revenue, costs, or risk. If there isn’t a direct impact, it’s probably noise.

2. Data advantage. Look at whether the company has proprietary, high-quality data on this particular problem. If the same problem can be solved by any company using any data, it’s probably not worth doing.

3. Workflow integration. Look at whether the output of the AI is actually integrated into real-world decisions. If the AI is simply used as a side project, as a curiosity, or as something to put on the company’s internal dashboard, it’s probably not worth doing.

If the answer to these questions is yes, then it’s probably worth doing; if the answer to any of these questions is no, then it’s probably noise.

4. Once you have a shortlist of use cases, what framework do you use to prioritize them across revenue growth, cost efficiency, and risk reduction?

Once we have a shortlist, my approach is to use a simple scoring framework across three dimensions:

1. Value size (P&L impact). What is the upside in revenue, cost, or risk, and how much is it really worth? Quantify it. The bigger, more clear-cut dollars should go to the top of the list.

2. Time-to-value. How long does it take for this idea to be deployed and start delivering value? Quick wins (30-90 days) should take priority because they create momentum, which helps fund other, longer-term bets.

3. Feasibility. Do we have the data, infrastructure, and people in place to deliver this idea, or is it a heavy dependency? High-impact ideas should be deprioritized if they are heavy dependencies, unless they are strategically important.

4. Scalability. Will this idea have legs across other areas of the business once it is proven out? The best use cases are those that are additive. One win should turn into many wins.

In other words, my goal is to look for high-impact, fast execution, and scalability. This combination of characteristics beats big, complex initiatives, which take too long to deliver.

5. What trends do you see emerging in the AI space and where do you see the industry heading?

A few things that are becoming clear are:

1. Automation instead of assistance. AI is moving from copilots to end-to-end workflow execution, where the real benefits materialize.

2. Vertical AI wins. Generic tools are becoming commodities; the advantage is going to industry-vertical solutions with proprietary data.

3. Data is the moat. The moat is no longer the model; it’s the data and the ability to continually improve performance.

4. Embedded AI. AI is becoming part of the system (CRM, ERP), quietly driving decisions instead of being a standalone product.

AI is becoming a part of the infrastructure for companies that embed it well; the benefits will compound for those that do; the rest will fall behind.

6. Which AI use cases are actually generating measurable revenue or operational efficiency, and which are mostly hype?

The most quantifiable AI projects are those that are closely tied to business performance:

1. Revenue: dynamic pricing, recommendation systems, and marketing efforts that increase sales.

2. Cost savings: automations in finance, supply chain, and customer support that reduce personnel costs.

3. Risk mitigation: fraud detection, credit scoring, and predictive maintenance that improve accuracy.

Mostly hype are projects where the outcome is fuzzy or difficult to measure: generic “AI strategy” initiatives, experimentation without key performance indicators, or tools that are simply adjuncts to the business with no real impact on operations.

It’s simple to see that if the AI projects aren’t directly tied to business decisions that impact the top line and bottom line, they are likely noise.

7. As an investor and advisor, how do you underwrite ROI for AI initiatives in financial models and deal theses?

As the AI ROI is underwritten, it is done so in a similar manner to other deals; however, it should be even more sensitive to the operational drivers and the risks.

First and foremost, the potential value, which could be revenue growth, cost savings, etc., being driven by this particular AI solution should be analyzed.

Secondly, the cost should be analyzed; this includes the cost of software, infrastructure, etc.

Thirdly, the time to value should be analyzed; this is very important, as the early wins will help offset the cost.

Fourthly, the ROI should be stress-tested for execution risks; this includes workflow risks, data risks, etc.

Lastly, once one is satisfied that the ROI is robust enough to offset all the risks, it is then time to consider whether it is worthy of inclusion in a specific financial model.

It is imperative to understand the difference between AI that is a true value multiplier and AI that is merely a flashy, expensive, and unprofitable solution.