AI educator Kavita Ganesan on uncovering AI opportunities in your business

Over the past year, rapid advancements in generative AI, including the groundbreaking launch of ChatGPT, have brought AI to the forefront of everyone’s attention. However, navigating the AI landscape can be intimidating for business leaders who are unsure of where to begin. The transition can feel overwhelming – from choosing the right problems for AI to solve, to building a robust data infrastructure and preparing teams for the change. This is where Kavita Ganesan comes in.

Kavita is an AI advisor, educator, and founder of the consulting business Opinosis Analytics. With a Ph.D. in Natural Language Processing (NLP), Search Technologies, and Machine Learning and over 15 years of experience, Kavita works with organizations to help them demystify AI and implement it into their business strategies. In the spring of last year – curiously enough, a few months before all the buzz started – she published The Business Case for AI, a practical guide for business leaders to launch AI initiatives that drive results.

In it, Kavita outlines a framework for identifying high-impact AI opportunities, emphasizing the importance of effectively assessing and framing problems to prioritize the implementation of AI solutions that are aligned with your business goals, as well measuring the impact and success of each AI initiative.

In today’s episode, we caught up with Kavita to talk about strategies for business leaders to seize the transformative potential of AI.

Here are some of the key takeaways:

• Start incorporating AI in your business by optimizing repetitive manual processes and addressing inefficiencies identified through customer feedback or other business units.
• To identify high-impact opportunities, assess where it makes sense to deploy AI, and see if they translate to tangible business gains.
• Before implementation, you’ll need to frame those opportunities to better articulate the benefits, the pain points you’re addressing, and which metrics will allow you to measure it.
• The next step is to call in experts to make sure it’s feasible. Only then can you rank all those initiatives and prioritize the most beneficial ones.
• Success in AI initiatives relies on three pillars: model performance, business impact, and user satisfaction.

If you enjoy our discussion, check out more episodes of our podcast. You can follow on Apple Podcasts, Spotify, YouTube or grab the RSS feed in your player of choice. What follows is a lightly edited transcript of the episode.

The AI bug

Liam Geraghty: Hello, and welcome to Inside Intercom; I’m Liam Geraghty. On today’s show, I’m joined by Kavita Ganesan, the author of The Business Case for AI: A Leader’s Guide to AI Strategies, Best Practices & Real-World Applications. And that’s exactly what we’re going to be talking about today. Kavita, you’re very welcome to the show.

Kavita Ganesan: Liam, thank you for having me. I’m really glad to be here.

Liam: I know you’ve delivered more than two dozen successful AI initiatives for a wide range of organizations – from mid-size to Fortune 500s. How did you get involved in the AI space in the first place?

“I became a software engineer, but I felt that something was missing – the whole algorithm development and problem-solving piece”

Kavita: My history with AI goes back to 2005 when AI was not really popular, nor was it even sexy or needed in the industry. I got intrigued by the problem-solving aspects of AI – even though the techniques may be the same, when applied to a different problem, the way you would solve it poses different challenges. That was appealing to me because I think, inherently, I’m a problem-solver. So I got deeper and deeper into AI in my master’s program. That’s where I got exposed to the whole AI space.

I became a software engineer, but I felt that something was missing – the whole algorithm development and problem-solving piece. That’s when I decided I needed to get a Ph.D. in AI because I wanted to specialize in this. And as I was about to graduate, in 2013-ish, data science started to take off as a field. That’s when I decided that instead of going to academic institutions or research labs, I would just go and solve industry problems. I think I’m a very practical, applied person, so I wanted to see these algorithms being put to good use. That’s where things really started. I delivered all these projects and worked on different problems from healthcare to other areas, like code.

“Generative AI has kind of put AI on the map for them”

Liam: It must feel like the rest of the world has just caught up with all this AI stuff in the last couple of months.

Kavita: Yeah, for many businesses, AI is a very new thing, especially for small businesses that have not been thinking about AI because they felt it was not relevant to them. Mid-size operations have been thinking about AI for a while, but didn’t know how to get started, and generative AI has kind of put AI on the map for them.

Where to start?

Liam: Let’s dive into your book, The Business Case for AI. Straight off the bat, you acknowledge the worries and concerns that leaders have around AI. We just released our report on the State of AI in Customer Service 2023, where we surveyed 1,000 support professionals, and found that 69% of leaders are planning to invest more in AI in the coming year. But so far, only 38% of leaders have already done so. That’s got to be a huge opportunity for early adopters to gain a real competitive advantage with all the benefits that AI brings, right?

Kavita: Yeah, that’s absolutely right. When applied to the right problems, you’re going to see significant benefits very early on. I think the challenge companies are facing now is finding the right problems within the business and applying AI in a way that’s going to give them value – not six or seven months down the road, but three months.

“It’s about understanding the space of what makes up AI, where you can apply AI, what type of problems you can apply it to, and where generative AI adds value”

Liam: What would you say to people on how to frame their AI thinking so that they’re not imagining robots taking over the world, but thinking about AI, as you say, as a practical tool for business?

Kavita: I think the first step is to understand what this beast is. Now, people think AI is generative AI, but generative AI is just one piece of that AI puzzle. There’s a lot more to AI. There’s traditional machine learning, NLP, computer vision. It’s about understanding the space of what makes up AI, where you can apply AI, what type of problems you can apply it to, and where generative AI adds value. Addressing that elephant in the room will help set the context or spark ideas on where you can apply AI in your business. I would say education is the first step, yeah.

Liam: If we say we’ve got over that hump and we’re onboard the AI train, so to speak, how do you figure out what AI could be used for in your company to improve existing business processes? Could you share some examples?

“Looking for existing processes that are inefficient is a good starting point”

Kavita: Sure. A lot of companies find value by starting with repetitive problems that are being manually solved. In customer service, routing a support ticket is a repetitive task, and it takes significant time for an agent to read through the ticket, determine which team to forward the ticket to, and send that preliminary data to the team so they can triage the issue. Finding those manual processes that are repetitive and require human-level thinking – that’s a key point – is where AI solutions can really make an impact in the short term because those problems are well-understood and likely have metrics you can use as a way to measure how it’s performing against the manual approach. Looking for existing processes that are inefficient is a good starting point.

Liam: You can always come up with that list, but you could also talk to your team and see what kind of blockers they have that can improve their day-to-day.

Kavita: Yes, just talking to different business units, understanding their challenges, and understanding what customer feedback they’re getting. You will detect inefficiencies and challenges even by analyzing customer feedback. Those are areas where AI can maybe help. Let’s say customers are having trouble getting the help they need because your support solution is not effective. That’ll give you a sense of, “Hey, maybe we should have a better search functionality that addresses the customer’s problems so they don’t have to go through our ticketing system.”

Spotting the right business opportunity

Liam: What advice would you give to people wanting to prepare for AI, become an AI-ready company, and put that knowledge into action?

“Framing of each opportunity will surface which ones are the most beneficial and which ones offer a marginal benefit that you can shelve for now”

Kavita: Becoming ready for AI has two parts. One is understanding where your opportunities are within your company. If you’re a mid-size operation, it’s about talking to the different business functions, understanding their challenges, and identifying and framing those opportunities. Is it in sales? Is it in HR? That will give you an idea of which area can be your competitive advantage. The second part is the foundational piece required for AI, which is getting your data infrastructure in shape. Maybe you’re not aggressively collecting data, so that needs to start, or you’re collecting data but your data stores are in silos and there’s no way for employees to access it in a holistic fashion. Identifying those gaps and combining that with the opportunities is going to give you a long-term way to get AI into the company.

Liam: It seems like ever since ChatGPT arrived on the scene, every product or business has slapped AI on the end of their names. In a sea of all of this, how do leaders find those AI opportunities? How do you weed out the ones that are not useful?

Kavita: Yeah, that happens when you find those opportunities and frame them – you are basically articulating the benefits of the opportunity and what metrics you’ll use to measure how you’re currently solving the problem. Framing of each opportunity will surface which ones are the most beneficial and which ones offer a marginal benefit that you can shelve for now. That articulation piece is very critical, and it’s step two of my “High-Impact AI Discovery Framework” (discussed in the book). First, you have an idea, or there is a potential AI opportunity. And step two is framing.

“Will introducing AI or any software automation provide a tangible benefit in that specific situation? Does it also make business sense?”

Liam: Could you talk a little bit more about the framework?

Kavita: This framework is a repeatable process for identifying high-impact AI opportunities, and it has four key steps. The first is first thinking about whether this is a promising AI opportunity. AI opportunities often solve complex decision-making problems, and that makes AI sense. But it also has to make business sense for you to go further. That’s when you look at the workload. Will introducing AI or any software automation provide a tangible benefit in that specific situation? Does it also make business sense? And then, does it have the foundational building blocks? Let’s say you’ve been doing this process manually. If it satisfies these three things, it’s a potential AI opportunity, but that by itself doesn’t mean you should go into implementation.

That’s where step two comes in, where you frame those opportunities. Essentially, you add a lot more detail to the opportunity. Articulating the benefits, the pain point you’re addressing, and what metrics you’ll use to measure it. That’s how you’ll know you’re achieving business success and data availability. But again, this doesn’t mean you go straight into implementation. You still need to ensure it’s feasible. That’s where your experts come in – step three. You’ll take it to your experts and say, “Hey, I have this opportunity. What do you think? Can it be implemented?” That’s where they’ll spot all the red flags like, “You have data, but the volume is not enough,” or, “This is too futuristic to implement right now.” That’s where they’ll put the brakes and give you more information. Once you have all that information, you can rank those initiatives and select the top initiatives, which is step four – ranking and prioritization. This is a very repeatable process, and I wanted this to be a big part of the book because I think people don’t currently have a way to do it systematically.

“The model itself is not the end. The model is a means to solving a business problem. That’s where business success comes in”

Liam: You mentioned it there, but I’d love to talk about what happens when a leader has tackled these issues and has implemented their AI strategies. What approach would you recommend to evaluate the success of their AI initiatives?

Kavita: Right now, success is fuzzy for most companies because leaders expect a financial ROI, and AI experts just want to see high-accuracy models. In my book, I talk about three pillars of success. One is model success. The model has to have a minimum acceptable performance. Otherwise, it’s not really solving the issue. If it has 50% accuracy, it’s just random. You want to ensure it’s doing the task and performing reasonably well. But the model itself is not the end. The model is a means to solving a business problem. That’s where business success comes in. And this ties in directly to your pain point. What are you looking to improve? Is it trying to analyze a support ticket? Is it trying to improve work-life balance for your employees? There are indirect ways to measure all of these. That’s what you need to be tracking for business success.

But model success and business success alone are not sufficient because, in the end, it’s the user who will be impacted. You want to also be talking to the users of the AI solution. It can be your vendors or employees – anyone consuming the AI output. You want to ask them what they think about the accuracy of the solution, ease of use, and anything that can surface problems, either in the model or in the workflow, because this can highlight adoption issues. If they don’t like the solution, they may go back to the old way of doing things. They may not want to use your AI solution although it’s accurate and achieving business success.

Human in the loop

Liam: What would you say to customer support leaders thinking of implementing AI who are a bit nervous or worried or concerned? What would you say to help them get over that?

Kavita: One theme I’ve seen amongst leaders is that AI systems are going to take over many jobs, even their own. Sadly, this is true, but I think AI systems are more likely to augment workflows than just replace jobs because we still need the quality assurance layer where humans come in. An AI system can help with customer support tickets, but what if it cannot resolve an issue? Humans need to be there. And how are AI systems learning? They learn from data. And who generates this data? Humans. We are a big part of this AI system, so we are very much in the loop for QA, data generation, and solving harder problems.

That’s one aspect. The other is to set your expectations correctly for each problem and think through the risks. If I make AI the sole decision maker in this scenario, what are the risks? Understanding the risk will help address some of the resistance to the adoption of AI in that scenario. Maybe in this scenario, it’s too risky, so you want to have humans in the loop to review what the AI has done.

Liam: I think we’re already seeing it, but with jobs that people might fear will be lost, AI is actually creating new jobs and roles – people monitoring the AI or, in our case, chatbot designers. It’s not all doom and gloom.

Kavita: Yes. I think roles may transition from doing really low-level jobs to doing high-level jobs. You’ll be more of a QA manager. So yeah, roles will eventually change.

Liam: And Kavita, what’s the one thing you’d want people to know about AI and business working together?

Kavita: AI and business, good question. In the research world, you often see one AI solution solving a problem. But in business, one AI solution may not be enough. You’ll need a hybrid solution. It can be a combination of an AI system, a rules-based system for edge cases, and maybe also humans. So, business solutions are often less elegant and more complex than research systems.

Liam: Lastly, where can people go to keep up with you and your work?

Kavita: The first place to go to will be my website, kavita-ganesan.com. That’s where you can learn about my book. It’ll also take you to my consulting page and some other podcasts I’ve done.

Liam: Perfect. Kavita, thank you so much for joining me today.

Kavita: Liam, thanks for having me.