David’s Sling: Mastering Just Enough AI to Triumph Over Goliath Projects with Avishay Meron at AIE 2024

What comes to mind when you hear the term ai, perhaps excitement bubbles up. As you envision gaining competitive edge. Maybe determination sets in.

After all, you have the team, the drive and the vision to make things happen. For example, for some, it's a sense of pride in leading the industry forward, enhancing your company's reputation. However, AI may also evoke feelings of being anxious or overwhelmed.

Some may have concerns like, I'm not sure how to identify opportunities, or How do I build a team that can execute and deliver? And these concerns might dominate your thoughts, and if you ever felt this way, they're actually not alone. You see, according to a recent survey by Orli, these concerns are primary factors preventing from executives from fully embracing ai.

When we examine these factors, we see that the top five are strategic in nature, while the rest are tactical. And it makes a lot of sense, right? Because a solid strategy forms the foundations on which plans are built.

So for example, before you start to tackle challenges of training models, you first need to identify opportunities and relevant business use case. And I think of our role as leaders, um, is to systematically address these concerns and craft a vision and position our teams for success. In this talk, our primary focus will be on strategy.

I will introduce a thinking framework designed to assist us in developing AI strategy. You see, what makes this framework effective is that it leverage leverages existing skills that we have as leaders, such as budget allocation, people management. You know, when my kids saw me preparing for this talk, they saw this slide and they saw the, the little icon of the slingshot.

And my 6-year-old boy told me, dad, you are good at computers. Why is there a slingshot there? Are you going to shoot cans?

Can I come too? So we're not going to shoot anything, but the framework is called Sling. Before I dive into it, allow me to introduce myself.

My name is Isha Maron and I'm the founder and CEO of AIM Consulting, where my main business is to execute and build AI strategies. Over the past 10 years, I have led hundreds of AI initiatives on a global scale, working for and with different organizations. My journey is deeply rooted in technology.

I have pursued three technical degrees and authored a dozen patents in the field of artificial and intelligence. But despite my background, the key message I want to convey today is that you don't need a technical background, and it's not a prerequisite to lead successful AI initiatives. In fact, I would argue that you are likely already possessed the essential skills needed to excel.

You got this because strategic skills such as identifying opportunities, managing risk, allocating budget, um, and effective communications are far more critical to success than technical skills. Technical skills can be delegated or outsourced. And so when you think about it this way, all of a sudden, all of a sudden, AI is not that intimidating, right?

If you're looking to embrace ai, you might just need a bit of contextual knowledge and understanding of how to leverage your strategic skills in this domain. As leaders, you already possess the right skill, whether acquired through business, school, or years of practical experience. The secret lies in how to apply these skills effectively within the context of ai.

And so my goal in this talk is to leave you with two key takeaways. The first is to convince you that you can leverage your existing skills to lead AI initiatives. And the second is to show you how, how to do so by exploring Slim.

And I want to begin by few examples that will hopefully achieve the first goal that convince you that you too can succeed in ai. So the first one is consider PayPal. For example, a key factor behind PayPal's strong brand is its seamless and secure shopping experience.

This is largely enabled by a sophisticated AI fraud detection model, which has been refined over a decade and continues to evolve. Tomer Barel, the fomo, EVP at PayPal played a significant role in maintaining the company's dominance in the payment industry. Through ai, Tomer realized that identifying blocking fraud is tightly coupled with user experience and therefore has strategic significance.

And Tomer studied economics and business, not technology. Another example is Cheryl Sandberg, the former COO transformed Facebook from a cool website to a social ad king. Facebook is the best place on earth for targeted campaigns.

You can reach any segment you have in mind. For instance, 35 to 45-year-old blue collar who suffer from insomnia in Brisbane. And this, this is a real segment IKEA is interested in for selling.

You guessed it, bad products. IKEA actually narrated their catalog thinking this is so boring that it might help people who struggle sleeping at night. And the ability to reach such laser focused audience is made possible through deep understanding of our behavior on social media and building an astonishingly accurate profile for each and every one of us.

This is achieved through ai and Sandberg led it. Sandberg does not have technical background. She studied art and business.

And the last example is Stuart Butterfield, the co-founder of Slack. Stuart spearheaded the company's AI effort to improve workspace communication because of features like predictive text and smart replies. The company exhibited remarkable growth.

And like Sandberg, Butterfield didn't come with a technical background. He didn't graduate from engineering school. He studied philosophy and business.

So I hope that by now I have convinced you that strategic skills are both necessary and sufficient to lead AI initiatives. So how do we begin? Is there a specific path we can follow?

Yes. The path is sling and I wanna dive into sling through an actual project I led. So picture this, in 2016, I was giving the task to design and build a system to automatically prioritize sales leads.

Some of you are familiar with this challenge. Imagine a file containing 500,000 leads, yet only about 1% of them are valuable. This means that only one in every 100 calls a sales rep makes might result in an opportunity.

The objective was clear, uncover these hidden gems and drastically reduce the number of calls needed to identify a viable lead. It means enabling the sales teams to create more opportunity with the same effort. I was really excited about this opportunity because successful execution had the potential to leave a positive dent on the company's financial performance.

And that meant promotion for me, right? And so I assembled a dedicated task force, and after a week of intense brainstorming, we presented a plan to the steering committee. Following the presentations, one of the leaders said a sentence that at that time seemed like a typical feedback, a praise for a job well done.

But his words kept resonating in my head for some time. And the more I reflected, I realized the depth of his comment. And what he said is this, I like your simple approach and I hope we can replicate it to other projects we have in line.

Listen, this project means a lot to me, and I'm counting on you team not to p**s off Jerry. It's not just encouragement. It's a distilled version of leading AI project successfully.

Let me break it down. So the first part, I like your simple approach. It shows understanding of AI concepts.

You see, foundational knowledge of AI allows us to make informed decisions. For instance, gen AI models are probabilistic by nature, meaning they generate different outputs when we feed them with the exact same input. If you operate in a highly regulated industry, you should be aware of this inconsistency and know that there are ways to deal with it.

You don't need to know how exactly, but be aware it is possible. In my case, the team proposed a supervised method to build a regression model. The steering committee may not have grasped all the technical details, but they, they understood the general concept and appreciated its forwardness.

She solidified The basics is not about mastering every detail, but about understanding enough to make informed decision. Learning is a skill you already possess, right? And you can leverage it for your AI journey as well.

You've got this. The second part of this sentence, sentence is, and I hope we can replicate it for other projects we have in line. This part confirms that research has been conducted for identifying business processes that are ideal candidates for automation.

The logic, the logical starting point for such research is to pinpoint processes that are heavily manual. And these processes should be listed and analyzed to determine the potential impact of their automation. And when you do so, try to be specific when measuring the effects, because this is what it means to be data driven.

For instance, it is not enough to say our goal is to deploy a customer service agent to reduce the workload on our customer support team. A more impactful approach would be our objective is to deploy a customer service agent capable for au autonomously resolving, resolving 50% of tickets generated daily. And there are several advantages for quantifying expectations.

And the first one is prioritizations. It. It it allows you to determine which project hold the most value and which you should tackle first.

Second, it gives you the ability to measure success. And the third is team alignment. When expectations are quantified, it becomes easier to communicate the project goals and align the entire team on what needs to be achieved.

And this task is not new to you, right? Methodical analysis of your business processes, collecting data, conducting interviews, you are fully capable of managing these activities. You got this?

The third part is, listen, this project means a lot to me. This part shows that among various potential initiatives, this specific project is a priority because it directly affect a key business metric. Uh, previously we listed the benefits of several projects.

Now we evaluate the cost, which enables us to effectively pri prioritize them. There are three main factors you, you should consider. The first is a human capital cost.

The second is data and compute. And the third is regulation, right? And if, if you sum all these up and you subtract it from the value, then you eventually get the net value of your project.

And this is a very effective mechanism for prioritization and for the data. And compute cost. AI models requires data and training, which then consumes computational resources.

Often cloud-based. These costs change depending on the project scales. Um, as for regulation, AI project face unique challenges such as data privacy, bias, transparency, it it depends on the industry that you're working in.

And, and the, the amount of effort you need to, to, um, to invest in this really heavily depends on your industry. And at this stage, the CTO and the legal team are probably the most valuable partners you have. Engage with them to gain a comprehensive perspective on all these costs.

Now you have visibility on both outcome and the cost, and you can effectively seize opportunities. You're likely already proficient in this process of prioritization. So you got this one too.

The fourth is I'm counting on your team. This outlines the formation of a specialized team. There are three main functions you need in the team.

You need a subject matter expert with deep understanding of the business dynamics. You need technical experts and you need someone to bridge and communicate between all these functions and stakeholders. And to illustrate the importance of having an SME in the team, consider this analogy.

Imagine being tasked with designing a harvesting machine without knowing anything about the crops it should harvest. Is it on ground? Is it on a tree?

What's its size? It is possible to piece this information together through trial and error and or by analyzing data, like watching a video of a farmer doing it. But having direct access to a farmer which can answer these questions is far more effective than, than trying to infer this information.

This is the reason you want an SME in your team. Regarding the technical stuff, the specific skills of the team depends on the project, but certain skills are universally essential. So for, for instance, in most cases, solution will be deployed to the cloud, right?

Which makes cloud engineering skills crucial. Another example is if the project involves automatic analysis of images or video input, then you need expertise, expertise in computer vision. This is where consulting, the CEO or hiring consultant can fill the gap if you don't have the technical background.

Once the team is set, our role is to champion it and support it. It's our responsibility to remove any obstacle in the team's path and provide all the necessary resources. This is nurturing a team.

And as leader, you are already well equipped for this role, right? You already know how to build teams and sustain effective teams. So you got this one too.

And the last part is not to p**s off, Jerry. This part is, is about governance and effective management of expectation. Consider Jerry as a stakeholder who will eventually utilize the solution you lead.

Jerry has defined business goals and processes and as your customer, he has the, he has high expectations from partnering with you. You see, Jerry is looking to you to resolve specific issue he faces. You must be aware of the risk factors that could affect your ability to meet his needs.

And a common risk in AI is inaccuracy. Due to the nature of AI systems, it means that solutions or productions, you or, or products you develop will never be 100% accurate. And in fact, if someone ever tells you that an AI solution is 100% accurate, it's a strong signal for fraud or a bug.

And so these inaccuracy can arise from different sources like bias, data, overfitting and so on. So make sure you start any project with a proof of concept with A POC because it allows you to validate assumptions, understand the challenges ahead, test in a controlled environment, and gather immediate feedback from the stakeholders. And as leaders, it's our responsibility again to set realistic accuracy levels, um, so your customers can adjust their business processes.

For instance, in my examples with the sales rep, I committed to doubling the opportunities they could generate within the same amount of time. I felt pretty confident to commit to it since after the team did two weeks. POC, we showed that we can triple the leads to opportunity ratio.

So by the end of the POC, you will have clear understanding on your ability to deliver at minimal level of accuracy. The outcome of your POC is actually a well informed go no go decision for the project. And so as we draw this talk to a close.

I really hope you find you found these insight and examples valuable. The common thread here is that success in AI stems from visionary leadership and applying strategic skills you already have. You have the experience and insight to guide your team.

So embrace this opportunity to lead with confidence. Thank you very much for your attention. You.