Being the Conversational AI person in an organization, or team, can feel like a tricky spot to fill. Some people are suspicious about the usefulness of what you’re doing, others are scared your technology will replace them, most of them honestly don’t care.

However, understanding and addressing these concerns is vital to effectively navigate and succeed in this role. By acknowledging the reservations, dispelling misconceptions, and demonstrating the tangible benefits of Conversational AI, you can gradually win over skeptics, alleviate fears, and foster a culture of collaboration and acceptance within the organization.

In this article, we turn to the perspectives of two experts of spreading knowledge about AI, Amr el Rahwan and Dr. Lilian Balatsou. Their insights on AI evangelization provide valuable guidance for professionals working in the conversational AI space, emphasizing the importance of fostering collaboration and facilitating organization-wide information flow.

In this double interview with Amr and Lilian, we cover a range of topics, including:

• Understanding the role of AI evangelists and their approach to different audiences.
• Educating and promoting AI literacy among various stakeholders.
• Effective methods and strategies for communicating AI concepts.
• The significance of hands-on experience in understanding AI capabilities.

The AI Evangelization Experts

Dr. Lilian Balatsou, originally trained as a neuroscientist, operates today as an AI evangelist, talking to different stakeholders about AI and influencing their strategic decisions. She’s also an activist for women in tech in Greece and is often called upon to talk about how to implement AI in her country and globally. 

Amr el Rahwan, as an AI expert consultant, supports international law enforcement and border security organizations in the design and implementation of information systems, to improve their safety.

Experts takes: AI Evangelization & Knowledge Sharing

Entering into the conversation we’ve had with Lilian and Amr, we delve into the experiences and strategies shared by them, offering actionable insights for organizations seeking to optimize their conversational AI practices. Let’s jump into the double interview.

What does your role as an AI evangelist entail?

Lilian: “As an AI evangelist, I talk about technology with diverse audiences, going from educators to entrepreneurs and big corporations. These are totally different audiences with totally different needs, which require tailored approaches. 

If we’re talking about a business, it all starts from analyzing:

• The organization’s use cases
• Their current investment strategy on the adoption of new technology
• Their long-term goals of sustainability and of technology. 

Based on the individual case, I design the strategy that will be followed to choose and implement the best technology, in terms of being the most sustainable for the business, but also the most disruptive. My goals change, depending also on the size of the organization:

• For smaller organizations, I want to identify the solution that is going to make the biggest impact in terms of KPIs and, most importantly, in terms of being truly transformative for the company.
• For larger organizations, I’m more focused on sustainability, as I try to understand what are the technologies that align with the rest of the strategy and are more likely to be adopted company-wide.

I’m basically breaking down the needs apart from the theory, talking about the technology, how it should be implemented, suggesting vendors, designing the strategy and the branding. All these activities culminate in connecting the B2B buyer with a vendor.”

‍

Amr: “Artificial intelligence can play different roles in law enforcement and border security. Broadly, there are two major areas where AI support operations in this field:

- Target identification in terrorism or serious crime cases. AI is leveraged both:

• Using biometrics, like facial recognition
• With (fuzzy) name matching, which relies on Name Entity Recognition and NLP. Name matching is a promising AI application in this field, as the technology is able to match variations of names in the forms of different spellings and pronunciations to the same identity. 

- Risk assessment, to find the unknown person of interest as they engage in behaviors that are considered suspicious and, thus, trigger the AI alert system.”

What’s the biggest challenge in talking about AI?

Amr: “The issue in talking about artificial intelligence is in the expectations. Movies and the media present AI in a way that is often distant from reality, creating high expectations for what it can do. When you explain that AI isn’t able to perform certain tasks or, at least, not with the same ease that they thought it would, people get shocked and tend to give you pushback. I show them that AI is not an absolute source of truth.”

‍

Lilian: “The most difficult thing about AI is its nature. We used to have this very grotesque, Hollywood-based, pop culture idea of AI, that was built through science fiction, or through journalists. This narrative has created two opposite factions of people, with some being completely terrified of AI and its potential, and others super enthusiastic about AI and its capabilities. This is why my goal is to educate a lot of people on AI and intelligence, encouraging everyone to be critical and have an opinion on the role of AI, both in our everyday's lives and in further research and development.

Escalating from simple to complex

Lilian: “Generally, there are three topics that I want to touch upon:

• I first start introducing what AI is and how it started, because I know that there's a lot of misinformation with the intention of moving beyond the buzzwords and making them AI literate.
• The second thing is contextualizing it to my audience, trying to make it closer to them, their own needs and speak in their language, whether that is the language of a university professor or of a CCO.
• The third thing is talking about some strategies and solutions that are particularly tailored and relevant to their case, while keeping some specific rules of either regulation or diversity in focus. 

People and companies want to adopt the solutions but they're not literate enough, to figure out on their own whether they truly need them and what kind of solutions would fit them. My role is to be the person that takes ownership of this implementation.”

‍

Amr: “One method is escalating, so starting out with simple topics and gradually moving into the more complex, but that isn’t the only approach. The approach I’m pursuing takes the opposite view of escalation. Rather than pushing someone to learn by escalating to complex topics, I’m trying to create a user-friendly application that allows them not to bother getting into those. Basically, try to push the software and the AI to do the complex work and provide a simplified layer, or end-user application.

For example, in my research about OSINT (Open Source Intelligence), I show that only officers with strong IT skills are able to use it effectively to reveal the identity of target persons of interest. Contrasting this hurdle, the newly introduced Person-Centric OSINT approach will allow all investigators to achieve great results, without being overwhelmed with learning advanced IT or OSINT.”

What do you evangelize about and think it’s important to talk about?

Amr: “As I realized that most people have a skewed, sci-fi vision of what AI is and can do, I created a video with four scenes from the movies where AI is supposedly used in law enforcement investigative scenarios. Three of them show applications of technology that don’t exist in reality and I use them to debunk the pop culture myths they propose. Only one of the scenes shows a truthful representation of it. This is because, in my experience, people tend to have very high expectations for AI and I need to bring them down, or specify what the current state of the art actually is.”

‍

Lilian: “First of all, I hope to be able to quell the high waves of enthusiasm that some people have, showing that we’re not even close to accessing the forms of super intelligence they envision when they hear someone talking about AI. The message I try to pass on is that we are on track and it is, most likely, going to happen, but we’re not there yet. 

The AI we have now is often mediocre, which is potentially even more dangerous, as it is not truly smart but it can very well simulate and give the impression of being super-intelligent. It’s also self-agnostic, meaning that it doesn’t know its own limitations and only simulates capacities on a surface level, which can really convince some people of its acumen.

At the same time, I also seek to reassure those who believe AI is the cause of a catastrophic series of events. It’s about communicating the nuances of a technology that is evolving and will likely not be either super bad or super good. It is extremely important to start having these conversations now, because we are on a track of growth and scalability. We should be working on creating AI technologies that are both smarter than they are now, but also ethically constrained and regulated, laying the basis for the next era of AI.”

How do you adapt AI educational content to different audiences and contexts you are asked to intervene in?

AMr: “Whenever I'm organizing anything like a training or a presentation, I always make sure to enquire about the backgrounds of the attendees and adapt my materials to try to cater to their interests. For example, if I’m speaking to the top management from the police, I won’t focus on the details of the investigations. Instead, I’ll speak more to how AI can show them predictions of the amounts of different types of crimes in their specific region, which can ultimately support them in decision making.

I try to be mindful about how everyone is trying to catch the information that's relevant to their experience. Those working on the technical side will be more interested in the algorithms themselves, looking to understand how to integrate them in their existing systems and how to build applications around that. If I’d start speaking about domain names, IPs and web protocols to investigators, they’d be very overwhelmed and, likely, also frustrated for being asked to learn about something that does not help them with the practicalities of their work.”

‍

Lilian: “Since my audiences can be very different between each other, I am very aware about understanding what each organization, and the different types of stakeholders within that organization, are interested in.

Many businesses really want to see the ROI, others are more focused on the digital transformational strategies, as they feel that just by adopting the solutions, they can already be one step ahead of competitors. The branding opportunities are another pull for some companies, who might be interested in representing their brand in an AI persona represented by a conversational agent. On the other hand, universities are purely interested in developing innovations, so they want to see the prospect in this line of research. 

Within organizations, there are also different types of stakeholders, which I distinguish between:

• The business stakeholders, who are more interested in seeing the results and impact on the end users, than understanding the process and the constituents of the enabling technologies;
• The technical stakeholders, who are focused on how to enhance their current processes and want to understand model performance and other technicalities. They are often focused on figuring out whether the company should develop an in-house AI algorithm, or choosing which partner they should rely on.”

What are the most effective ways to evangelize about AI?

Amr: “I try to communicate using as many methods as possible: writing blogs, publishing research papers, providing workshops and training, through conferences and presentations. In my experience, the most effective ones of all are training and workshops, because they allow for a direct interaction with the audience.

Presentations can also be effective but they need to be interactive, which can be achieved by asking questions directly to the audience, and/or explicitly allowing interruptions. So, I make sure that everyone feels comfortable with interrupting any time they have a thought or question. However, it isn’t always possible to have this freedom of interaction, as, in some cases, there will be time restraints which can only be met following a more structured approach.``

‍

Lilian: “Since my services are often more directed towards consulting and strategizing, my contributions are usually delivered either in the form of a webinar or a training. I don’t rely on prepared materials to be sent over.”

What’s an underrated approach to teach AI literacy that you think should be leveraged more often?

Amr: “Other than trying to simplify the more complex contents and using visual representations, I find that using examples from everyday, real life can be a really effective method. I encourage my audience to jump into the mud and start to test the technology themselves. What I see is that the hands-on approach makes them often quickly realize how the technology is truly able to reduce complexity for them, which makes them excited and supportive about the introduction of AI in their operations.”

‍

Lilian: “I wish there would be more hands-on experience available to people at all computer literacy levels to play with AI and understand from first-hand experience what AI is truly capable or not. It’s about getting their hands dirty with the technology, not in a bland science fair way, but really in a way that helps them understand and formulate opinions.”

‍

Looking to figure out how conversational AI can help your business achieve your desired outcomes and drive growth? Feel free to reach out via voice message or good old email.

And if you’re still hungry for knowledge, follow us on LinkedIn for weekly updates on the world of conversational AI, or check out our article about successful stories of empathetic AI automation, where we talk about the AI assistants of Woebot, Flo, and Gyant.