Everyone, and thank you very much for having me. Thank you very much for the organizers of, Women in Tech Global Conference.I'm delighted to have a speech today here at this conference regarding, implementing AI in nontechnical teams. I will briefly tell you about myself. My name is Julia Malich. I'm a head of, public policy at Volt, and currently, I am the AI champion on our team. What does it mean? It means that, I am working on the solutions that we as nontechnical people are trying to implement to improve our productivity, to improve our, day to day work, and I will tell you about these, in just a second. A little bit, else about myself. I have over thirteen years, in government relations, in strategic partnerships, and advocacy. My core experience is lying and building, also AI governance, practices in our global policy team. That's, the initiative that I've been running for over two years.

I have completed the University of Oxford's AI and digital transformation government program that was, not an easy one to complete, but, I liked it very much. And, happy to say that I am a leader of the Bold Women in Tech, leadership board and, one of the authors of, Bold Women in Tech, report at Bold. And, if you want to connect with me, I will be happy to have you on my LinkedIn so you can use this, QR code, if you wish to connect. So do you remember the year of 2023? We had a team meeting, and I raised the question. What do you think about AI? What should we do about AI? Should we be specific? Should we implement something? And you know what I received in response? Don't we have more important things to discuss? That was the answer. That was the first answer. At that time, all tools that we had, were trained only on specific part of dataset.

They could not, connect with some services in real time and so on. And it appeared, at that time, it appeared to be just a fun thing to exercise with AI. And guess, you know, I I I I'm sure you know what happened over the last two years. But, yeah. Two years later, our our CEO, Marcus Willich, announced from stage, during our summer summit that 80% of Bold employees use AI weekly. So that's pretty good number, 80%. But the question is, actually, who are the 20% who don't? The 20% who don't use AI, who are they? And this is my answer. These are heavy nontechnical teams, policy teams, lawyers, maybe communications people sometimes. But nowadays, I believe we also see a rise, in usage of, AI in terms of communicating.

People traditionally not involved in technology processes and teams and those who haven't built a road map for adopting AI. I'm curious to know, in the chat, which teams in your company are less AI ready and less, or maybe had less time to, bring those changes, to the organization. So, if you can write in the chat, I will be happy to see that. Then I will brief you briefly walk you through the shared infrastructure we have at the moment because it all starts from the basics. It starts from what is actually available in the company that a specific team can build upon. So what is our shared infrastructure? Of course, it's generative AI tools like Gemini, chat, GPT. We have also started recently, Claude. It was it happened just couple of days ago.

What do we do with this, generative AI tools? So not just rolling this out to the teams, but also providing, first, training, second, discussion channels, third, champions who help others. For instance, we have on Slack, specific channels with the people who are keen to explore the AI tools and are keen to ready and ready to help other, people to navigate through all the tools that are available and what to do with them and what could be the specific solutions for your, specific and tailored tasks.

And, of course, there is a group that is responsible for company wide rollout. We do not have a top down approach, and I believe this is a very smart move because teams are different. Finance team is not, an engineering team, and they have absolutely different tasks, And they have absolutely different AI readiness, and AI literacy, in in these different teams. So what's important is that we identify champions in each another, in each team, and then these champions try to build something for specific tasks of this specific team. So we have a bottom up approach. Internal communications. Internal communications, include Slackbot and Rojo, which is navigating through the Confluence pages. Our company is having a lot of information stored on Confluence, and, also, Slack, is an official communication channel at Bolt. So we do search information via these boats.

We have, because of this, we have the ability to quickly generate a response. And, this enables structured search, so you no longer go to numerous decks. You don't you no longer check, hundreds of pages. You just ask, Slack bot to find something for you or find something in this specific channel. What was mentioned, you can summarize what was discussed and what are the action steps. So it's quite useful. And, when connected, especially with Claude, this can be a game changer in terms how you interact, with your colleagues, on the daily basis. And, of course, automation tools. One of the latest introduced tools, is n a 10. The main principle is to automate everything possible. We have n eight n working hours where people that have this specific knowledge help others to assemble n eight n workflows.

Also, champions, volunteer guild that help others, on Slack if there are any problems. And, of course, we do the promotion among nontechnical teams. So what are the AI principles before we go to the specific initiatives? First of all, written as simply as possible. The goal is build AI that is safe, fair, privacy protecting, transparent, and accountable, and, of course, with appropriate human oversight. So what is our checklist for every project? We identify who may be affected, customers, partners, employees. We determine the level of risk. If AI could negatively affect a person, we need to assess the likelihood and severity of this harm. The third one is define human oversight, validation, privacy protection, and transparency. And this establishes clear mechanisms, before development begins. And, of course, we document everything. Generally speaking, as we are a technology company, we are data driven a lot.

We document everything. We document decisions. We write, numerous documents for everything, what we are working on, and, we store and keep this information properly, with control of everything, the datasets, the outcomes, and so on. And, specifically, when we talk about the initiatives, specific initiatives for specific teams, what are the key requirements before we start building anything, before we start working with, anything in terms of let let's assemble something that will enable us to do this and that.

First, it's human in the loop. You are the one responsible for anything that goes out, for instance, to your colleagues or to your customers or to your partners. The person that is sitting behind the computer is ultimately responsible for the outcomes of an email sent or a message generated via Claude. So human is always in the loop, whatever you build. Safety and security. We need to validate the system before and after deployment. We need to plan protection against edge cases and errors. Privacy. We need to anonymize data and comply with GDPR, of course, every time, everywhere. We need to implement, data protection and access controls. We need to retain data only as long as necessary. And for that, we need to have clear deadlines. Also, transparency about AI use.

Once, actually, I joined the meeting that was held by AI, and that was quite weird because I was not, I was not warned about that. That was outside, of, our organization. But I was quite surprised. What? How is that even possible that I'm not, warned about the meeting that is being, held by an AI tool? So we need to be transparent about usage of AI. Nondiscrimination principle. We check datasets and monitor decisions for nondiscrimination for all the cases and all the projects that we build. And the last one is social and environmental good. So we build solutions that benefit society. That is the basic principle, and we, of course, consider environmental impact. What should I say else? These key requirements and principles, they are following the general logic of, AI, digital act, AI services act.

So whatever is in the legislation that any, I believe, any technological company is following. So and we are not an exclusion here. I forgot the last one, of course, accountability. We document internally everything. We record everything, and we assign what is important. We assign an owner for each AI project. Now when I started thinking about how to implement AI in our team, what are key three things that I needed to think about at that time? What are we actually doing? Do we have meetings? Do we have meeting notes? Do we write documents? Do we monitor legislation? Do we meet do we meet jointly, internally in the team? What are our processes that are present on a daily basis? So first thing a person needs to do is to audit what you actually in your team and you personally, what you are doing.

Walk through current procedures, where you are inefficient, where are decisions slow. Maybe you have some duplication of information. Maybe you do one thing for several times. Where can you speed things up and audit? Second thing, before implementing, go and talk to your people. Go and talk to your team. Gather ideas. Run a hackathon. Let the team share their experience and their workflows so you build together. Do not, implement a top down approach. It will not work. So if you implement, for instance, let's use these and your team does not need it because they do it in a faster way, It's not gonna work. So first and the second, audit and talk. And the third one, you can actually what you can do? Concrete initiatives can have clear owners. You can engage your team, those who are AI, keen to learn something new, AI savvy people. You can engage them and build this together.

Because one person is good, but when you have the guild of people to work on the project, I believe it's so much better. Then I've briefly spoken about, what we think through when, providing some guidance on how to, implement an initiative or a project in your team. Last two things, account for GDPR and confidentiality, of course, and prioritization. You cannot implement everything, absolutely everything. You need to start with one project, maybe two projects, complete them or reiterate or pivot. You cannot implement in Salesforce and Gmail and all at once. Pick the most important ones for you. Next, you set the rules upfront. What is your AI policy in the company? One page. What's allowed? What's not? Why? People need a very simple one pager, what they can and cannot do. Then red teaming. Test everything. Test AI outputs before they reach stakeholders, internal or external. Doesn't matter.

Test it. Confidentiality boundaries. Again, people need to understand what what they can insert inside and and what they can't. And approval processes. Fast approval process that records decisions without creating bureaucracy. This is a game changer. And for instance, if you run, at the end of the day, a session with Claude, please complete my day and give me this and that, post it on Slack. You will have, then approvals and feedback from your team at the earliest. So what we as a team automated, example, our public policy team, ministry briefings, press content, meeting coordination, regulatory monitoring, multilingual research, tender bids. And these are the concrete examples. Sometimes it's easier to use the external provider. Sometimes it's not. Now from my perspective, also, cloud, is competing with all the external providers that have been built, before.

It's possible to work on a joint cloud projects at the same time. That's one of the, examples that we implemented. Then NA 10 as a tool, we enabled, regulatory monitoring that gives us automated alerts in less than twenty four hours. That's one of the examples of the simple stream that was assembled, by me personally, and it gives us updates on what is happening in one of the countries and in one of the ministries. Then we've been talking about the tools. Now let's, quickly, talk about the skills. So you chose the tools, created a safe environment, tried to build a prompt library for your team, curated and tested prompts for policy talks, for policy tasks, pinned on Slack, used daily. Share this knowledge with the people. They will appreciate it. And, of course, run, live training for everybody. What's available? What is the solution? What we've organized? What we've provided?

And that's how it can look like, in practice, on the timeline. So two weeks, you can kick off. You can pilot in several more weeks. So pilot can be ended in one month. Then you measure, and then you make a decision whether it works or not. Several projects of ours were, declined or deprioritized because we made a decision to do so, because we we saw that there is a possibility to do something else. There is a possibility to do it in a better way, maybe with Claude, maybe with something else. So six week cycles, work, very good when they are with assigned owners, clear adoption goals, and concrete success criteria. Last but not least, champions. Champions. Champions. And once again, champions. Enjoy working together with your team, AI savvy. You can when you engage these people, they will help you in running this.

Have you can also have proper communication rhythm. You can send weekly Slack updates, monthly showcases, quarterly leadership reviews, so everybody is updated because the tools are changing all the time. Build some kind of a digest that is, working specifically for your team. Design the training. Not generic AI courses, but, specifically, what is your team is doing? Is is your team do you have a finance team or a legal team? They have specific use cases, so it's better to tailor specific trainings, not just regular how to write emails. This will be more efficient. And, of course, measure real impact. Course completion, response time, cycle time reduction, and employee confidence. So before we started, we've measured where we are.

And after one year of the initiatives we've been implementing for the whole year, we will be measuring it again, how people feel, what they want, are they comfortable with working. And, also, we have, a new dashboard that is calculating the tokens, for all the platforms that we have, for all the gen AI tools that we have. So I've briefly talked about that. And key takeaways, nontechnical teams are the biggest untapped opportunity. Start with an audit. Rules accelerate speed. They don't slow it down. Clear rules are always good for any initiative. Build something together with champions and measure what matters, quality, time saved, risk reduced. Show impact to your leadership, not just we've implemented something, but showcase that we did this. This impacted this side of our work. We've saved a lot of working hours and so on. And as I said at the very beginning, every function can join the 80%, not just engineers.

Thank you very much.