Why does Intel need an AI community?
Intel Israel decided to strengthen a few areas, and AI was one of them. We're a very large organization and we collect a huge amount of data, but we weren't using the treasures we'd gathered, because the volumes are truly enormous and they're hard to work with using standard BI tools. So the company wanted to create internal impact with AI, and the way to do that was to encourage as many departments and groups inside Intel as possible to use it.
How was the community founded?
When Intel Israel decided to strengthen AI, we set up working groups to figure out how. One thing that came up: this is a topic that's handled at a very high level, and it's very hard to bring it down to the ground, to the people, to everyday use. Why does it matter that it reaches the ground? Because the most interesting, highest-impact projects in the organization will come from the people sitting at the keyboard writing code. That's the point where we realized we needed to build a community.
Ilan Harari and Inbar Vest, who were part of the working groups, took it on themselves to build the community. They recruited people to lead it. I'm one of them, and off we went.
When we started thinking about a name, we asked ourselves what we wanted, and we realized we wanted hands-on work with AI: not an abstract idea, but something concrete and practical. We picked the slogan "Our hands on AI." It fit the first year. In the second year, people got it and wanted something more meaningful, so we switched to "AI CAN MAKE A DIFFERENCE."
When we built the community, we wanted it to be relatively exclusive. We started by thinking about different kinds of AI knowledge, then asked potential members to fill out a form with the questions that mattered to us. We sent a message to the whole organization, and the way people answered the questionnaires shaped how we defined our community. The main criterion: anyone who had taken a serious course was a fit. It was important to us that everyone spoke the same language and that the training sessions weren't too basic.
The foundation for this whole process is management backing: budget, time, focus. And later on, people to show results to, because they genuinely care.
What is the AI community and how does it work?
We have 3 goals in the community:
- Teach: we host experts from inside Intel who build up our members' professional knowledge. We have clear tracks and a methodology you can learn.
- Quarterly meetups: we dig into topics beyond the AI work happening inside Intel, sometimes with external speakers.
- Projects: we initiate, encourage and launch projects from within the community, and help members get them off the ground. We connect members with mentors from inside Intel who help them with concrete projects, and we've had some great successes.
We measure each of these areas and define what success looks like for us.
When did you know the community was working?
On the professional level, we've reached a point where the organization really understands what AI is. For example, we wanted the people who build the data retrieval to know what AI is and realize there's a lot of hidden data that isn't being used, so they'd want to connect to a process that makes use of it. Working in the community develops them as engineers, and the toolbox we give them lets them deliver solutions. Today, because we've taught them, people understand what AI requires, so they know how to ask themselves good questions instead of jumping straight into projects. And when they do decide to launch an internal project, they get all the tools they need to make it succeed.
On the organizational level, before COVID we held a kickoff conference with 200 people. There was a talk by our community and another AI solution for the organization. We showed them how our AI projects were built without us being data scientists, and the response was very warm.
On the community level, take what's happening now: we're recruiting mentors for our community, and we have an amazing pool of experts who want to take part. It's a lot of fun. Intel has a huge number of AI experts, and thanks to the same working group where the idea for the community was born, we know many of the "spearheads" across Intel. When we reach out, they're genuinely happy to help, because they know our work.
Cracking the win-win: how do you make sure both the community and the organization get value?
The win for the community: activities that broaden horizons, professionally and personally. The organization lets us grow in new directions and be innovative, and there's a lot of fun and enjoyment in the work.
The win for the organization: embedding innovation in the most effective way possible. AI isn't a slogan. It's a way of thinking that changes reality.
What challenges did you face?
- On the professional level, one of the biggest problems with AI is this: someone tells you they have an AI problem, and once you get into the details you realize very simple software would solve it. The other possibility is a problem so complex it's hard to model at all. We had to find answers to the professional difficulties of working with AI day to day, and we looked for what's in the middle. Then, on the community management side, you have to check that the ROI makes sense and is worth the investment.
- We found that the hardest thing is getting people involved. At Intel you get lots of emails you never open, never read, just ignore. Our toughest battle is breaking through the noise and staying focused and sharp.
- Intel moved to Microsoft Teams, and for some people that made things harder. We knew from the start that we wanted Teams, because you can put questionnaires, recordings and files there. We expected the group to buzz with questions and answers, but in practice that hasn't really worked, because people don't really know how to work on a feed. Every so often there's a short burst of activity and that's it: a question is asked and answered. So that's something we're still working on, figuring out how to reach end users more smoothly.
- This is hard work that takes a lot of administration, so we bring a lot of people in with us. Luckily, the group itself is a joy to work with: people who aren't afraid of hard work. You don't have to convince them. Very often there's agreement and an open atmosphere about what needs doing and who to bring in.
Tips for building a successful community
- When people think about AI, they think about complex tasks. We actually started with validation, to prove that AI capabilities move us a step forward.
- One of the things that helps the community succeed and stay full of active members is making it an exclusive club.
- We collect feedback through questionnaires and hallway conversations. Whatever people write or tell us, we act on. We're always trying to improve.
- You need a winning team: people you'll love working with on such a challenging project, people you share a common language with.