5 Key Lessons learned from setting up an AI Community

Which questions should you consider? — Sharing our main learnings.
(We pivoted on the way. If you are only interested in the 5 main learnings, jump right to the bottom!)

Dont try to do the job all alone! Find multiplicators! Photo by Duy Pham on Unsplash

We had been running Data Science projects for a few years, when we decided to start an AI readiness journey for the whole company. After a first big momentum and an Christmas inspiration special, we asked various stakeholders what they wished most from the next steps. Apart from getting inspiration and easy learning these were the most mentioned answers:

“We need to learn from each other!”
“Let’s exchange!”
“It would be dumb to reinvent the wheel, if we have done it before.”
“I want to know, what worked here and what didn’t!”

So we decided to start to build up an AI Community — or the AI Ambassador program as we called it.

The beginning or: How fuzzy is too fuzzy?

It was pretty a blast: The first announcement of the foundation of an AI community made 15% (our goal was 5%!) of our builders organization (consisting of engineers, product managers and UX designers) sign up. Even though you couldn’t have described it fuzzier:

“The AI Ambassador program shall create a strong, high-speed learning and frequently exchanging community.”

We didn’t want to design something people actually didn’t want or didn’t feel the need for. So we put our bet on co-creation: Right after sign up we send the interested people an email back and asked, what they expected from the program and what they wished that would happen.

From this feedback we designed our first session and drafts for the later ones, which happened on a monthly basis.

(P.S.: We called it “AI Ambassador program”, because 1) we wanted members to act as Ambassadors for their business units. That means to come to the community get togethers, learn and then bring the learnings back to their business units. And 2. because we believe that combining strengths from different origins makes fantastic innovations come true. But of course for this you need teams that speak a common language and know about the skills of the others! Members should get to know the others “cultures”, skills and languages like Ambassadors would get to know other countries.)

A few insights

We have learned a lot from successes and failures over the past year where we’ve gotten some things immediately right, but frequently also had to take some iterations to discover what worked best for Scout24.

Not brussel sprouts and not kiwi berries, but the most lovely fruit! Feijoas. Only available in New Zealand though. It taught me: Combining strength from different origins makes fantastic innovations come true. Another reason was our conviction that the best products result from teams that speak a common language and know about the skills of the others. Photo by Elena G on Unsplash

Have multiplicators
For us one of the main reasons for setting up a community before a learning program was as stated above the multiplication effect we expected from AI community members. We hoped that the members — or Ambassadors as we called them — would go back to their business unit and start telling and teaching all employees there about what they learned. We also formulated this in our Community kickoff, but I’m pretty sure we haven’t made this expectation clear enough, since we saw many members didn’t share their learnings!
How we iterated though: We started to prepare presentations Ambassadors could easily share and present and later recorded the sessions so that sharing knowledge became even easier. We also opened up the sessions to people whom it would be relevant to after some cycles.

Split up into tech and product community?
— Of course we got this question! Yet, we consciously decided against this after considering the possible effects. Our conviction was and is that the best products result from teams that speak a common language and exchange frequently amongst different functions. Yet of course you can be sure that this is a challenge, because the probability that there will be different knowledge levels in the room is much higher.

It normally needs a common basis to become friends. And so it does with buddies. Therefore some core questions need to be answered. Photo by Hannah Rodrigo on Unsplash

5 Key lessons learned

1.Make sure you only go “community first” for the right reasons and with the right fundaments.
Andrew Ng suggests in his AI transformation playbook to go with learning first after pilot projects. We consciously decided — since hearing from the experiences of others is learning too, and we had wide experiences from past projects and good practices to share — that we’d go with “community first”. But make sure you take this decision consciously! These questions might help you with answering the question:

• What’s the main goal you need/want to achieve?
• Do people have some general knowledge and can speak “one language”?

(Note: Even if you go community first, you’ll definitely will need to do trainings sooner or later to build up broad and deeper knowledge!)

2. Evaluate the interest of your employees in an AI Community.
Are you getting a similar interest from all business units? Is there a balance between product, engineering and data science folks interested in joining? How diverse are the different knowledge levels?
Depending on these numbers choose one of the following three approaches:

1. Let VPs or Heads of name the representatives from their business units/functions.
2. Let those ones interested in the community apply and do a proper screening process.
3. Be super open and welcome anybody who is interested in your community.

3. Make sure people have the right expectation what it means to be a community member.
Is the goal of the community that you want to set up…

• loose exchange amongst the members?
• knowledge distribution to the business units?
• joint decisions?
• building up knowledge?

Make it very clear in your kickoff that this is your expectation and make sure you get the buy-in from the future members. You’ll need a strong commitment.

4. Have clear topics and clear responsibilities in the community sessions.
To define the topics of the sessions the answers to the following questions will help you:

• What are people interested in most?
• What will help them most in their daily job?
• Which sessions were outstanding? (If you have done any before)

Maybe you even want to create “Topics of the month” (more on this in a later article) and communicate those on all different channels. Independently, let the people with most knowledge contribute with their learnings of past and ongoing projects.

Also: Make sure you measure the feedback in each sessions and define the topics for the next get together(s) with the feedback you get!

5. If you want to establish a Buddy system, make sure the connection creates enough value.
Make sure people have similar problems or can truly help each other, when trying to establish a buddy system.

We asked people to fill in the following information our community page:

• What practical experiences do you have in regards to ML & AI?
• What are your current challenges in regards to ML & AI?
• Which challenges have you solved? Which learnings can you share?

Make sure people fill their knowledge and challenges in as quickly as possible (maybe in one of your first sessions) so that building the connection can take off early and people see the value the exchange creates!

Background

We started our AI Readiness Program @Scout24 in 2018 to make the company truly AI ready. As AI Evangelist my mission is to bring the right ideas to meaningful life.

It means also educating and inspiring the company as well as finding the right structures and processes with the teams in order to run smooth and fun AI projects.

Reach out to me if you want to get to know more! https://www.linkedin.com/in/thejuliabutter/

If you enjoyed this article, tap the claps 👏 button :)