Kevin Kelly recently closed his The Next 30 Digital Years Long Now talk saying, “It is easy to predict the next 10.000 Startups! Take topic X and add AI.”.
Salesforce put a stake into the ground by boldly claiming that you can now create your first AI-powered community using their platform.
So far details are scarce what exactly Einstein is doing to help your community to be more successful. I added an “Unproven” stamp on a copy of their marketing page. When you click on the demo link, you end up with a video from 2014 with no word of AI or Einstein in it.
Well, what could machine learning do to enhance your community?
Engaged customers bring in three times the value over regular customers. If you do it right, they will happily help you create your best products and proudly promote them too. AI can do a lot to move more customers into the engaged category.
How do you nurture passionate customers? By treating them like friends and family. You give them a voice and an opportunity to influence your product as well as your community activities. You make them shine by allowing them to share their expertise via blogs, webinars, and sessions at events.
The rocket fuel for machine learning is real world data, that it can dig into, derive patterns, make suggestions, continuously test and fine tune these results.
Thankfully online communities provide that rich data set that makes your AI hum. Every post, like, connection, answer, link, … can be fed into the machine learning algorithm.
There is one caveat though: in a community, you form these deep connections, the lasting friendships that create the engagement that you are looking for when you meet face to face. That real world activity is not, or only peripherally reflected in your community’s online activity. Something to keep in mind and adjust your algorithms accordingly.
Here are a couple of ideas where machine learning can support your community efforts:
- Improved search results, tailored to each member based on their past searches, posts, interests, and expertise level, is low hanging fruit.
- Connecting members based on their expertise, behavior, and region. You liked this content, you may like this person that wrote the following content. Or you both liked this content and live in Buenos Aires, you may want to get to know each other and possibly create a meetup.
- Discovering up and coming community members to nurture and engage with. Feed the algorithm the behavior of your top community influencers and tell it to find similar activity in other topics or regions that you would like to develop a more engaged community in.
- On the fly creation of topic communities: We want to improve this product. Targeted beta customer invitation to all members that had a positive impact on that product’s community discussions.
- Individual newsletters summarizing the community activity for every member. Targeted individual summary newsletter for every member derived from interests, prior activity and connections to other community members.
- Highlighting problem areas. You can train algorithms to point out areas of conflict in your community before they escalate, which gives you the ability to swoop in and defuse the situation before it gets out of hand.
What ideas do you have that am I missing?
You need to be vigilant and flexible and overwrite the algorithm suggestions when needed, or you end up with an escalation like United did last week. John Robb points out how the reliance on the algorithms made the situation worse: the United employees on the ground were following the systems’ suggestion and offered max $800 for someone to give up their seat, it wasn’t enough and they continued with the protocol, with frustrating results.
What other community management function could be enhanced with machine learning? For me, AI enhanced community management is the right perspective that should guide our platform development. You need the personal touch of a good community manager to get to the deep engagement with your members. Let the AI help these managers shine.