Salesforce, The A.I. First Platform
Sadly I didn’t manage to make the trip to San Francisco last Tuesday and attend Salesforce’s first developer conference ever: TrailheaDX.
Thankfully though, Salesforce had thought about us all professionals working on their own customers’ orgs, writing code and configuration, deploying from sandbox to sandbox to production, training future users or simply not being able to find a budget to travel to the US... All User Group leaders were sent in advance a TrailheaDX “Viewing Party Kit” pack including all sorts of swags and the secret details for the keynote webcast.
As a co-organiser of the Thames Valley DUG meetup in the UK, I was one of the happy attendees to this remote keynote and haven’t been disappointed…
The keynote content is, of course, a little different from what you would find at a World Tour event: less sale and more technical information. I particularly appreciated some insight to the Winter and Spring ’17 releases… There was also a mention of the new cloud, the e-Commerce Cloud, just one week after the announcement of the Demandware acquisition. If you want to know more about TrailheaDX in general, I recommend you read Alba Rivas’s recent post on the subject.
But make no mistake, the big announcement was about Artificial Intelligence. Salesforce has seen quite a lot of innovations along the years: “the social platform”, “mobile first”, “API first”, … now “AI first” as announced by Alex Dayon. Unlike the previous phases though, “AI first” is not just an architectural update, it’s an entire new paradigm. One that I would say is taking componentisation to the next level and is about to blow your consultant mind off!
“AI first” is not just an architectural update, it’s an entire new paradigm
With a componentised (Lightning based) solution, implementers don’t write bespoke code any more but reuse well tested components built by niche specialists developers and deliver a solution which is declarative, still highly granular. Componentisation doesn’t mean that we won’t need developers any longer but, typically, developers will build and sell multiple small components to many customers around the World rather than one big custom app to one unique customer. Implementers (IT or SI) will either re-use components or build low code applications. This is very much in line with the evolution of programming languages and the never ending addition of abstraction layers, IDE, libraries and tooling to simplify the developer work, now all packaged in one component.
So, what’s all about this AI first platform? Well, the AI components that implementers will use in their solutions will bring some capabilities to help on a specific task (like "guess the most realistic probability to close a deal") but without being directed on how to provide this information. It will works thanks to the clever algorithms (Artificial Intelligence) embedded in the component accessing a huge amount of data and the relevant CPU power to process it. The first attempts may fail (for instance a few 80% opps may not close in the end) but then the system will learn and improve with time... without modifying its code!
The system will learn and improve with time... without modifying its code!
As you read more about Salesforce AI-first platform in the coming months, try and figure out use cases for yourself or your customers. What will you do when Salesforce releases the technology to its customers? No doubt Salesforce's new strategy means that it will develop its A.I. footprint and continue acquiring specialised startups, so, stay tuned!
[Edit 1]: Salesforce has just released a new Trailhead module about AI basics. It's a fun way to get a better understanding of AI and how could it help your customers.
[Edit 2]: Salesforce provided more details about it's AI offering ("Einstein") at Dreamforce 16. Salesforce's AI algorithms will do the leg work for you. The big amount of data required to train these algorithms will be provided by Salesforce not by the customer. The Data Scientists who build the AI algorithms will also be working behind the scene, in Salesforce "labs" (Data Scientist as a Service anyone?). This means that in my example of "opportunities with a real chance to close" the solution will be already trained before go-live and not six month later thanks to trends and patterns already known by Einstein.
I hope you enjoyed this post. Don't hesitate to ping me on Twitter if you have any comment or question. Bye for now!