Main content

Salesforce Connections 2024 - "Commerce should not be an island" - how customers are personalizing commerce and making sense of Copilots

Jon Reed Profile picture for user jreed May 23, 2024
Salesforce Connections 2024 brought the AI conversation to a head - via a customer-facing Copilot for Shoppers demo I've featured here. But for Commerce Cloud customers, there's a bigger goal in play: get off of "commerce island," and kill all the B2B portals while we're at it.

Jenna Flateman Posner, COO Solo Brands at Salesforce Connections
(Jenna Flateman Posner of Solo Brands)

As I write this, Salesforce Connections 2024 is just past the halfway point, but we can obviously say this is the year of the Copilot. 

With the Commerce Cloud Copilot heading for a planned general availability (GA) of June 2024, and Marketing Cloud's Copilot GA in July 2024, you could not get very far in the McCormick conference center without hearing how these Copilots will change your day-to-day. 

However, I see a different underlying theme at this year's show. But for now, what do customers think? First, the news releases: 

Wednesday featured a slew of keynotes across products. Needless to say, I have plenty of pesky questions. The Commerce Cloud news is my focus here, due to my sit down with Alex Bucher, VP Product Management. First, the Commerce Cloud details

Salesforce Checkout: "A unified, one-click checkout platform connected to customer data, Salesforce Checkout is built on the Einstein 1 Platform to provide shoppers with more consistent, cross-channel experiences throughout their buying journeys." 
Headless commerce capabilities for B2B: "Developers can implement headless commerce with out-of-the-box APIs in Commerce Cloud that help merchants create engaging B2B sites that rival traditional consumer shopping experiences."
Composable Commerce Enhancements for B2C: "Developers can combine or create templated, composable, and customizable headless approaches to B2C ecommerce site development with new powerful Salesforce APIs and out-of-the-box personalization functionality."

Enterprise-grade commerce experiences - are brands getting closer?

I see a theme behind this news: whether it's B2B digital commerce or B2C, it's time to meet/exceed the high bar Amazon has set around direct-to-consumer personalization and logistics, but with enterprise-grade solutions, and, hopefully, a human touch when needed - an area where brands can (and should) strive to do better than Amazon. Oh, and those enterprise-grade solutions should be accessible to smaller customers with lean teams as well. Being "headless," they should integrate with Amazon, Shopify, SMS and more.

But that's my view. So I asked Bucher: what is his message for those who are not here in Chicago? As he told me: 

What I'm hoping people take away from today: it's really the convergence of data, AI, marketing and commerce all really coming together. I know it sounds somewhat like marketing speak, but it really is building that end-to-end journey. When you think about commerce, it's the transaction piece, the bottom of the funnel - but there's so much to the full customer lifecycle that you bring to life. And that's where a lot of our innovations are driving us.

The Commerce Cloud Copilot is not yet GA - but we still need to talk about results. After all, Salesforce has been developing AI-related platform enhancements for 12 years. So we shouldn't have to wait to talk about success with digital commerce. During the Commerce Cloud keynote, Salesforce cited the example of Siemens Customer 360: >23% win rate, and a 2x increase in first-time case resolution. Moving from older systems is paying off: according to Salesforce Commerce metrics, "customers saw a 29% increase in digital revenue by shifting from legacy channels." 

Keynote customer demo - Einstein Copilot for Shoppers in action

Salesforce did something different during the Commerce Cloud keynote: they had a customer provide context to a demo. Jenna Flateman Posner, COO Solo Brands (pictured above), spoke to a scenario with Einstein Copilot for Shoppers capabilities: personalizing a shopping interaction on WhatsApp. 

Another challenge for this demo? Showing how personalization can work when you don't have first-party, "logged in" customer data. This demo involved clicking on a personalized Instagram ad of a Solo Brands portable Mason stove. Clicking through to the Solo Brands site shows a 3D version of the stove. But the demo shopper gets distracted, and doesn't add the item to their cart. However, it's not over. As Salesforce explains: 

Sometime later though, on my phone, I receive a notification from WhatsApp - it's the Solo bot, powered by Copilot for Shoppers re-engaging... Solo was able to recognize who I am, even though I was shopping as a guest user and had never logged in. They're able to piece together my past purchase history, my previous interactions with their site and my engagement through Instagram. And on top of all that, because I'm a loyal customer, they offered me a 20% discount.

But will this stove fit in my backyard? We ask the Copilot: 

Copilot for Shoppers - demo
(Copilot for Shoppers - demo in action)

We snap a patio picture, and send to Copilot - and Copilot provides the product specifications. It also recommends the Solo Adirondack chairs to help complete the set - and sends along a demo photo of the chairs and stove in the patio backyard:

Salesforce Connections - patio pic
(Copilot for Shoppers - WhatsApp chat, providing product specs and customized image)

So we ask Copilot how much this will cost, and get a response with a pre-built card payment link, and the promotion applied. In this case, the order is processed by signing into to our existing Amazon account. So how is this Copilot working out for Solo Brands? As Posner told attendees: 

Let me tell you what really has inspired me. First off, in no particular order, the ability to leverage the loyalty and trust that Amazon has generated with their consumers, and transition into our environment is paramount. 

"We have a ton of awful AR - it's nice to see a really good one"

Posner says that there is a "nuanced value proposition" here involving inventory. This enables Solo Brands to better utilize their existing inventory, and potentially take some bigger swings as well with bigger risk categories and styles - with the knowledge that the B2C environment "can help drive throughput." 

For Posner, it's not just about Copilot for Shoppers, it's about how all the pieces fit together: 

One-to-one personalization, product recs, upsell, cross-sell opportunities, all at the intersection of AR. I know we have a ton of awful AR [Augmented Reality]. So it's nice to see a really good one. Putting our product in context of the consumers' buying process, and giving them confidence in the sales process that it's the right product for them, is definitely going to help conversion. 

Opening up a new, highly interactive channel like WhatsApp is also a big win: 

While we're all leaning into AI, we're leveraging this aggressively as we can, [with a mix of one-to-one and one-to-many] with email and SMS, it's hard money. It's starting to get a little commoditized. So the idea of having a new communication channel that requires a level of back and forth, and engagement and communication with the consumer that's so intimate, that is a really great opportunity to drive that affinity, and that customer engagement to really reinforce the brand points.

My take - on commerce islands, and killing all the B2B portals

Salesforce has advanced its AI messaging considerably; I hear less hyperbolic statements, such as AI eliminating UIs. Instead, Salesforce AI demos had a comprehensive feel, with careful touches: such as disclaimers notifying users they are dealing with generative AI output to review. They are the first I've seen to offer prompt templates, to make it possible for business users to import the proper data into the prompt query (via RAG methods, etc) - without coding.

But not all customers will be ready for Copilots just yet - and there are pricing considerations to evaluate (e.g. a Commerce Cloud customer would buy consumption-based Data Cloud and Einstein "credits"). As I see it, the big story from this show is not really AI, but how far customers can take their business by moving off of siloed ways of working. As per the Commerce Cloud keynote: "Commerce should not be an island." As Bucher told me: 

When you think marketing plus commerce plus data plus service all of these areas together... You need to remove commerce from simply that island - and think about the full funnel result, the full customer lifecycle. 

And when you get off that island, guess what? The data platform you have in place is exactly what AI will need for an accurate/better business result. An even stronger message came from a B2B customer that was quoted (anonymously) during the keynote:

Kill all the portals. 

I can't blame a customer for feeling that way; old school B2B purchasing portals are well past their expiration date. Bucher adds:

People expect a direct, consumer-grade experience no matter what they're buying online, including business purchasing. Think about a salon refilling their products. Why should that experience be any less personalized than if you were to just go buy those same products yourself on a retail site?

Hyperpersonalization is a big theme in all of this. During a recent podcast, I aired out my grievances against hyper-personalization hype, including the risk of the "creepy" factor, and the difficulty predicting the real-time context of consumers, given that our lives and priorities can shift on a dime. 

Salesforce has some thoughtful rebuttals to my grievances here. Short version: Salesforce believes it can keep the downsides of hyperpersonalization in check with human oversight of AI systems. And, in some cases, AI can help with the real-time context, not by trying to predict it autonomously (which is my concern), but by pulling real-time data from the consumer into the scenario.

One keynote example: adjusting an outdoor equipment web page to family-friendly equipment in real-time, when the shopper tells the AI bot they are shopping for the entire family. That is indeed proper context. There seems to be a better understanding here of how to design AI to incorporate human feedback, not fantasize that we can live without the human element, correcting/modifying our interests and preferences. It's encouraging to see designs that are pulling in this human preference info, but in a non-clunky way.

But of course, human-in-loop designs do impact the ROI-of-AI calculation. As it turns out, humans can be pretty useful creatures, but we are not always cheap to sustain. So, we'll have to look at the ROI of this type of customer-data-informed AI carefully. 

During a CMO panel today, Salesforce CMO Ariel Kelman made the case that using this kind of customized prompt with customer data via RAG can reduce, or even eliminate the need for enterprises to train/fine tune their own models, a potentially huge expense. I also heard good things about tangible gen AI benefits from the other CMOs on the panel, but this will need fleshing out. That's for another time. Look out for more Salesforce Connections coverage shortly from our team.

A grey colored placeholder image