How Salesforce Runs Einstein GPT Pilots (ft. Gucci GPT)

Share this article...

Einstein GPT has made headlines in what seems like a constant cycle of GenAI-enabled capabilities and use cases. It kicked off with Einstein GPT being unveiled and the Salesforce AI Cloud roadmap including no less than 16 capabilities rolling out in fast succession – with use cases including Slack, sales, service, marketing, commerce, analytics, and app builders. Salesforce are ‘covering all bases’ by bringing generative AI to an array of personas across the organization you already support in improving their work lives.

With all this exciting technology, the question on everyone’s lips is: When do we get to use it?

Salesforce have brought Einstein GPT to where it is with incredible speed and, behind the scenes, have been running pilots with key brands – one of which is Gucci.

With an oversubscribed Service GPT pilot (where hundreds of organizations requested to join, lining up ‘around the block’), what can we learn from organizations trialing Einstein GPT to roll out similar experiments in our own organizations? Plus, what results did they achieve? Let’s take an inside look at Einstein GPT trials.

How Einstein GPT Trials Work

Salesforce believe it’s important to get users into the product early on. This is known as the pilot phase – the first phase of public testing, involving a small sample of customers who have opted in/been nominated and are a good fit to test drive the functionality.

Once the pilot phase has been completed, the results gathered, and iterations made – the functionality goes into Beta (this time to a wider, unrestricted sample of customers) before finally moving into General Availability (GA).

With GenAI (GPT) being a technology that has been adopted so rapidly – faster than has ever been seen before – pioneering GenAI in CRM, Salesforce are keeping their eyes on the prize.

The trials involve performing A/B tests between two groups of users: one group using the Einstein GPT technology and another group in equivalent roles not using it (i.e. the ‘as is’ control group).

As it’s all cloud-based, trials offer rich user interaction data. Salesforce have been very hands-on with pilot customers. There needs to be a decent sample size in order to get statistically significant results while ensuring there’s an ‘apples to apples’ comparison.

Gucci GPT

Gucci was Salesforce’s first pilot customer for Einstein GPT, creating what’s been termed ‘Gucci GPT’.

Case handling efficiency with Einstein GPT (Service GPT) was 30% higher compared to users not using the technology – a very promising pilot. Salesforce also piloted their own Einstein GPT technology for CodeGen with their engineering teams; coincidentally, a 30% productivity increase was recorded.

From there, Salesforce have improved upon the technology, influencing their AI roadmap – and will be doing further rounds of pilot research.

Salesforce believe that these figures will increase with time as Einstein GPT becomes more tightly integrated into the user interface of apps.


Picture this: You’ve just enabled Einstein GPT in your business, and your users are suddenly 30% more productive at their jobs. That’s an impressive number – especially when there’s scope for that to increase as more use cases are released.

With an increasing number of use cases for Einstein GPT across Slack, sales, service, marketing, commerce, analytics, and app builders, knowing about the rapid pilot programs that Salesforce are conducting is fascinating.

How will you be rolling out similar experiments in your own organizations?

Add Comment