AI can yield many opportunities for businesses. It can also cause challenges. Salesforce has helped its customers with AI since 2014 through consistent investment and innovation. This has led to various features being made available; allowing customers to harness its benefits, faster. Up until recently, this has focused on predictive AI (e.g. Lead Scoring). In recent months, generative AI has come to the fore. However, if you’re new to Salesforce AI, where do you start?
Salesforce recently hosted an 8-webinar series called ‘Salesforce AI Explained’. Want a quick summary of the best bits? Read on!
Benefits of AI
Salesforce's latest State of IT report found that 86% of IT leaders expect generative AI to soon play a prominent role at their organizations.
So, what’s the fuss about? In short, AI, if used responsibly and, in the right context, can be transformative. From the 8 AI Explained webinars, this seems to come down to four key benefits: better decision making, user empowerment, system futureproofing and enhanced customer service.
The potential benefits are not limited to a single group of users. The webinars outlined how AI capabilities have been woven across the platform. For example, in the sales webinar. it was reported that 61% of salespeople believe generative AI helps or will help to deliver better service and could save up to 4.5 hours per week. The key theme is how AI can empower teams to work more efficiently and help businesses provide better outcomes for their stakeholders.
Challenges of AI
A signifcant minority of organizations are architected for AI success, with only 2% reporting no barriers to using their data for AI use cases. There are a variety of barriers, from data quality to funding to talent, but outdated IT infrastructure leading to data silos and fragmented systems tops the list.
Realising the benefits of AI isn’t without its challenges. From the webinars, there were 5 key recurring themes:
- A lack of training/knowledge within teams.
- System readiness.
- Privacy and security
- Data issues (i.e. data availability and quality).
- Accuracy (i.e. of model outputs).
These challenges are not surprising. For example, the recent MuleSoft 2024 Connectivity Benchmark report found that, on average, organisations use 991 applications across their digital estate. Data will almost certainly be inconsistent and siloed. Given AI requires high-quality data to make accurate predictions, or generate helpful content (e.g. replying to a customers question), this can cause various hurdles. For example, impacting the accuracy of a model’s output.
As with any new technology, upskilling is also needed to both setup as well as use the new tools. In the case of AI, it also comes with data security, privacy and ethical considerations. If these factors are not considered carefully, it can damage businesses (e.g. data breaches, harm brand reputation, etc). To get the benefit of AI, a number of hurdles must first be overcome.
Solutions
The webinars outlined how Salesforce can help businesses get up to speed and utilise AI capabilities.
NOTE: Bear in mind that the release dates etc are as of March 12th 2024. Due to the speed at which Salesforce is releasing new features, please ensure you check Salesforce documentation for further updates.
Einstein Trust Layer
A high-level summary of the key features is provided below. This helps to ensure personal identifiable information (PII) and other sensitive data is kept secure by masking and preventing retention within the Large Language Model (LLM) used to generate the output.
| Feature | Description |
|---|---|
| Prompt | A request made to a generative AI model to perform a specific task/action. |
| Secure Data Retrieval | Access data securely (honouring access controls) for use in prompts. |
| Dynamic Grounding | Contextualise and tailor prompts by merging in business data (grounding). |
| Data Masking | Mask PII and other sensitive data before sending to a LLM. |
| Secure Gateway | Secure interface to route requests to the desired/supported LLM. |
| Zero Retention | Prompts & outputs are not stored within LLMs in the trust boundary. |
| Generation | LLM output returned. Masking removed (output 'rehydrated') with business data. |
| Toxicity Detection | Scans and scores the output to mitigate against harmful content. |
| Audit Trail | Audit trail of steps throughout the process. |
CRM Apps
You may have noticed that the Einstein Trust Layer diagram says ‘CRM apps’ on the left hand side. This is because Salesforce is embedding AI capabilities across the platform. The AI Explained webinars covered a large number of capabilities. Here are some of the main take-homes.
Learn More
As you can see, the webinars covered a lot of ground! Thank you to Salesforce for hosting the webinars. If you want to view the webinars for yourself, they are available here.
Summary
The Salesforce AI Explained webinars outlined many of the benefits, challenges and potential solutions to use AI. However, it is clear that Salesforce will be continuing to innovate and add more features. If you’re at the beginning of your Salesforce AI journey, now is a good time to start. Focus on small use cases and choose the right tool for your need. This will help in gaining experience with the tools and help when more capabilities become available in future.
Bonus Penguin Fact 🐧
The life expectancy of a Humboldt penguin (Spheniscus humboldti) is typically between 15-20 years in the wild. However, this past week, ‘Spneb’ the penguin celebrated her 36th birthday! Spneb is believed to Europe’s oldest Humboldt penguin. Click here for more info!
Want More?
Find this article useful? Why not Share!
Or check out some of our other articles: