Following on from our previous article – “Artificial Intelligence – The Model T”, Salesforce and other vendors have announced a raft of new AI related products.
Announcements are one thing; product releases are another. Announcements are like a stake in the ground advising customers that “the next big thing” is coming, whereas product releases are the actual product being delivered. There can be many months between these two dates.
AI Cloud
On June 12, 2023, Salesforce Announced the AI Cloud.
“AI Cloud is a suite of capabilities optimized for delivering trusted, open, and real-time generative experiences across all applications and workflows.” (announcement link)
What does this mean and how is it different to other vendor products and capabilities?
The challenge today with AI is trust. How can we be sure that:
- Data used for AI processes is handled securely (eg GDPR)?
- AI Hallucinations do not occur?
- Toxicity and Bias do not occur?
Privacy
When data is to be processed via an AI tool, often this data (eg a customer contact or account record) is sent off to an external AI tool, the data processed and the results returned. The issues here are about data governance and data sovereignty. Who has the control and ownership of the data and where is the data being processed? If the data is being processed offshore, does this offshore center comply with the expected privacy standards where the data originated?
Hallucinations (mistakes)
AI Hallucinations occur when an AI Model makes a mistake. This can be caused by inaccurate or inadequate training of the model, by unexpected inputs or deliberate attempts to confuse the model.
Here are some early examples of AI generated hands by Stable Diffusion, a popular AI image generator:
Fortunately, most of the AI image generators can now accurately generate hands. This however has caused some controversy with the generation of “deep fakes” and a fine line between the use of factual images and parody or satire.
The Washington Post published an article in March 2023 about fake images of Donald Trump’s arrest being posted to Twitter and being viewed over 5 million times, even though the event never occurred. (link)
Other issues
Toxicity and bias are other issues facing AI. Unintended toxic responses or responses that have a bias (whether intended or not) are two more challenges for Generative AI. If an image model is trained on young, white, male images then it may struggle to generate images that contain diversity of race, gender and age.
These issues of hallucinations, toxicity and bias also occur because the AI models can’t actually think and don’t have common sense (and probably never will). As a result, we need to employ safeguards to prevent or limit their occurrence.
The Einstein GPT Trust Layer
To address these concerns, Salesforce has developed a solution called the Einstein GPT Trust Layer, a key component of the new AI Cloud:
“AI Cloud’s new Einstein GPT Trust Layer resolves concerns of risks associated with adopting generative AI by enabling customers to meet their enterprise data security and compliance demands, while offering customers the benefits of generative AI.” (link)
The trust layer will separate sensitive data from the Large Language models (LLMs) while still allowing enterprise customers to benefit from the potential of generative AI. Note: Very few details of exactly how this trust layer will operate are currently available.
The trust layer will also allow customers to choose a variety of Large Language Models to perform their AI processing:
- Salesforce LLM (by Salesforce Research)
- Public third-party models including OpenAI, Amazon Web Services, Anthropic, Cohere and others
- Bring Your Own Model – where customers build their own specific models
AI Across the Enterprise
With the Einstein GPT Trust Layer in place (general availability in June 2023), Salesforce has announced a suite of AI enhanced products as follows:
- Service GPT is in pilot today, and will be generally available in June 2023.
- Sales GPT is in pilot today, and will be generally available in July 2023.
- Marketing GPT will be in pilot in June 2023, and generally available February 2024.
- Commerce GPT is in pilot today, and will be generally available in July 2023.
- Apex GPT will be in pilot in June 2023.
- Flow GPT will be in pilot in October 2023.
- Slack GPT is in beta today, and will be generally available later this year.
- Tableau GPT will be in pilot in November 2023.
It’s not cheap!
Getting onboard the AI Cloud is not a cheap exercise with “Starter” kits priced at $360,000.
Salesforce Ventures (the investment arm of Salesforce) have doubled their Generative AI Fund from $250M to $500M with investments to date in: Anthropic, Cohere, Hearth.AI, You.com, Humane and Tribble.
Other Alternatives
There are many other alternative AI products that can integrate with Salesforce. However, without the Einstein GPT Trust Layer – you will have the responsibility to ensure the privacy, accuracy, bias and data governance standards are in place to protect your customers.
Sourceforge.net has a list of 24 AI Tools for Salesforce
Sales Outreach Tool: Mailshake
Sales Pipeline Tool: Peopleglass
Booking calls with prospects: MixMax
Developer Tools include: GitHub Copilot, ACCELQ’s AI-Powered No-Code Test Automation Platform
The final word
For a final perspective, tune in to the entertaining Ike Wagh for his recent video: A.I changes everything for Salesforce.
KeyNode Solutions is a full-service Salesforce Consulting Partner company. We deliver the ultimate Salesforce solutions to grow your business: implementation, integration, customization, and development. Our best-in-class Salesforce managed services for ongoing support available nationwide.
