Return to site

Dreamforce 2023.

Gimme an 'A', gimme an 'I', waddawegot...?

· TRA Blog

Day one at Dreamforce (possibly the only global tech event where grown adults queue to hug a bear, Einstein, an elephant and other mascots) where Salesforce has doubled down on AI with major announcements around generative AI solutions.

Salesforces’ next generation of Einstein offers Salesforce users the ability to bake generative AI into their workflows, underpinned with a very strong focus on ethical AI and trust.

Einstein Copilot, according to Salesforce is an ‘out-of-the-box conversational AI assistant built into the user experience of every Salesforce application’ and Einstein Copilot Studio essentially gives Salesforce customers the ability to customise their Copilot solution. All of this sits on a fresh re-architected platform, ‘Einstein 1’, that provides a framework for handling metadata.

There’s certainly a huge amount of noise in the market around GenAI and the stand out differentiator for me is Salesforce’s Einstein Trust Layer. From our work in this space we know enterprises have concerns around how much access vendors have to the customer’s data, how that data is being use to train large language models (LLMs), hallucinations created by GenAI tools and auditability of
content and outputs. Salesforce have been super clear that through Trust Layer customers retain control and oversight of their data (Salesforce has a zero data retention approach), can undertake toxicity analysis and have the ability to undertake analysis and audits of any generated content.

Whilst some other genAI solutions point to an ethical or some other form of ‘trust’ capability around genAI, Salesforce has been an early mover in this, having established clear principles and tenets pointing to ethics and accountability.   

The other differentiator for me is the Trailhead community. Salesforce can point to an extremely passionate user community and the company has a growing number of AI-focused certifications and courses available to users and partners.

Less clear is the commercial model. Details are a little sparse right now (although perhaps more will come over the next day or so) however as with any ‘aaS’ consumption model, the devil is absolutely in the details. Cloud bill shock experiences mean businesses are vastly more knowledgeable about potential pricing gotchas and will be looking for transparency on this front. The lack of clarity certainly won’t help organisations looking for support around building the business case, at a time when tech
budgets in APJ are under pressure from softening economic conditions.

broken image