Salesforce and Microsoft Boost Conversational AI in Contact Centers and Enterprise-wide

Ask not what ChatGPT can do for you, ask whose branded generative AI and large language model-based solutions are most appropriate for your company. Then tailor your prompts to make sure the response explains how GPT-based resources can improve customer experience, speed completion of mundane tasks and introduce operational efficiencies throughout the business. Then brace yourself for detailed responses that note how Salesforce and Microsoft, two giants of cloud-based computing, employ their own branded renditions of #GenerativeAI based on OpenAI’s library or LLMs and generative AI resources.

CRM is My Copilot

The story starts on March 6 with the introduction of Microsoft Dynamics 365 Copilot, which employs Azure-based renditions OpenAI’s resources to “jumpstart” or accelerate composition of messages or responses to queries that arise in the course of everyday business conversations. Microsoft has been refining the “copilot” model for years, having made the GitHub Copilot generally available to the developer community in June 2022 after providing a technical preview in the form of an extension to popular “integrated development environments” (IDEs) roughly a year earlier.

The “co-pilot” paradigm is powerful. As Marcus Schmidt, Principal Program Manager of Dynamics 365 Customer Service, explained in a briefing “it is the agent’s 24-hour assistant” always there to answer questions, make suggestions and draft responses for the agent to review. He was referring to how it appears in the agent workspace in Microsoft’s Digital Contact Center offering. Yet Monday’s announcement provided concrete use cases that lean heavily on Microsoft Dynamics 365 Copilot to automate business workflows and automation in CRM as well as ERP systems by applying text generation, sentiment analysis and workflow automation to accelerate completion completion of repeatable, mundane tasks.

Microsoft has packaged the copilot for implementation throughout the enterprise. In early February it showcased the integration of Azure-based GPT 3.5 into its Viva Sales platform. It employed the model to generate suggested email responses for salespeople to use in common use cases, such as replying to specific questions about a products price or availability, taking into account promotions and deadlines. This is now marketed as “copilot” for Dynamic 365 Sales and Viva Sales.

For the Marketing department, Copilot will be baked into Dynamic 365 Customer Insights and Dynamics 365 Marketing. Core to this use case is the ability for marketing professionals to use their own words to “curate highly personalized and targeted customer segments” by carrying out a dialogue with the customer data platform. They can also use a “query assist” feature to use their own words to describe a target segment. Use cases for Copilot permeate Dynamics 365 Business Center, where it can be used to streamline creation of product listings and catalogs for online commerce. In this case the marketer can tailor product descriptions for online storefronts that can be published to Shopify “in just a few clicks”. Microsoft’s vision for Copilot in the Supply Chain Center includes its ability to flag internal issues such as weather that can effect the supply chain and provide predictive insights to help planners draft email alerts to share with impacted partners or customers.

Nuance Making ChatGPT Part of the Mix

In a closely related development, Microsoft’s Conversational AI subsidiary Nuance announced a new enhancement to Nuance Mix, the development and run-time platform for chatbots, speech-enabled IVRs and and other Enterprise Intelligent Assistants. Called “Conversation Boosters”, powered by GPT, it enables developers who use Mix to create virtual assistants or conversational IVRs that can provide answers or responses to a wider range of topics than was previously possible.

The new service is available as a preview today and the workflows for specific use cases are works in progress. In general, the expectations is for GPT’s large language model to track down and incorporate domain-specific information from websites, documents or live agent logs. Then the generative element of GPT will provide accurate, meaningful responses in a conversational way. This approach will evolve over time and, based on all-too-common experience with LLMs that hallucinate, enterprise professionals are reticent to allow GPT-based resources to operate without human oversight and curation of output.

For Salesforce, GPT Augments Einstein

Microsoft’s announcement precede by one day Salesforce’s claim that Einstein GPT is “The World’s First Generative AI for CRM.” Setting the headline aside, you’ll perceive that Salesforce and Microsoft have defined well-grounded use cases for OpenAI’s Generative AI throughout every business enterprise. For Salesforce that means “out-of-the-box” Generative AI capabilities in all of its clouds as well as the ability to generate instant conversational summaries based on conversations with external customers as well as internal discussions over Slack. As described in the launch event at Trailblazer DX 2023, “ChatGPT in Slack adds another layer of leverage where conversations happen faster, threads get explained faster, and work gets done more efficiently”.

The objectives of Einstein GPT are remarkably similar to Microsoft’s co-pilot. Employees will turn to Einstein to “generate personalized emails for salespeople to send to customers, generate specific responses for customer service professionals to more quickly answer customer questions, generate targeted content for marketers to increase campaign response rates, and auto-generate code for developers. Instead of leveraging Microsoft Dynamics and Nuance, Salesforce will be augmenting brands like Tableau, MuleSoft and Slack by employing OpenAI’s models to learn more about real world usage and respond appropriately.

Excitement over the availability of Einstein GPT was palpable among attendees at Trailblazer DX 2023 (#TDX23). The value of employing Einstein to draft sales letters, answer questions or generate code seemed obvious. The ability to train Einstein with a company’s own corpus of data seemed especially valuable, as did Salesforce’s commitment to security, privacy and ethics. But the proof will be in the real-world implementations. Of greatest importance will be defining how to keep live employees involved in the curating and refining Einstein’s answers. It is a topic that was brought up on stage at #TDX23, but unlike Microsoft’s Copilot, Einstein GPT’s answers were not regarded as suggestions. They were the finished product: a landing page, a chart in Tableau, scheduling a Slack meeting.

Bottom line: Generative AI’s Future Starts Now; Humans Included

Microsoft and Salesforce are giving the CRM, Contact Center and Collaboration worlds a glimpse of the possible ways Generative AI will impact key, measurable business outcomes. They will not be alone in providing easy to implement toolsets to improve customer experience, employee productivity and business efficiencies. Adoption of services like summarization will be swift. Discovery of areas where human intelligence is required to tailor prompts and refine output will be speedy as well. At base, LLMs and Generative AI platforms are like “autocomplete” on steroids. Their increased use is inevitable, but human “feedback” will always be required to tailor responses, ensure accuracy and prevent bias



Categories: Conversational Intelligence, Intelligent Assistants, Articles

Tags: ,