Welcome to this week’s No Jitter Roll, our regular roundup of product news in the communication and collaboration spaces. Leading off this week, we highlight: some of the news out of AWS re:Invent 2024, including the Amazon Nova family of models and updates to Amazon Bedrock, Verint's launch of a CX/EX scoring bot and AVOXI's two-way international calling with local Caller ID capability.
In partnerships, we highlight the integration of Salesforce Contact Center with Amazon Connect, ServiceNow's expanded collaboration with AWS Amazon Bedrock, RingCentral's new Zendesk integration's, Movework's Agentic AI-based collaboration with Box, the call analytics capabilities delivered via a SIPPIO-Akixi collaboration and the dynamic, AI-powered avatars arising from AKOOL and LiveX.
In reports, we cover an Amdocs survey that provides some insight into consumer and communications service provider expectations around Agentic AI.
Click the links below to jump to specific sections:
AWS re:Invent took place this week In Las Vegas. NJ covered the key Amazon Connect news, as well as the Salesforce and ServiceNow partnerships in this edition of NJR. In this post, NJ highlights the new Amazon Nova AI model family, improvements to Amazon Bedrock and updates to Amazon Q Business.
AWS introduced its new generation of foundation models called Amazon Nova which will be available in Amazon Bedrock. These models include:
Amazon Nova models are available in Amazon Bedrock in the US East (N. Virginia) AWS region. Amazon Nova Micro, Lite, and Pro are also available in the US West (Oregon), and US East (Ohio) regions via cross-Region inference. Amazon Nova Premier is still in training with expected availability in Q1 2025.
All of the Nova models are integrated with Amazon Bedrock (which also provides access to more than 100 third-party models), which means that customers can experiment with and evaluate not just the Nova models but others, as well.
The Nova models all support custom fine-tuning which allows customers to point the models to examples in their own proprietary data that have been labeled to boost accuracy. The models also support distillation, which enables the transfer of specific knowledge from a larger, more capable ‘teacher model’ to a smaller, more efficient model that is accurate, but also faster and cheaper to run (because it is smaller). (See here and here for more on distillation.)
“What I like about the Nova releases is they're not just focused on ‘frontier scale’ models, but are instead focused on smaller models that can be adapted to specific tasks and knowledge contextualities,” said Bradley Shimmin, Chief Analyst, AI & Data Analytics with Omdia. “I think Amazon is just catching up on the model side. They had their Titan family of models for quite some time and those did not receive the same degree of support from the community that we've seen for the Meta Llama models, IBM Granite, and even from Google with its Gemini and Gemma family of models.”
Amazon Bedrock is a fully managed service for building and scaling Gen AI applications with foundation models, both Amazon’s own models as well as many from third parties. AWS’s announcements for Bedrock introduce several new capabilities several of which are highlighted here.
This safeguard, available via Amazon Bedrock Guardrails, uses mathematical, logic-based algorithmic verification and reasoning processes to verify the information generated by a model so that outputs align with known facts and aren’t based on fabricated or inconsistent data. That is, it helps prevent factual errors due to model hallucinations. (The linked site provides a detailed description of how it works.) This feature is available today in preview in Amazon Bedrock Guardrails in the US West (Oregon) AWS Region.
This new capability helps customers implement safeguards for generative AI applications by configuring policies for denied topics, content filters, word filters, PII redaction, contextual grounding checks, and Automated Reasoning checks, to tailor safeguards to specific use cases and responsible AI policies. The existing content filter policy in Amazon Bedrock Guardrails can now be used to detect and block harmful image content across varied categories including hate, insults, and violence. Multimodal toxicity detection with image support is available today in preview in Amazon Bedrock Guardrails in the US East (N. Virginia, Ohio), US West (Oregon), Asia Pacific (Mumbai, Seoul, Singapore, Tokyo), Europe (Frankfurt, Ireland, London), and AWS GovCloud (US-West) AWS Regions.
There are several new capabilities within this category. With Amazon Bedrock Data Automation (preview), you can build automated intelligent document processing (IDP), media analysis, and Retrieval-Augmented Generation (RAG) workflows. Potential insights that can be derived include video summaries of key moments, detection of inappropriate image content, automated analysis of complex documents, and more.
Additionally, Amazon Bedrock Knowledge Bases now processes multimodal data (text and visual elements) and supports GraphRAG. A knowledge graph is a collection of data that, as AWS defines it, “offers a structured representation of real-world entities and their relationships.” (See here for more.) RAG is a technique that essentially ‘points’ an LLM at external knowledge sources so that the LLM returns information more relevant to the task at hand – e.g., accessing an enterprise’s product FAQ versus Wikipedia. GraphRAG essentially points the LLM at a knowledge graph. Per AWS, GraphRAG provides several advantages: the knowledge graph already contains curated and interconnected information, and its structured nature enables more precise querying and retrieval, which helps ensure the LLM produces better responses.
Amazon Bedrock Knowledge Bases also now support structured data retrieval. This capability extends a knowledge base to support natural language querying of data warehouses and data lakes so that applications can access business intelligence (BI) through conversational interfaces.
These new capabilities make it easier to build comprehensive AI applications that can process, understand, and retrieve information from structured and unstructured data sources. For example, a car insurance company can use Amazon Bedrock Data Automation to automate their claims adjudication workflow to reduce the time taken to process automobile claims, improving the productivity of their claims department.
“What they're basically doing is helping you set up a RAG system that you don't have to wire up by hand. You don't have to figure out some of the subtleties that are still very much in the realm of data science – like, how do I pick the right embedding model? How do I select the correct overlap for how I'm chunking this information?” Shimmin said. “Instead, they’re basically saying: We are going to wire this stuff together so that you can more rapidly adapt our models, or our partner models, to include the knowledge and tools that that you need to get whatever use case built that you're trying to build.”
This allows customers to build and orchestrate multiple AI agents to work together on complex multi-step tasks that require specialized skills. The AWS blog post provides a detailed example of how a ‘social media campaign manager agent’ (the blue box below) can be created and used to coordinate the activity and action of several specialized sub-agents – one for creating content, one for timing when the content should post, etc. Multi-agent collaboration in Amazon Bedrock is available today in preview in all AWS Regions that support Amazon Bedrock Agents, except AWS GovCloud (US-West).
Available soon, Amazon Q Business will add the capability to simplify the creation and maintenance of complex business workflows. With this capability, users only need to describe their desired workflow using natural language, upload a standard operating procedure (SOP), or record a video of the process being performed. Amazon Q Business then uses Gen AI to automatically author a detailed workflow plan from those inputs. Then, with the recommended workflow, users can review, test, modify, or approve.
With Amazon Q Business plugins, users will be able to connect to third-party apps and perform specific tasks related to supported third-party services directly within their web experience chat. These plugins are accessible through Amazon Q Apps. Examples of plugins include Salesforce, Microsoft Exchange and Teams, Google Calendar, etc.
Amazon Q Business can now be used directly from a web browser and other supported messaging and collaboration applications. Administrators need to enable the browser extensions for users of the Amazon Q Business application. The new Amazon Q Business plugins are available today in all AWS Regions where Amazon Q Business is available. The new capability to orchestrate workflows in Amazon Q Business will be available in preview soon.
With respect to Bedrock, Omdia’s Shimmin said that what AWS is doing is making their platform much more adaptable in terms of model selection.
“That to me is critical for the enterprise. What they need is the ability to have total ownership and control of transparency and governability of their solutions without getting locked in. What we're really seeing from with this set of announcements, is this recognition that you don't have to pledge allegiance to a model manufacturer. You may love Anthropic Claude, but maybe that's not the best solution for project 21 out of 51 inside the business. It may not even be the right one. If it's the right one today, maybe it's not the right one tomorrow,” Shimmin said.
“What matters most is that control with flexibility…to be able to swap in and out, to move models in and out of a given use case or a given project without breaking the project or having to refactor all your code and all the other technical debt you would have to if you were trying to switch, for example, from Anthropic to OpenAI.”
For a complete list of the many announcements see the AWS News Blog and their top announcements post.
The provider of CX automation solutions announced the addition of a new CX/EX Scoring Bot running on the Verint Open Platform. This bot uses Verint’s AI models to provide real-time scoring of customer experience, agent satisfaction and the emotional connection between customers and agents for every call. Supervisors have access to a real-time dashboard that provides a CX/EX summary and trends by agent and by team, configurable real-time alerts, the ability to listen to live or recorded calls and the ability to search for calls.
The cloud voice software provider for contact centers announced that its offering, AVOXI TrueLocal, can deliver two-way international calling with guaranteed local caller ID across more than 100 countries. AVOXI also noted that TrueLocal coverage has expanded to India, Pakistan, Sint Maarten (Dutch Part), Macedonia, Republic of the Congo, Ethiopia, Burkina Faso, Liberia and Bahrain.
According to Barbara Dondiego, CEO of AVOXI. “Seeing a recognizable area code and receiving regulatory-compliant calls from trusted brands increases confidence in voice as the preferred communication channel.”
AWS announced the Preview of Salesforce Contact Center with Amazon Connect. With this integration, companies can integrate their Salesforce CRM data and agent experience with Amazon Connect’s voice, digital channel, and routing capabilities.
ServiceNow and AWS announced a new connector that enables the use of multimodal models developed and trained on Amazon Bedrock for Gen AI-powered workflows in the Now Platform. This connector allows customers to use their choice of third-party models, based on their specific workflow needs, such as summarization, advanced analytics, or code generation. Data remains private and secure through ServiceNow and AWS.
With the launch of Zendesk Talk Partner Edition for RingEX, RingEX’s voice capabilities and call data are now directly embedded within the Zendesk customer support platform. (In March 2024, RingCentral MVP was rebranded as RingEX.) The Zendesk Talk Partner Edition for RingEX was designed for departmental use by informal contact center users in sales, marketing, and support.
The provider of a ‘universal AI copilot’ or search and automation announced its integration with Box. Powered by Moveworks' Agentic Reasoning Engine, customers can ask for the information they need (e.g., document, surfacing an insight, analyzing information) in Box using plain language prompts.
SIPPIO provides PSTN calling in Microsoft Teams Phone and Zoom Phone; Akixi provides call and contact analytics. This partnership will allow service providers and resellers to offer businesses real-time insights into call activity, performance, and customer engagement.
AKOOL provides a Gen AI platform for creating business videos; LiveX AI provides human-like AI Agents for customer-/employee-facing interactions. This collaboration combines LiveX AI’s real-time conversational AI Agents with AKOOL’s dynamic avatar technology to create virtual agents “see, hear and respond like humans.”
Amdocs released a global study, Rethinking Brand and Customer Experience in the Agentic Era, which examined consumer and communications service providers (CSPs) expectations for the agentic era. The study is based on survey, interviews and focus groups comprised of 120 CSP leaders and 7,025 consumers aged 18-69 in 14 countries across North America, Europe, and Asia. Amdocs collaborated with McCann Tech agency and Coleman Parkes in the research.
In this ‘agentic era,’ AI agents autonomously perform tasks, make decisions, and interact with customers on behalf of businesses and individuals. The report argues that AI agents are gradually becoming brand representatives. According to Gil Rosen, Chief Marketing Officer, Amdocs, each “brand interaction with a customer via AI agents should be seen through the paradigm of brand, context, and customer.”
Amdoc’s report found that 48% of consumers expressed concerns that AI agents could be used primarily for cost-saving purposes. The report also compared consumer responses with CSP responses. For example, 45% of consumers expressed concern about the use of AI for customer care and sales, while CSPs estimated this concern to be nearly 60%. Put another way, consumers in this study were more receptive to AI agents than the CSPs estimated.
Conversely, consumers have high expectations of AI agents, which exceed those of CSPs in every surveyed aspect: 80% of consumers expect a high level of empathy from AI agents, compared to just 43% of CSPs. This gap is also significant for agent professionalism (85% vs. 44%), quick issue resolution (87% vs. 51%), and achieving first-time resolution (74% vs. 27%).
In case you missed them, here are some of our top stories:
We’re excited to announce that registration is now open for Enterprise Connect, March 17-20, 2025, in Orlando, Florida! Next year’s event will be better than ever, featuring an enhanced Monday program with an inspiring Luminary Keynote, coupled with industry keynotes; New Strategic Impact Summits; and more enterprise-led sessions. This is your best opportunity in 2025 for learning, peer-to-peer networking, and product sourcing. Don’t miss it!
