#37: Hit me with your best AI- Website AI Podcast, SWARM OpenAI release, Perplexity Spaces, NVIDIA LLM Release, Napkin AI

The newsletter and GTM AI Podcast are sponsored by GTM AI Academy AND AI Powered GTM .

Business Impact> Learning Tools

We have the SLACK COMMUNITY around AI , come hang with your fellow GTM professionals, get insights, and content free to help in your teams and roles.

The GTM AI Demo Tools Library is a free resource for you to view AI tools and demos to see what is out there to help your team or to help you in your own workflow

We changed the bi-weekly Collab on Thursdays to now every other Thursday happening THIS week in the morning and afternoon, if you want to attend, let me know and I will invite you!

Please note, there is SO much happening every week in AI, I literally had 25 updates and chose to narrow it down to not overwhelm you but there is A LOT going on.

Now with all that being said, lets move forward with todays newsletter which is:

We have #37 GTM AI Podcast with Eldar Agayev of Hachly diving into AI chatbots, the next generation of website experience that drives sales.

OpenAI SWARM release

Perplexity Spaces Release

NVIDIA LLM Release

GTM AI Tool of the week: Napkin AI

Some AI posts from this last week in case you missed it:

Sales teams are lagging in adoption of GEN AI

Your CRM is lying to you and how to fix it

And now the podcast:

You can go to Youtube , Apple , Spotify , or here on Linkedin as well as a whole other host of locations to hear the podcast or see the video interview.

AI Chatbots: The Future of Website Sales Conversion

Hey everyone, I just had an incredible conversation with Eldar Agayev of Hachly and I've got to tell you, this guy's story and what they're doing with AI chatbots is pretty fascinating.

Eldar started coding chatbots at 16 over in Ukraine, building stuff for Counter-Strike of all things. He ended up moving to the UK, got into computer science, and while most people were trying to figure out what to do after graduation, this guy had already built 15 different chatbots with over a million users on Telegram.

Now, what really got me excited was how they're completely flipping the script on how we think about website chatbots. Forget those annoying customer service bots we all hate - Hachly is all about driving actual sales. They've built this system that can be up and running in 15 minutes (not kidding) and actually understands what visitors are doing on your website to have real, contextual conversations.

The business impact? They just helped an enterprise dev agency land a $65K contract through their chatbot. That's not just answering questions - that's generating real revenue. And here's what's wild - Eldar sees a future where we're not even clicking around websites anymore; we'll just be having conversations with them. Imagine just telling a website what you're looking for instead of hunting through pages of content.

This is something I am seeing more of, is augmentation to the buyers journey and this isn't about replacing salespeople. It's about making them more effective by handling all that initial qualification and engagement. Your sales team gets to focus on what they do best - closing deals with qualified prospects - while the AI handles all the early-stage stuff that eats up so much time.

We dove deep into some amazing tech stuff too, like how they're thinking about multi-agent frameworks and the future of AI interfaces. Eldar shared a case where they turned 6,000 website visitors into a $50-60K pipeline for one of their clients - all through their chatbot.

Whether you're a startup or an enterprise, if you've got a website (and who doesn't?), there's a way to make this work for your business.

I've got to tell you, I haven't been this excited about a sales tech innovation in a while. It's not just another tool - it's a completely different way of thinking about how we engage with website visitors and turn them into actual opportunities.

- Evolution from building Telegram chatbots to creating an AI-powered sales platform

- Focus on sales conversations rather than customer service

- 15-minute deployment time without complex setup needed

- Use of context-aware AI to understand user intent and website behavior

- Success case: $65,000 enterprise contract secured through chatbot

- Vision of future website interaction being primarily conversation-based

- Multi-agent framework potential for future development

- ROI example: 6,000 website visitors generating $50-60,000 pipeline

- Emphasis on complementing rather than replacing human sales teams

Key Quotes

"We ultimately help businesses to generate more sales from their website by engaging every visitor with context-aware messaging... We ultimately make the experience for visitors better."

"Being a salesperson in the next 10 years will change... I think we just give you more leads. So there's more things to work with. And we don't actually close deals... but we drive more of these prospects from where they were just browsing the website."

"I imagine in 10 or 15 years, we're going to have websites where you have a button at the top right corner, you just click on it and you can actually talk to the website instead of browsing... You can actually talk and ask, okay, I want to find an article about X."

"Overall, the cost of software will go down dramatically because of AI... I have AI on every front that helps me optimize my schedule and my time. So yes, I think AI for scheduling more leads and getting you more revenue will help any company of any size."

OpenAI Swarm Multi-Agent Capabilities

OpenAI released for testing SWARM on GitHub .

Overview of OpenAI Swarm

OpenAI’s Swarm is an experimental framework aimed at creating, orchestrating, and deploying multi-agent systems efficiently. Unlike traditional approaches, Swarm introduces a lightweight, scalable solution, focusing on dynamic agent collaboration and tool execution. This design allows developers to create sophisticated agent-driven processes while maintaining control and minimizing complexity.

Swarm from OpenAI is not just a simple chatbot framework; it’s a powerful tool for creating modular, task-specific agents that collaborate and coordinate dynamically. It leverages OpenAI’s Chat Completions API but goes a step further by enabling sophisticated handoffs and context management. Each agent within Swarm is equipped with unique instructions and functions, allowing them to specialize in different tasks and share context seamlessly.

Swarm’s real strength lies in its combination of lightweight operation, customization options, and scalability. Its applications go beyond customer support or sales automation; Swarm’s flexibility enables complex, scalable implementations in various domains.

1. Lightweight and Scalable:

? Swarm is designed to be lightweight and client-side, avoiding heavy backend reliance, making it easy to deploy at scale without complex infrastructure dependencies.

2. Agent Coordination and Handoff:

? Swarm’s core mechanism revolves around the concept of agents and handoffs. Agents encapsulate specific instructions and tools, and can intelligently pass control to other agents based on the context. This allows for modular design and simplifies building multi-step processes.

3. Customizable and Transparent:

? Developers have fine-grained control over context, steps, and tools through flexible agent configurations. Each agent can execute tasks, invoke external functions, update context variables, and switch control without maintaining state between calls. This leads to greater transparency and predictable behavior.

4. Multi-Model and Multi-Functionality:

? Each agent can be fine-tuned for specific tasks or workflows, providing dedicated functions and instructions while remaining independent. This modular approach enables more effective handling of complex or diverse tasks.

5. Designed for Complex and Real-World Solutions:

? Swarm is optimal for scenarios where encoding all logic into a single model or prompt is challenging. It uses the Chat Completions API to offer context-dependent and tool-based actions, supporting varied real-world applications like customer support, personalized recommendations, and more.

How OpenAI Swarm Can Help Various Teams:

1. Sales Teams:

? Benefit: Swarm can automate routine sales support tasks by acting as a proactive assistant that manages multi-step inquiries, identifies potential leads, routes them to the right resources, and provides data-driven insights.

? Example: A Swarm setup could handle initial triage of inbound requests, categorize potential leads based on interest, and transfer complex queries to a human sales agent, streamlining sales efforts and improving lead conversion rates

2. Customer Success:

? Benefit: Using Swarm, teams can deploy support bots that identify and handle different types of customer issues dynamically. This ensures accurate routing and faster resolution times.

? Example: A multi-agent setup could differentiate between simple troubleshooting and billing issues, enabling agents to hand off conversations efficiently while keeping records intact.

3. Marketing Teams:

? Benefit: Marketing teams can leverage Swarm to create dynamic, personalized customer interactions that incorporate a variety of tools to manage campaigns, gather feedback, and analyze consumer behavior.

? Example: Marketing agents could engage customers based on context, track sentiment, and provide tailored offers in real-time, improving engagement rates and campaign effectiveness.

4. Enablement:

? Benefit: Swarm can enhance enablement strategies by acting as a virtual coach or roleplay partner. It can automate parts of training and feedback loops by simulating different customer interactions.

? Example: Enablement professionals could configure agents to conduct simulated sales scenarios, guiding new hires and offering feedback, reducing onboarding times.

5. Revenue Operations (RevOps):

? Benefit: Swarm’s ability to handle context-based workflows and functions can assist RevOps in automating key data collection and reporting processes.

? Example: Agents could automatically retrieve and summarize key sales metrics, track pipeline stages, and generate customized reports for decision-makers, enhancing operational efficiency.

6. GTM Leaders:

? Benefit: For GTM leaders, Swarm provides an opportunity to deploy custom strategies and workflows that are optimized for scale without sacrificing control.

? Example: Leaders could leverage Swarm’s modular agents to implement strategic pivots, automate processes across departments, and ensure alignment with overarching GTM objectives.

Testimonials & Case Studies:

OpenAI’s documentation and repository provide examples showcasing Swarm’s applications. For instance, there’s a customer support bot example with specialized agents that work together to offer a seamless service experience. In an airline scenario, Swarm is used to categorize and resolve passenger inquiries dynamically, demonstrating its potential in real-world support automation.

Additionally, the framework has received positive feedback for its low overhead and adaptability. Developers appreciate its client-side design and transparent operation, making it easier to debug, iterate, and scale multi-agent systems.

Overall, OpenAI’s Swarm stands out as a framework for businesses looking to orchestrate sophisticated agent-based solutions, enhancing productivity, efficiency, and customer interactions.

Perplexity AI: Spaces Deep Dive

Perplexity AI’s Spaces offers a powerful organizational tool that acts as a dynamic knowledge hub, allowing users to efficiently group related content by topic or project. Whether you’re working individually or with a team, Spaces empower you to create tailored repositories where research, files, and insights are kept together and easily searchable.

Key Features & Benefits of Spaces

1. Organized Research: Spaces allow you to group Threads and files by specific projects, topics, or interests, keeping related conversations and documents in one place for easy reference and deeper context.

2. Seamless Collaboration: With Spaces, you can invite team members or stakeholders as collaborators or viewers. Collaborators can add new Threads or follow up on existing ones, enhancing collaboration.

3. Integrated Search Capabilities: For Perplexity Pro users, Spaces search can be customized to use either web sources, your uploaded files, or a combination of both. This tailored search option delivers more contextually relevant and comprehensive answers.

4. File Upload & Storage: Users can upload key documents to Spaces and use them as sources for contextually accurate answers. Perplexity Pro users can upload up to 50 files per Space, while Enterprise Pro users have additional flexibility.

5. AI-Enhanced Context: You can set up a customized AI persona to handle inquiries in a specific way, allowing for more tailored responses that align with the project or research’s context.

Use Cases for Different Teams

1. Sales Teams:

? Benefit: Spaces help sales teams organize their knowledge base with customer-specific Threads and store files like contracts, proposals, or emails. Sales reps can quickly reference previous conversations, streamline communications, and get more personalized responses based on comprehensive file and web-based insights.

? Example: A sales team could use Spaces to maintain project-specific knowledge hubs, store signed contracts, and easily access customer preferences from previous conversations during follow-up calls.

2. Customer Success:

? Benefit: Spaces allow customer success teams to set up dedicated repositories for each client, organizing historical data and files. When customers have inquiries, team members can retrieve detailed context without having to sort through separate systems.

? Example: Customer success reps can quickly search within a Space dedicated to a specific customer, drawing on uploaded documents and relevant web-based insights to resolve issues faster and more effectively.

3. Marketing Teams:

? Benefit: Marketing professionals can use Spaces to store and organize campaign assets, research materials, and consumer insights. They can set up collaborative Spaces for running marketing experiments, gathering team feedback, and refining strategies based on contextual searches.

? Example: The marketing team could run Spaces that group insights from different campaigns, keeping creative briefs and performance data side-by-side for fast iteration and insight gathering.

4. Enablement:

? Benefit: Enablement professionals can create Spaces that serve as training repositories, organizing threads and files related to sales training, playbooks, and learning resources. This enables trainers to dynamically update enablement content and keep all training materials together.

? Example: Enablement leads can easily update and manage training threads, providing personalized instructions and insights by combining course files with the latest industry trends from the web.

5. RevOps:

? Benefit: Revenue Operations teams can leverage Spaces to centralize operational metrics, reports, and analytics. They can set up custom AI personas to answer internal queries using uploaded spreadsheets or operational documentation.

? Example: RevOps can maintain Spaces for quarterly reports, pulling insights based on uploaded files and external analytics tools to answer complex revenue-related queries in real-time.

6. GTM Leaders:

? Benefit: For GTM leaders, Spaces provides an organized hub for managing strategy documents, playbooks, and key project threads. Leaders can share key files and discussions with other team members for review and provide strategic guidance efficiently.

? Example: GTM leaders could create a Space for GTM strategy discussions, enabling stakeholders to reference key playbooks, review strategies, and access context-driven insights during critical meetings.

Practical Prompts Examples for Spaces

1. Plan a business trip:

? “I have an upcoming business trip to Japan and Australia. Cross-reference my saved travel docs and original packing list to update my packing list for this trip, considering weather conditions and latest travel guides.”

2. Summarize educational content:

? “Summarize the main points from last week’s reinforcement learning lecture. Add insights from related public web resources.

3. Review contract proposals:

? “Compare contractor proposals for my kitchen renovation, considering pricing, timeline, and materials.”

NVIDIA OPEN FRONTIER CLASS MULTIMODAL LLMs

The paper “NVLM: Open Frontier-Class Multimodal LLMs” introduces NVIDIA’s NVLM-1.0, which brings three distinct architectures: NVLM-D (Decoder-only), NVLM-X (Cross-attention), and NVLM-H (Hybrid). These are designed for vision-language tasks and general text-based tasks, marking a production-grade shift towards multimodal AI models. The NVLM-1.0 framework excels in processing high-resolution imagery using a unique tile-based dynamic tagging approach, enhancing OCR and multimodal reasoning. Outperforming top open-access models like LLaVA, it sets a new standard in vision-language alignment and multimodal capabilities.

Key Features and Technical Innovations:

1. Three Architectures for Flexibility:

? NVLM-D: A decoder-only architecture focusing on efficient text-based generation.

Think of this as a focused engine for generating text efficiently. It’s designed to handle straightforward text-based tasks like writing responses or drafting content quickly.

? NVLM-X: Employs cross-attention mechanisms, ideal for deep multimodal interaction.

This architecture is like having a connector that deeply analyzes and integrates information from both text and images. It’s perfect for understanding complex documents or visuals combined with text.

? NVLM-H: Combines strengths from both decoder and cross-attention architectures for robust multitasking.

A versatile multitasker that blends the capabilities of both NVLM-D and NVLM-X. It’s built to seamlessly switch between pure text tasks and complex multimodal (text and image) processing.

2. High-Resolution Image Processing:

? NVLM introduces a tile-based dynamic tagging approach to enhance OCR capabilities, dynamically understanding and extracting key elements from detailed images or documents.

3. Extensive Model Training:

? The model utilizes diverse datasets with specialized curation for text and multimodal data, enhancing visual-language alignment.

4. Superior OCR and Multimodal Reasoning:

? The advanced architecture enables NVLM to surpass benchmarks against leading models like LLaVA and Llama 3-V, indicating better OCR capabilities and integration of text-image reasoning.

5. Open-Source and Scalable:

? By offering comprehensive architecture descriptions, open-sourced training details, and dataset accessibility, NVLM-1.0 promotes reproducibility and collaboration.

Potential Impact on GTM Teams:

1. Sales Teams:

? NVLM’s OCR capabilities can streamline sales documentation by extracting critical information from contracts, agreements, or product images, speeding up negotiations and deal closures.

2. Customer Success:

? Enhanced document recognition allows customer success teams to easily retrieve key client documents, troubleshoot visual data issues, and personalize solutions faster.

3. Marketing Teams:

? The hybrid architecture and multimodal capabilities allow marketing teams to create more engaging campaigns with deep context from both visual and textual data.

4. Enablement:

? NVLM’s adaptability allows enablement professionals to create interactive training materials integrating visuals and text, elevating the learning experience.

5. RevOps:

? RevOps can leverage NVLM’s comprehensive insights from sales, marketing, and customer success data to make data-driven recommendations that align visual analytics with key performance metrics.

6. GTM Leaders

? Leaders can deploy NVLM to analyze both qualitative and quantitative insights simultaneously, refining strategies based on a full-spectrum view of customer interactions and feedback.

Here’s a comparison of NVIDIA NVLM-1.0, Claude, and ChatGPT tailored for GTM professionals:

1. Internet Access:

? NVLM-1.0: No direct internet access; primarily designed for integrated multimodal tasks.

? Claude: No direct internet access during live interaction, but trained on broad datasets.

? ChatGPT: Can be integrated with browser plugins for dynamic web access.

2. Context Windows:

? NVLM-1.0: Limited by architecture; mainly optimized for multimodal data processing.

? Claude: Large context window (~100,000 tokens in Claude 3) for long conversations.

? ChatGPT: Context window ~32,000 tokens in ChatGPT-4 Turbo for handling extended interactions.

3. Multimodal Capabilities:

? NVLM-1.0: Specialized in handling high-res images and OCR with advanced dynamic tagging.

? Claude: Predominantly text-focused, with some expansion towards multimodal understanding

? ChatGPT: Supports image input (beta) but not as specialized in OCR or high-res tasks as NVLM.

4. Testing and Performance:

? NVLM-1.0: Benchmarked to excel in vision-language and OCR tasks, outperforming LLaVA and Llama 3-V.

? Claude: Excels in reasoning and long-form text generation but not tested in vision tasks.

? ChatGPT: Well-rounded in text and evolving in image recognition; top performer in general conversational AI.

5. Language and Multimodal Training:

? NVLM-1.0: Tailored multimodal architecture designed for precise text-image alignment.

? Claude: Generalized language model with reinforcement for human-like text-based interactions.

? ChatGPT: Trained on diverse text sources with gradual improvements in multimodal capabilities.

6. Key Advantages for GTM Professionals:

? NVLM-1.0: Ideal for high-res image recognition, OCR-heavy use cases, and content that needs precise text-image relationships.

? Claude: Best for longer discussions, strategic planning, and deep text analysis with a human-like approach.

? ChatGPT: Versatile for real-time research, customer interactions, and creative content; excels in accessibility and integrations.

In essence, GTM leaders can leverage NVLM for multimodal tasks requiring image and text precision, Claude for strategic planning and deep reasoning, and ChatGPT for dynamic and broad-spectrum conversational AI. This comparison aligns each model’s strengths with potential needs in sales, marketing, enablement, and operations.

GTM AI TOOL OF THE WEEK Napkin AI

Napkin.ai is an AI-driven visual content creation tool that turns your text into visuals like diagrams, flowcharts, and infographics, making it an ideal companion for storytellers, educators, marketers, and content creators.

ALL of the extra visuals in the newsletter today was made by Napkin AI all for free during their beta.

Overview:

Napkin helps users quickly generate visuals directly from written text, eliminating the need for manual graphic creation. It’s ideal for professionals who want to present complex ideas more clearly and visually.

Key Features:

? Text to Visuals: Transforms written content into visuals (infographics, diagrams, and flowcharts) without needing to write prompts.

? Fully Editable Visuals: Users can customize visuals with fonts, colors, icons, and connectors.

? Exports & Integrations: Export visuals as PNG, SVG, and PDF files. Direct integration with platforms like Google Slides, Word, and Notion.

? Collaborative Editing: Allows real-time editing and collaboration for shared projects.

? Sparks Feature: Instantly visualize concepts with auto-generated visuals based on your content.

? Simple UI: An intuitive and modern interface that focuses on ease of use for everyone, from designers to marketers.

Here is an example from the Prompting Guide on the GTM AI Academy :

Impact on Teams:

1. Sales: Sales professionals can quickly create visual aids to simplify pitches, making it easier to explain complex sales processes and value propositions.

2. Customer Success: Napkin’s visuals help CS teams illustrate solutions, processes, or tutorials for clients, leading to clearer communication and better onboarding or troubleshooting guides.

3. Marketing: Marketers can leverage the tool to create social media posts, infographics, and impactful presentations that resonate with audiences and drive engagement.

4. Enablement: Sales enablement teams can develop training materials with visual aids that enhance learning and retention, reinforcing sales and customer success processes.

5. RevOps: RevOps can streamline complex process diagrams or flowcharts to illustrate operational workflows or data-driven strategies, improving team alignment.

6. GTM Leaders: Leaders can visually map out go-to-market strategies and initiatives, enabling clearer communication and alignment among cross-functional teams.

You may or may not be seeing more visuals as a result of this cool tool ;) And yes visuals today were provided from Napkin!

Let me know what else you wanna hear or see!