Meta Llama 3 Launches Open Source, GPT-4 Vision Reaches All Plus Users, Copilot Studio Goes GA

#3 Microsoft: Copilot Studio Reaches General Availability

Microsoft Copilot Studio, the low-code custom AI agent builder, is now generally available. Build agents with specific knowledge sources, behaviors, and escalation logic. Pricing: $200 per 25,000 messages per month.

Executive angle: Start the 30-day free trial and build the customer service triage agent from Article 5 before committing to a subscription.

Read more ->

#4 Meta: Meta AI Arrives in WhatsApp Search Bar and Across Meta Apps

Meta rolled out Meta AI, powered by Llama 3, into the WhatsApp search bar and across Instagram, Facebook, and Messenger in the US. Available without any separate app download.

Executive angle: Test Meta AI in WhatsApp for quick questions and compare response quality to ChatGPT on similar prompts.

Read more ->

#5 Google: Gemini 1.5 Pro Expanded Access in Workspace AI Studio

Google expanded access to Gemini 1.5 Pro in Workspace AI Studio for developers and business teams. The 1.5 Pro model offers a 1 million token context window, enabling analysis of entire document repositories in a single prompt.

Executive angle: Evaluate Gemini 1.5 Pro via Google AI Studio for any use case requiring analysis of very long documents or large data sets.

Read more ->

02 / Top AI Updates | 3 Deep Dives

# 01 / 03

Meta Llama 3 and Meta AI: What Open-Source AI Means for Your Business Stack

Meta released Llama 3 on April 18, 2024 in 8B and 70B parameter versions, both freely available for download and commercial use. The 70B version matches GPT-3.5 Turbo on multiple benchmarks. For executives, the strategic question is not about the Meta AI consumer product: it is about when open-source AI is the right choice over proprietary API subscriptions.

Business Impact

Eliminates: Per-token API costs at high volume and third-party data processing requirements for sensitive data workflows

Time back: Variable: significant cost savings for organizations with high-volume automation running on GPT-3.5 class models

Best for: Organizations with high-volume AI automation needs, strict data residency requirements, or domain-specific fine-tuning needs

Executive Action Steps

• Try Meta AI in WhatsApp or at meta.ai to experience Llama 3 quality without technical setup.
• Identify any high-volume internal task currently running on a paid API (GPT-3.5 or similar) that is cost-sensitive.
• Evaluate Llama 3 8B on cloud infrastructure (AWS, Azure, or Google Cloud) for that specific task.
• Compare accuracy and cost versus the current API approach before switching production workloads.

Why it matters: When an open-source model reaches GPT-3.5 quality, the calculus for high-volume tasks with strict data control needs shifts materially. The decision is not about raw capability: it is about volume, cost, and data residency.

Deep dive: Meta Llama 3 and Meta AI: What Open-Source AI Means for Your Business Stack ->

# 02 / 03

GPT-4 with Vision: Analyzing Images, Charts, and Documents in ChatGPT

GPT-4 Turbo with Vision is now available to all ChatGPT Plus subscribers. Upload any image directly in the chat and ask GPT-4 to analyze it: financial charts, screenshots, scanned documents, whiteboard photos, and competitive marketing materials. The analysis happens in plain-language conversation without any technical setup.

Business Impact

Eliminates: Manual extraction of insights from visual data before writing a briefing or narrative

Time back: 15 to 30 min per chart or report that previously required manual analysis and narrative drafting

Best for: Executives who regularly review financial dashboards, analyst reports, or competitor materials and need written narrative from visual data

Executive Action Steps

• Open ChatGPT Plus and select GPT-4 in a new conversation.
• Click the image icon in the chat input bar and upload the chart, screenshot, or document page.
• Send a specific analytical prompt immediately after the image uploads (see Article 4 for the full prompt).
• Ask follow-up questions without re-uploading: GPT-4 holds the image in context for the full conversation.

Why it matters: Vision analysis converts visual data into structured, shareable narrative in under 2 minutes. The output lands directly in board decks, investor updates, or executive emails with minimal editing.

Deep dive: GPT-4 with Vision: Analyzing Images, Charts, and Documents in ChatGPT ->

# 03 / 03

Microsoft Copilot Studio is now generally available. A visual, low-code builder for custom AI agents with specific knowledge sources, conversational behaviors, and escalation logic. Non-developers can build, test, and deploy a working agent in an afternoon. The highest-ROI starting point is a customer service triage agent trained on product documentation and FAQ.

Business Impact

Eliminates: Human handling of tier-1 customer support questions that have known answers in existing documentation

Time back: 15 to 25 hours of human support time per month for teams handling 500 or more support contacts

Best for: Organizations with a defined tier-1 support queue where 25 to 45 percent of questions have documented answers

Executive Action Steps

• Start the 30-day Copilot Studio free trial at copilotstudio.microsoft.com.
• Create a copilot, name it, and add knowledge sources: upload your top FAQ and product documentation.
• Configure escalation logic: a topic that triggers on frustration phrases and collects customer information.
• Test with your 20 most common customer support questions, fill knowledge gaps, then deploy via the web chat widget.

Why it matters: Tier-1 support deflection is among the highest-ROI first applications for a custom AI agent because the knowledge base is defined and the use case is measurable.

Deep dive: Microsoft Copilot Studio: Build Custom AI Agents Without Coding ->

03 / Pro Tip

GPT-4 Vision: Financial Chart to Board-Ready Narrative in 2 Minutes

Tools: ChatGPT Plus (GPT-4 Vision) | Time: 2 Min | Level: Beginner

Upload any financial chart or dashboard screenshot to ChatGPT (GPT-4 selected) and use this five-section prompt to get key trends, flagged anomalies, anticipated board questions, a ready-to-use narrative paragraph, and a recommended action. The output converts a raw chart into a complete briefing asset in under 2 minutes.

The Prompt

Why it works: Five structured sections turn a 20-minute manual analysis into a 2-minute AI output. Section 4 pastes directly into board decks. Section 3 becomes pre-meeting prep. Section 2 flags items for finance review.

Best for: Executives who regularly review financial dashboards, earnings charts, analyst reports, or vendor metrics and need to convert visual data into written briefing content.

Deep dive: GPT-4 Vision: Financial Chart to Board-Ready Narrative in 2 Minutes ->

04 / Productivity Gem

Copilot Studio Customer Service Triage Agent: Built in an Afternoon

Tools: Microsoft Copilot Studio | Saved: 15-25 Hr/Month

A four-step process to build a no-code customer service triage agent in Copilot Studio that handles tier-1 questions from your product documentation and escalates complex cases to human agents with context already collected. Deployable as a web chat widget on your support page in under an afternoon.

• Create the agent at copilotstudio.microsoft.com, name it, and add a description defining its customer support role.
• Add knowledge sources: upload your top FAQ documents and product documentation PDFs or connect your SharePoint support site.
• Create an escalation topic that triggers on frustration phrases and collects customer name, email, and issue description before confirming human follow-up.

The Prompt

Time savings: Deflects 25 to 45 percent of tier-1 support contacts for teams handling 500 or more monthly inquiries, saving 15 to 25 hours of human support time per month.

Deep dive: Copilot Studio Customer Service Triage Agent: Built in an Afternoon ->

Open source hit a new capability benchmark, vision analysis went from preview to standard, and custom agent building became a no-code afternoon project. The access floor for serious AI capability dropped again this week.

Executives who act this week will upload one financial chart to ChatGPT Vision and run the Article 4 prompt before the next board meeting, start the Copilot Studio free trial, and test Meta AI in WhatsApp on 3 questions they would normally take to ChatGPT.

Forward this to one executive you respect. They will thank you next week.

Premium Upgrade