Microsoft unveils three new foundational models!

Sponsored by

How Jennifer Aniston’s LolaVie brand grew sales 40% with CTV ads

The DTC beauty category is crowded. To break through, Jennifer Aniston’s brand LolaVie, worked with Roku Ads Manager to easily set up, test, and optimize CTV ad creatives. The campaign helped drive a big lift in sales and customer growth, helping LolaVie break through in the crowded beauty category.

Learn more

What’s a week without yet another product release? An impossibility 😼. 

Forward thinkers, welcome to another issue of the Neural Frontier.

This week, we’re covering Microsoft’s new foundational models, Slack’s huge AI-driven update, and OpenAI's financial moves. 

Let’s dig in!

In a rush? Here's your quick byte: 

🤖 Microsoft unveils three new foundational models!

💻  Salesforce rolls out an AI-filled update for Slack.

💸 OpenAI closes a deal to raise $122B!

⚡ The Neural Frontier’s weekly spotlight: 3 AI tools making the rounds this week.

🤖 Microsoft unveils three new foundational models!

Microsoft is doubling down on its in-house AI efforts with the release of three new foundational models — covering text, voice, and image generation.

The move signals a clear shift: even with its deep partnership with OpenAI, Microsoft is building its own AI stack to compete more directly.

⚙️ The three new models

Microsoft’s MAI (Microsoft AI) team introduced:

• 📝 MAI-Transcribe-1: Speech-to-text across 25 languages, and reportedly 2.5× faster than Azure’s existing transcription tools.

• 🎙️ MAI-Voice-1: Audio generation model, which can generate 60 seconds of audio in 1 second. It also supports custom voice creation

• 🖼️ MAI-Image-2: Image (and reportedly video) generation model.  Previously tested in MAI Playground, now rolling out more broadly

All three models are being deployed via Microsoft Foundry, with some also available in MAI Playground.

💸 Competing on cost and practicality

Microsoft is positioning these models around lower cost vs competitors, practical, real-world use cases, and a “human-centered” approach to AI. 

Example pricing highlights:

• Transcription → from $0.36/hour

• Voice generation → from $22 per 1M characters

• Image generation → from $33 per 1M output tokens

The pitch is a strong performance, but cheaper and more usable at scale.

🧪 Built by a new AI unit

These models come from Microsoft’s MAI Superintelligence team, led by Mustafa Suleyman.

The team was formed in late 2025 and is focused on building multimodal AI systems, production-ready models, and AI that aligns with how people actually work and communicate.

🤝 Still partnered with OpenAI — but more independent

Despite this push, Microsoft isn’t abandoning its relationship with OpenAI. Instead, it’s taking a hybrid approach:

• Build its own models

• Continue using OpenAI’s models

• Choose the best tool for each use case

Think of it like its chip strategy: 👉 Microsoft both builds and buys.

This move reflects a broader shift in the AI race:

• Big players don’t want to rely on a single provider

• Owning the full stack (models + infra + apps) is becoming critical

• Cost efficiency is emerging as a major competitive lever

Overall, Microsoft is positioning itself as a full-fledged AI lab and platform player.

💻  Salesforce rolls out an AI-filled update for Slack.

Salesforce is turning Slack into something much bigger than a chat tool — rolling out 30 new AI features that push it toward becoming a full-fledged work execution platform.

At the center of it all: a heavily upgraded Slackbot, now evolving into a true AI agent for the workplace.

🤖 Slackbot becomes an agent, not just a helper

Slackbot isn’t just answering questions anymore — it can now draft emails, schedule meetings, search inboxes and Slack channels, transcribe and summarize meetings, and generate action items automatically

This builds on earlier updates, but now positions Slackbot as something closer to a workflow engine inside Slack.

🧩 Reusable “AI skills” (the biggest unlock)

The standout feature is reusable AI skills.

These let users define tasks once and reuse them across workflows.

Examples:

• “Create a budget” → pulls relevant info, builds a plan, schedules a meeting

• “Prep client update” → gathers context, drafts message, assigns follow-ups

The key idea is a prebuilt skill library from Salesforce and custom skills users can create themselves — all triggered with simple commands inside Slack. This turns Slack from chat to a command center for repeatable workflows

🔗 Built on MCP + Agentforce

Slackbot now acts as an MCP (Model Context Protocol) client, meaning it can connect to external tools and systems. It integrates with Agentforce (Salesforce’s AI agent platform) and other enterprise apps and internal systems

This allows Slackbot to route tasks to the right tools or agents, pull context from across systems, and execute workflows end-to-end without manual coordination.

🖥️ Beyond Slack: context from your entire workday

Slackbot can now operate outside Slack itself, pulling signals from your calendar, conversations, deals, and CRM data, and even desktop activity patterns. Using this context, it can:

• Suggest next steps

• Draft follow-ups

• Flag important tasks proactively

Clearly, Slack is moving from a Communication tool to an AI-powered work platform. The goal? Seemingly to make Slack the place where work happens, not just where it’s discussed.

💸 OpenAI closes a deal to raise $122B!

OpenAI has closed a massive $122 billion funding round, valuing the company at $852 billion — one of the largest private raises ever. 

The move comes as the company prepares for a likely IPO, and everything about this announcement reads like a public market dress rehearsal.

🏦 Who’s backing it — and why it matters

The round was co-led by heavyweight investors including SoftBank and Andreessen Horowitz, with participation from firms like TPG, T. Rowe Price, and strategic players such as Amazon, Nvidia, and Microsoft.

Notably, about $3 billion came from retail investors, and OpenAI is being added to ARK Invest ETFs — giving everyday investors exposure before it goes public. That’s unusual for a private company and shows how aggressively OpenAI is trying to broaden its shareholder base ahead of IPO.

⚙️ Where the money is going

This isn’t a defensive raise — it’s fuel for an expensive race. OpenAI is pouring capital into:

• AI chips and large-scale compute infrastructure

• Data center expansion

• Hiring top-tier AI talent

It also expanded a $4.7B credit line (currently unused), signaling that it’s optimizing for scale and flexibility, not survival.

📊 The numbers they want investors to see

OpenAI shared metrics that feel very intentional — almost like an S-1 preview.

• $2B in monthly revenue

• 900M+ weekly active users

• 50M+ paid subscribers

• Ads already generating $100M+ ARR in weeks

On the business side, enterprise revenue now makes up 40% of total, and the company expects it to match consumer revenue by 2026 — a sign that OpenAI isn’t just a consumer app, but a growing enterprise platform.

🧠 The positioning: OpenAI as the “AI superapp”

OpenAI is increasingly framing itself not just as a model provider, but as the default interface for AI. That means:

• Chat → search → workflows → execution, all in one place

• Expansion into agentic systems powered by models like GPT-5.4

• New monetization layers, including ads and enterprise tooling

OpenAI wants to own how people interact with AI entirely, not just power it behind the scenes.

⚡ The Neural Frontier’s weekly spotlight: 3 AI tools making the rounds this week. 

1. 🖥️ Perplexity Computer is an AI-powered digital worker that autonomously plans, delegates, and executes entire multi-step workflows — so you don't have to.

2. 🎨 StoryMotion is an AI-powered animation tool that turns diagrams, docs, and ideas into polished explainer videos in minutes — no video editing experience needed.

3. 👨‍💻 Verdent is an AI-native coding agent that plans, writes, and executes complex software tasks in parallel — so developers can ship faster without losing focus.

Another week…

Another barrage of product releases. A shocker? At this point, not anymore 😏. 

Truly, it is an amazing time to be alive. Not just to witness these products, but to test and integrate them into our workflows.

We can’t wait to see what’s on the menu next week. And we’re gonna go out on a limb and say, neither can you. 

As usual, you know where to find us, same time, next week. See you then! 👋

PS: If you’re the “friend” this mail was forwarded to, and you enjoyed it, hit the Subscribe button to see more content like this every week 🙂.