14: Meta's 10B AI bet
And why enterprise agents are (still) useless.
Hey, it’s Marc.
Salesforce’s new study shows LLMs like GPT-4o still flop at complex CRM tasks, nailing single steps but failing 65% of the time in multi-step workflows. Worse, most agents miss basic data privacy cues.
Just a week back, Apple dropped a study claiming that reasoning models — Claude, DeepSeek, o3 — fail at real reasoning. [Read the full story]
The takeaway: Enterprise AI isn't plug-and-play, yet.
Top Agent signals this week:
Manus AI introduced "Manus Playbook," a tool designed to function like a mini-app
RSM to invest $1B in AI agent automation
KPMG launches WorkBench
Atlassian adds CLI to Rovo Dev AI agent
👉 Get your brand in front of 30,000+ decision-makers — book your ad spot now.
Top Boardroom Reads This Week
How we built our multi-agent research system (Anthropic)
A practical guide to building agents (OpenAI)
Apple’s viral AI paper. Reality or fraud? (Ignacio de Gregorio)
How to build AI Agents (2025 Guide) (Max Berry)
How AI agents are changing financial research and analysis (IBS Intelligence)
Power Moves
Meta invests $10B in Scale AI now valued at $29B
Scale AI raised a major investment from Meta, valuing the company at $29B+, with founder Alexandr Wang joining Meta’s AI efforts. Jason Droege, ex-Uber Eats strategy lead, takes over as interim CEO to accelerate Scale’s data-centric AI solutions. Scale AI isn't just another AI startup. They supply training data and human feedback to 𝘦𝘷𝘦𝘳𝘺 𝘮𝘢𝘫𝘰𝘳 𝘈𝘐 𝘭𝘢𝘣 — including Meta, OpenAI, and the U.S. government. They all depend on Scale's army of contractors to label the data that makes AI actually work. Scale went from $760M revenue in 2023 to $870M in 2024, and they're projecting $2B this year.
So what? Meta’s bet confirms the enterprise demand for human-supervised training data, a key scalability choke point across industries. Expect more “data-as-infrastructure” plays and rising valuations in AI tooling firms. [Analysis]
Mistral’s New AI Stack
Mistral AI unveiled Mistral Compute — a full-stack infrastructure platform offering sovereign, on-prem AI environments, now backed by thousands of NVIDIA GPUs.
So what? This is Europe’s biggest push yet to break from U.S.- and China-dominated AI clouds. For enterprises and governments seeking control over AI deployments (think defense, pharma, finance), Mistral Compute offers a build-it-yourself alternative—complete with orchestration, APIs, and regional compliance baked in. [BLOG]
Vendor Spotlight: Adaline - The LLM Operations Platform
Adaline just opened its doors to everyone after a year of closed development, offering what could be the missing piece in enterprise AI deployment. [RELEASE]
Why it matters: Most companies are stuck in AI pilot purgatory. They can build a demo, but can't ship reliable AI products at scale. The gap? Operations tooling that treats AI like the new software category it is.
What Adaline does: Think GitHub for AI instructions instead of code. The platform handles four critical functions:
Iterate: Collaborative prompt development (no more Slack threads with prompt versions)
Evaluate: Systematic testing before deployment
Deploy: Production-ready AI application management
Monitor: Real-time performance tracking
The strategic angle:
For CTOs: Adaline treats prompts like code — version control, testing, deployment pipelines. Your engineering practices finally extend to AI.
For business leaders: Cross-functional teams can collaborate on AI applications without technical bottlenecks. Product managers and domain experts work directly alongside engineers.
Zooming in: Adaline is offering $1M in API credits to first 100 enterprise workspaces. That's roughly $10K per qualified team — meaningful budget for initial deployment and testing, creating a low-risk evaluation window.
Customer base: Salesforce, DoorDash, Discord, HubSpot, McKinsey & Co.
Developer traction: 100K+ developers during closed beta
Market timing: Going public now, ahead of the enterprise AI deployment wave
Strategic verdict:
Buy if: You're past AI experimentation and need production operations tooling
Wait if: You're still figuring out basic AI use cases
Watch if: You're building internal AI ops tools (potential build vs. buy decision)
Adaline could be the Rails moment for AI operations — the infrastructure that makes the next wave of applications possible.
Top AI Vendors This Week
👉 We’re tracking 1000s of AI and blockchain vendors so you find the right vendors, partners, and bet on the right tech. Sign up for early access.
More from 51:
That’s all for today.
Thanks,
Marc & Team