Today's edition is about survival.

So let's start with a question:

BCG just dropped their report on the widening AI value gap, surveying 1,250 companies across the globe.

AND

Only 5% of companies are generating substantial value from AI. Just 5%.

Meanwhile, 60% are burning cash on AI initiatives that deliver essentially zero measurable business impact.

But here's the part that should terrify every executive reading this:

That 5% isn't just winning. They're pulling so far ahead that the gap is becoming unbridgeable.

These companies spend 26% more on IT. They dedicate 64% more of their budget to AI. They expect 2x the revenue increase compared to everyone else.

The winners create a compounding cycle. AI gains fund more AI capabilities. Those capabilities generate more gains. Those gains fund even more capabilities.

The laggards are locked in the opposite cycle. Lack of value means no reinvestment. No reinvestment means falling further behind.

Scale this over 18 months, and you see entire industries bifurcating into winners and losers.

Before we go deeper, let's cover what happened this week.

1. Raycast Keyboard for iOS

Raycast launched a custom keyboard for iOS that puts AI commands, snippets, dictation, and quicklinks directly into any app. The keyboard integrates Whisper-powered voice dictation in 50+ languages and syncs with Raycast for Mac. You can trigger AI chat, paste saved snippets, or access quicklinks without leaving Messages, Mail, or any app. Works system-wide on iOS.

My take: This solves the biggest limitation of Raycast on iOS. Now you can trigger AI commands without app-switching. This is how AI should work on mobile.

2. cto.new, Completely Free AI Coding Agent

Engine Labs launched cto.new with $5.7M in funding as the first completely free AI coding agent. Provides unlimited access to GPT-5 Codex, Claude Sonnet 4.5, and Gemini Pro. Integrates with GitHub, Linear, Jira, and Slack for task planning and code review. Already outperforms Claude Code on Terminal Bench and is used by 6,000+ developers at Adobe and Docker.

My take: Direct shot at Cursor and every paid coding assistant. If they can sustain this economically, the entire AI coding market has a problem.

3. alphaXiv: NotebookLM for arXiv

alphaXiv announced a NotebookLM-style feature that turns dense arXiv papers into conversational summaries. It pulls context across thousands of related papers to explain motivations, connect findings to state of the art, and surface key insights like a professor who’s read the whole field. The goal: higher‑fidelity research communication and faster comprehension directly on top of arXiv.

My take: This is the right UX for preprints. If it reliably cites sources and avoids hallucinations, it could become the default way researchers and practitioners skim, compare, and understand new work without slogging through PDFs.

Back to the report, and the brutal reality it reveals.

Picture two retail companies. Both have the same revenue. Same market position. Same access to technology in 2024.

After 18 months, Company A is still "piloting." Company B is 3.6x ahead on shareholder returns and 2.7x ahead on return on invested capital.

I see this pattern everywhere. The companies that spread AI thin across dozens of use cases generate almost no value. The companies that go deep on a few core workflows generate transformative returns.

The difference is focus.

Here's the stat that exposes why most AI initiatives fail.

70% of potential AI value comes from core business functions. Sales, marketing, manufacturing, R&D, supply chain.

Yet most companies start with the easy stuff. An AI chatbot for internal IT support. AI-powered expense report processing. GenAI tools for writing emails.

These things create some value. But they won't drive shareholder returns.

A global beauty company built the industry's first large-scale virtual beauty assistant. Integrated into websites across 20 markets and eight brands. AI-powered consultation with real-time personalization.

Deployed in under 12 months. Expected $100 million in incremental revenue. Double the ROI of traditional e-commerce.

A leading electronic device manufacturer created a centralized "company store" of agentic AI solutions across 200+ factories.

Core agents can be repeatedly used across different workflows. They support 80% automation in complex operational tasks. $300 million in projected EBIT impact.

Most companies are optimizing emails. The winners are inventing entirely new revenue streams.

And here's what frustrates me. Everyone knows where the value is. 70% in core functions. Yet most AI budgets still go to support functions because they're easier to implement.

The report reveals something that wasn't even a category last year.

Think about that timeline. We went from "agents don't really exist" in 2023 to "agents represent nearly a third of AI value" in 2028. That's a 36-month transformation.

But here's the trap. 46% of companies are "experimenting" with agents. Only 16% are actually deploying them with tangible value.

The gap between experimentation and value is where companies die.

A leading insurance company partnered with AI infrastructure providers, foundation model companies, and data integration specialists to deploy agentic AI across underwriting and claims.

They orchestrated a multivendor ecosystem to reshape how the entire function operates. Dramatically faster underwriting and enterprise-wide transformation through coordinated agents.

Honestly, I think we're underestimating how fast this will move. The report projects 29% by 2028. I think we'll hit that by late 2026.

BCG's research reveals that most roadblocks involve people, organization, and processes. Not technology.

The 10-20-70 rule. 10% of your focus should be on algorithms. 20% on technology infrastructure. 70% on people, processes, and organizational change.

But most companies obsess over which model to use. Which vendor to pick. They treat AI like an IT project.

Meanwhile, the actual barriers:

Future-built companies establish joint business-IT ownership. 1.5x more likely than laggards.

They appoint chief AI officers. 3x more likely.

They upskill more than 50% of employees in AI. Laggards only upskill 20%.

Everyone wants the sexy AI models. Nobody wants to do the hard work of organizational change. But that's where all the value lives.

Here's the uncomfortable truth about AI and jobs.

Some jobs will disappear. That's already happening.

Future-built companies are six times more likely to dedicate structured time and programs for AI learning. They're planning to upskill 50%+ of staff in 2025. They're redesigning workflows so humans and AI agents work together with clear roles and accountability.

Laggards are hoping people "figure it out" while rolling out ChatGPT licenses.

This creates a terrifying dynamic. As AI gets better, the performance gap between top performers and average workers narrows in structured tasks. AI automates the routine stuff, so everyone performs similarly.

But in complex, judgment-driven work, AI amplifies the edge of top performers. The best get better. The average stagnate.

If you're not systematically preparing your workforce for this shift, you're building a company full of people whose skills are actively depreciating.

I keep thinking about what happened during the internet transition. Companies that retrained their workforce thrived. Companies that ignored the skills gap died. This will be faster and more brutal.

More than 50% of future-built firms operate on a single enterprise-wide data model. Only 4% of stagnating companies do.

You can't run sophisticated AI on fragmented, siloed, poor-quality data.

The companies generating value spent years building unified data foundations with strong governance, clear ownership, and centralized oversight.

The winners started building their data foundations in 2018-2020. They're reaping the benefits now. If you're starting today, you're already years behind.

This is the dirty secret. Everyone wants to deploy the latest LLM. Nobody wants to spend two years cleaning up their data infrastructure.

The 5% have opened a value gap that is actively widening. Every quarter they reinvest AI gains into more capabilities, more talent, more infrastructure.

The window to catch up is measured in months, not years.

Future-built companies achieve 9-12 month time-to-impact. Laggards take 12-18 months.

Future-built companies have 62% of AI workflows already deployed. Laggards have 12%.

The performance gap on revenue growth is 1.7x. On three-year TSR, it's 3.6x. On return on invested capital, 2.7x.

Ask yourself: Are you in the 5%, or are you part of the 60% pretending pilots equal progress?

The market doesn't care about your AI roadmap. It cares about your P&L.

And right now, for most companies, AI is just expensive experimentation with no returns.

The value gap isn't coming. It's already here. And it's swallowing companies whole.

Until next time,
Vaibhav 🤝

If you read till here, you might find this interesting

#AD 1

Deel’s simplified a whole planet’s worth of information on global hiring. It’s time you got your hands on our international compliance handbook where you’ll learn about:

With 150+ countries right at your fingertips, growing your team with Deel is easier than ever.

Get the guide

#AD 2

Deel’s free HR Budget Planning Guide helps you plan smarter, scalable HR budgets with benchmarks, best practices, and a global-ready checklist.

Build a Smarter HR Budget