Keye is doing DD with AI right

FIRST PASS

2024: The State of Generative AI in the Enterprise (Menlo)

Menlo Ventures’ State of Generative AI in the Enterprise report revealed dramatic shifts in how enterprises have approached AI adoption and investment.

The headline number—$13.8B in enterprise GenAI spending in 2024, up 6x from $2.3B in 2023—signals a clear transition from experimentation to execution.

The report, which surveyed 600 U.S. enterprise IT decision-makers, shows that while 72% of leaders anticipate broader AI adoption, more than a third still lack a clear implementation vision. This tension between growing investment and strategic uncertainty defines the current state of enterprise AI adoption.

Some fascinating findings:

1. 60% of GenAI investments still come from innovation budgets, but 40% now come from permanent allocations

2. Organizations have identified an average of 10 potential use cases, with 24% prioritized for near-term implementation

3. The application layer is growing faster than the foundation model layer, with $4.6B in spending (up 8x from 2023)

4. Enterprises are prioritizing ROI (30%) and industry-specific customization (26%) over price (just 1%) when selecting tools

Merlin’s Notes: The shift from innovation to permanent budgets suggests we're entering a phase in which GenAI moves from "nice-to-have" to "must-have." This means portfolio companies need concrete AI strategies now, not later.

It should come as no surprise that enterprises emphasize ROI and customization; however, it’s still worth reiterating that many are taking a sanguine view despite the eye-watering investment sums. For what it’s worth, I continue to believe that vertical AI presents one of the greatest value-creation levers that operators should pay attention to.

The report also reveals that implementation costs (26%) and data privacy (21%) are the top reasons for failed pilots. Operators should not overlook implementation planning and data governance frameworks when rolling out AI initiatives across a portfolio.

The goal should be to avoid the common pitfall of underestimating the technical and organizational complexity of AI adoption. Here are the key actions we recommend:

1. Shift from opportunistic to strategic AI adoption by developing comprehensive roadmaps

2. Focus on industry-specific use cases where AI can create demonstrable value

3. Build robust data infrastructure and governance frameworks before scaling AI initiatives

4. Plan for the full cost of implementation, including change management and training

HIGHLIGHT

Secret to AI Profitability is Training it for Highly Specialized Tasks (Bloomberg)

Bloomberg recently published an insightful piece revealing a crucial trend in AI monetization: the growing demand for highly specialized domain experts to train AI models for specific industry applications.

The article highlights how AI service providers are moving beyond basic data labeling to tackle complex, industry-specific challenges—from training algorithms to detect early-stage lung cancer to helping John Deere's subsidiary optimize pesticide usage in agriculture.

• Companies like iMerit (backed by three Silicon Valley billionaires) are building teams of domain experts across the globe

• Scale AI, the largest player in this space, recently raised funding from Meta and Amazon at a $14B valuation

• Specialized experts (like radiologists training AI models) can earn $1,200 for a few hours of work

With the stakes as high as they are, the article reinforces that AI's real value isn't just in general automation, but in solving specific, high-value problems where domain expertise meets technology.

JARGON BUSTER

USE CASE SPOTLIGHT

How Torrens University Used GenAI to Transform Online Learning at Scale

Torrens University, Australia's fastest-growing university, partnered with Microsoft to revolutionize its online learning platform using generative AI, resulting in remarkable efficiency gains and cost savings.

The goal? To standardize and enhance the accessibility of online courses while creating a superior digital learning experience for students across all devices.

In just 16 weeks of discovery and implementation, the university's transformation team achieved what would have typically required thousands of hours of manual work. They developed a custom solution using Microsoft Azure OpenAI that analyzed and standardized over 1,200 courses and 60,000 web pages.

The impact was substantial and immediate:

1. Saved up to 20,000 hours of manual work

2. Reduced cost by A$2.4 million

3. Achieved 100% mobile responsiveness

4. Met Web Content Accessibility Guidelines 2.0 AA rating

5. Standardized curriculum delivery across all courses

The project's success relied on several key strategies:

• Strategic AI Implementation: Rather than using AI for basic chatbot functionality, Torrens developed a sophisticated GenAI model to analyze and reorganize course content at scale.

• Human-in-the-Loop Approach: Academics reviewed AI-standardized content to ensure quality and accuracy.

• Comprehensive Integration: The team built a custom integration with Microsoft Teams for seamless class creation and enrollment management.

As Eoghan Hogan, Director of Transformation at Torrens University, noted: "In a cost-sensitive world, the fact that we achieved 100 percent mobile responsiveness and a completely accessible and inclusive environment while saving A$2.4 million is phenomenal."

This case demonstrates how organizations can leverage GenAI to not just reduce costs but also significantly improve service delivery and user experience. The success has prompted Torrens to explore additional GenAI applications, including enhanced student support systems and personalized content delivery.

TOOL OF THE WEEK

Keye

If you've ever been involved in any kind of investment due diligence, you know the drill: mountains of data, tight timelines, and the constant worry that you might miss something crucial. Enter Keye, a YC-backed startup that's reimagining how PE firms conduct due diligence using AI.

Keye bills itself as more than just another AI summarization tool. Instead of simply condensing information, it performs complex analyses across entire data rooms—think cohort analysis, retention metrics, and risk identification—all while maintaining audit-ready documentation.

What sets it apart is its PE-native approach. Founded by former Vista and Goldman dealmakers who've managed over $20B in transactions, Keye understands the specific challenges of PE due diligence. The platform can:

• Transform raw data into thousands of analyses instantly

• Process entire data rooms and deal folders to surface critical insights

• Generate organized data packs with tagged risks and opportunities

• Export findings to fully linked Excel spreadsheets with complete audit trails