What Should LPs Actually Ask About AI?

Author: By Galit Farkash, Director of Investor Relations Glilot Capital

Date: 16/11/2026

Knowledge-Hub

Notes from the IR Desk

In my previous post, Has Venture Capital Mistaken Speed for Progress?, I argued that AI should compress friction, not judgment.

Since then, I have kept returning to the same issue from the LP side.

AI now appears in almost every venture conversation. Managers describe AI-native portfolios, AI-enabled investment processes and access to the companies building the next generation of infrastructure and applications.

For LPs, the question is no longer whether they have exposure to AI. Most already do.

The more useful question is what that exposure actually represents.

Does it provide access to companies an LP would not otherwise own? Does the manager’s specialization lead to better selection and disciplined pricing? Does its network help turn technical promise into an enterprise product? And can that advantage eventually produce distributions?

These questions reveal much more than the number of AI companies in a portfolio.

Exposure Is Not an Investment Strategy

Demand for AI exposure is not in doubt. S&P Global reported that more than 75% of LPs planned to deploy capital into AI over the following 12 months. At the same time, 82% of the increase in AI investment during 2025 went into rounds of at least $1 billion. Capital is entering the sector quickly, but much of it is concentrating around a relatively small number of companies.

A portfolio may contain several managers with different names, geographies and strategies while still carrying exposure to many of the same companies or parts of the AI stack.

So the first question an LP should ask is not simply, “How much AI exposure does this fund have?”

It is: What exposure do we already own, and what will this manager add that is genuinely different?

The answer may lie in stage, geography, technical domain or access to companies before they become widely competed assets. It may also lie in the ability to identify an important category before the market has agreed on its name.

Without that differentiation, adding another AI-focused fund can increase concentration while creating the appearance of diversification.

Does Specialization Change What the Manager Sees?

Sector expertise is often described through credentials: years in the industry, technical backgrounds, investment history or the size of an advisory network.

Those things matter, but LPs should look for evidence that they change the opportunity set.

Does the manager consistently meet relevant founders before a formal fundraising process begins? Do experienced operators refer founders because they trust the firm’s understanding of the market? Can the team recognize a problem that appears narrow today but could become a substantial category?

This is particularly important in cybersecurity, where AI and security are rapidly becoming one investment conversation.

AI is changing how software is created, how attacks are launched and how security teams operate. Agentic systems introduce new questions about identity, permissions, visibility and accountability. Products are forming around problems many enterprises are still learning how to define.

In that environment, early access is not just a matter of receiving a deck first. It comes from being close enough to CISOs, technical leaders and founders to see where existing tools are failing before the problem becomes a recognized category.

Recent Glilot conversations have focused heavily on this convergence. At Black Hat, founders, CISOs and technology leaders were increasingly addressing AI and cybersecurity as parts of the same problem.

The value of those relationships is not the event itself. It is the information produced by repeated, candid conversations with the people implementing these systems inside real organizations.

That is what LPs should test: not whether a network exists, but whether it improves what the manager sees and when it sees it.

Does Domain Expertise Improve Selection?

Access matters only if the manager knows what to select.

AI has made it easier to build a convincing demonstration. Software development is faster and testing an idea costs less. Those are meaningful advances, but they also make surface-level signals less useful.

A polished product can still lack a durable customer need. Rapid adoption can reflect experimentation rather than dependency. A company can use sophisticated AI without possessing proprietary data, a defensible workflow or a credible route into enterprise budgets.

The same applies in cyber. An emerging threat does not automatically create a viable company. The product still has to fit into a crowded security architecture, survive long procurement processes and earn the confidence of customers who cannot afford careless decisions.

A specialist manager should be able to explain:

  • Which assumptions were validated directly with customers
  • Why the problem requires a new company rather than a feature from an incumbent
  • What becomes more defensible as the company scales
  • How the product moves from a pilot to a system on which customers depend
  • Which technical or market risks could invalidate the thesis

Building an AI prototype is becoming easier; building a lasting enterprise business is not. Execution, domain knowledge and demonstrated customer value therefore carry more weight, not less.

An LP should expect a specialist to explain not only why an investment was made, but what evidence would have caused the firm to walk away.

A specialist earns confidence by knowing where its conviction ends.

How Does the Thesis Reach Distributions?

A sector thesis can be compelling while still being incomplete.

LPs invest in a fund, not a collection of interesting technological predictions. They need to understand how sourcing, selection, ownership and portfolio support translate into realized value.

That question has become more prominent as liquidity remains constrained. Recent manager-selection discussions have put repeatability and DPI back at the center: between two funds with similar total value, the manager that has returned more capital is easier for an LP to support.

This does not mean rushing companies toward premature exits. My original argument still applies: operational speed should not be confused with the time required to build an enduring business.

But patience must be paired with a credible realization strategy. LPs can reasonably ask how a manager thinks about strategic buyers, follow-on reserves and the point at which holding a company creates more expected value than seeking liquidity.

The question is not, “When will every company exit?”

It is: How does this strategy create credible paths from early insight to returned capital?

What LPs Are Really Underwriting

Dynamo’s 2026 survey found that 63% of LPs planned to increase their alternative allocations, and 78% expected to use fund managers. LPs still value manager expertise, and they also want greater transparency and control over where their capital is going.

LPs are not merely underwriting exposure to a popular sector. They are underwriting a manager’s ability to interpret a fast-moving market before consensus forms, maintain discipline while capital rushes in and help technically strong companies become durable businesses.

For a specialist, domain knowledge should create earlier access, improve selection, make diligence more informed and make portfolio support more relevant.

If specialization cannot be observed in those outcomes, it may be positioning rather than an investment edge.

The View from the IR Desk

IR sits in an important place in this conversation.

We hear the excitement around AI, but also the questions beneath it. LPs want access to the companies shaping the next technology cycle. They also want to understand concentration, valuation, defensibility and the path to liquidity.

Those positions are not in conflict. They reflect the same responsibility: allocating capital to genuine long-term advantage rather than following the language of the moment.

The best question for an LP is therefore not, “Does this manager invest in AI?”

It is:

What does this manager understand, access and improve because of its specialization – and how will that advantage ultimately return capital?

Sector exposure is increasingly easy to find.

A repeatable edge is not.

Scroll to Top
Contact Info

Fill out the form below, and we will be in touch shortly.

Contact Information
I am:
Company Stage:
Company Sector: