At roughly the same time, Reuters examined Unitree, the Chinese robotics company that has turned advances in quadruped robotics into lower-cost, increasingly capable machines.

The juxtaposition raises a useful question for investors and founders working in physical AI:

What if the scarce asset is not ultimately the intelligence itself, but the ability to turn that intelligence into something that reliably performs useful work in the physical world?

That distinction may become increasingly important as the underlying technology improves and becomes more accessible.

Intelligence is only one part of the problem

Physical AI is attracting substantial capital. Crunchbase estimates that companies in the category raised $47.4 billion across 521 deals during the first half of 2026, nearly four times the dollar amount invested during the second half of 2025.

The logic is easy to understand. AI can increasingly perceive, reason and adapt, while industries such as construction, manufacturing, logistics and agriculture continue to face pressure around labor availability, cost and productivity.

But a better robotics model does not necessarily produce a better robotics business.

We have already learned a version of this lesson in software. Access to a strong foundation model does not, by itself, create a defensible application company. Distribution, workflow, proprietary data, customer behavior and product execution still determine where value accrues.

Physical AI adds several more layers.

The commercial chain looks something like:

Intelligence → hardware → reliability → workflow integration → distribution → customer ROI

A company can be strong at the first step and still fail at the steps that follow.

The more important question may therefore be: as robotic intelligence becomes more capable and more available, where in that chain will durable advantage actually accumulate?

Why Gravis is interesting

What stands out about Gravis is not simply the autonomy technology. It is the problem the company has chosen to solve.

Rather than asking construction companies to replace existing equipment with an entirely new robotic architecture, Gravis is developing technology that can automate heavy machinery already operating on jobsites.

That changes the adoption problem.

A construction company does not need to buy into a vision of a fully autonomous jobsite. The proposition can be much simpler: take an expensive machine you already own and make it more productive.

This may be a useful pattern to watch across physical AI. Some of the stronger companies may not begin by asking customers to redesign an industry around a robot. They may begin by inserting intelligence into workflows that already exist.

The underlying technology can be complex while the customer proposition remains simple.

Customers are generally not buying autonomy for its own sake. They are buying higher utilization, lower labor costs, fewer accidents, faster throughput, or another measurable economic outcome.

That changes the question investors should ask.

How much should we care about how impressive the robot looks, versus how quickly the customer gets paid back?

The physical world is less forgiving

Physical environments introduce constraints that software businesses do not face in the same way.

Jobsites change. Weather changes. Terrain changes. Equipment is damaged. Sensors fail. Components wear out. Humans improvise.

A system that performs well in a controlled demonstration still has to function in an environment where conditions are variable and failure has real-world consequences.

This is where the physical AI problem becomes broader than intelligence alone.

Once the model works, a second problem begins: can the company make the system work reliably, repeatedly and economically in the customer’s environment?

That requires a wider set of capabilities: hardware engineering, field operations, installation, supply chains, maintenance, channel relationships, customer training and industry-specific workflow knowledge.

These capabilities can look operational rather than technological. But that distinction may become less useful over time.

If an operating system allows a company to deploy faster, perform more reliably, collect better proprietary data and improve the product faster than competitors, then operational capability may itself become part of the moat.

Unitree raises a related question

The Unitree story makes the same point from another direction.

Reuters reported that important technical foundations behind some quadruped robotics capabilities originated in U.S.-funded research that was openly published. The research mattered, but access to the research alone did not determine who built a scaled commercial product.

Unitree combined technical capability with manufacturing, component sourcing, iteration speed and lower cost.

That should broaden how investors think about defensibility in physical AI.

Proprietary technology still matters. But perhaps the more useful question is not whether every important layer is proprietary.

It is whether the integrated system becomes difficult to reproduce.

A company may use increasingly available models, components or research and still build a meaningful advantage through data, manufacturing, distribution, integration and installed base.

If that is true, diligence focused too heavily on the uniqueness of the underlying IP may only tell us part of the story.

What this could mean for investors

Physical AI requires a broader diligence lens.

The first question remains obvious: does the technology work?

But that may increasingly become table stakes. Investors also need to understand what happens after the demo.

How long does deployment take? How much customization is required for each new customer? Who installs and maintains the system? How often does it fail? How much must the customer’s workflow change? How long until the customer earns an economic return?

And one question may matter more than most:

Does the 100th deployment become meaningfully easier than the first?

That gets to the heart of scalability.

A services-heavy robotics business can grow revenue while becoming more complicated with every customer. A stronger model may behave differently: every deployment generates operating data, the data improves performance, better performance reduces intervention, deployment becomes faster, margins improve and the installed base grows.

The loop could look like this:

Deployment → proprietary real-world data → better performance → easier deployment → larger installed base

If that flywheel exists, deployment is no longer just an implementation function. It becomes part of the compounding advantage.

This also raises a more nuanced diligence question. Investors often discount operational complexity because it can signal poor scalability.

But in physical AI, some forms of operational complexity may create defensibility.

The key is distinguishing complexity that compounds advantage from complexity that simply consumes people and capital.

What this could mean for founders

For founders, the implications are somewhat different.

There is an understandable temptation in robotics to pursue the broadest possible vision: general intelligence, general-purpose robots and systems capable of operating across many environments.

Some important companies will be built that way.

But not every physical AI company needs to start there.

There may be considerable strategic value in beginning with one expensive problem, one clear workflow and one customer for whom the ROI is obvious.

The questions become more practical.

Can the robot do one valuable job well, every day? Can the customer deploy it without redesigning the organization? Does the product become harder to displace once integrated? Does using it generate proprietary data that improves the system? And does solving the initial problem create a credible path into adjacent workflows?

A narrow entry point does not necessarily imply a narrow company. It can be the wedge.

Founders may also need to think differently about the organization they build. If deployment is central to the moat, then product and AI talent alone will not be enough.

Someone needs to understand the customer’s environment deeply. Someone needs to own deployment, reliability and servicing. Someone needs to make sure the economics work for the customer, not merely that the robot works technically.

Those functions may appear less central than the model itself.

They may prove just as important to the outcome.

The most valuable robots may eventually look boring

We naturally pay attention to robots that run, jump, dance or demonstrate capabilities we have not seen before. Those demonstrations show how quickly the frontier is moving.

But many of the most valuable physical AI systems may eventually look less remarkable.

They will perform repetitive work inside factories, on construction sites, in warehouses and on farms. The customer will care less about how sophisticated the underlying AI is and more about whether the machine works when needed, improves productivity, reduces cost and can be serviced when something breaks.

That is why the current wave of physical AI may require a subtle shift in how we evaluate these companies.

The technological breakthrough gets a company into the race.

Deployment may determine who wins it.

As intelligence becomes more accessible, the harder question may no longer be who can build the smartest machine.

It may be who can turn intelligence into useful work, again and again, at scale.

These are very different companies at very different stages.

But together, they tell us something important about defense technology:

Capital is no longer the primary constraint.

That is an encouraging development. It is also the point at which founders and investors need to become more disciplined.

Defense has crossed into the venture mainstream

Defense technology spent years outside mainstream venture.

Government sales cycles were considered too long. Hardware was considered too capital-intensive. Procurement was opaque. Many investors saw the category as fundamentally inconsistent with venture returns.

That has changed decisively.

PitchBook recorded $19.8 billion invested across 262 defense-tech deals in the first quarter of 2026. Helsing’s latest round included venture firms, growth investors, financial institutions and a major Canadian pension investor.

Capital is now entering defense from nearly every part of the private-market stack.

The obvious debate is whether this has created a defense-tech bubble.

I think that is the wrong level of analysis.

A good market can still contain bad investments

A sector can be structurally attractive and still contain badly underwritten companies.

Equally, a company can carry a high valuation and still be rationally financed if the capital materially increases its probability of becoming a category-defining platform.

Government demand may be rising. Procurement priorities may be changing. Software, autonomy and lower-cost manufacturing may be creating entirely new product categories.

None of that tells us whether a particular company is appropriately priced—or whether it is actually retiring the risks that stand between technical promise and commercial scale.

The right question is not whether defense technology is receiving too much capital.

It is what uncertainty each additional dollar of capital is retiring.

What is the next dollar actually buying?

In traditional software, a large financing might accelerate product development, customer acquisition and international expansion.

In defense, capital may need to do considerably more.

It can fund:

• Technical development and testing

• Certification and security requirements

• Manufacturing facilities and inventory

• Specialized engineering talent

• Supply-chain resilience

• Deployment and field support

• The working capital required to serve government customers

These are legitimate reasons to raise large amounts of money. In some cases, capital itself becomes a competitive advantage because smaller companies simply cannot fund the required development and production cycle.

But fundraising can also blur the distinction between technical promise and commercial proof.

A company can use abundant capital to expand into more programs, build capacity before demand is established or postpone difficult decisions about which customer and product have the clearest path to scale.

Capital can retire risk. It can also finance unresolved risk for longer.

Knowing which is happening is the real underwriting challenge.

The milestones are not interchangeable

Defense companies progress through several distinct forms of validation:

1. Technical capability has been demonstrated.

2. The capability has been validated under operationally relevant conditions.

3. A customer has agreed to conduct a paid pilot or initial procurement.

4. The customer has made a repeat purchase.

5. The product has converted into a larger, enduring procurement program.

6. The company can deliver at the required cost, cadence and reliability.

These milestones may be directionally related. But they are not interchangeable.

A successful test is not a production contract.

A production contract is not a program of record.

A framework ceiling is not funded revenue.

Announced factory capacity is not demonstrated throughput.

Political support is not durable purchasing authority.

That distinction matters because the headline surrounding a defense company often collapses several of these milestones into a single impression: validation.

The underlying evidence may be much narrower.

Procurement remains procurement

Germany offers a useful example.

Its government approved an initial €540 million strike-drone order involving Helsing and Stark Defence. But lawmakers also reduced the proposed long-term procurement framework from €4.3 billion to €2 billion because they wanted to preserve parliamentary control over future commitments.

That is not evidence that demand is weak.

It is evidence that even in a period of strategic urgency, procurement remains procurement. Budgets, oversight, testing, performance, politics and competing priorities continue to matter.

A government can believe deeply in a mission while remaining cautious about a vendor, a program structure or the timing of a budget commitment.

Market urgency does not eliminate customer discipline.

What founders need to demonstrate

The weakest version of a defense startup’s narrative is:

Governments are spending more. Our technology is strategically important. Therefore, our market is large.

The stronger version is much more specific:

• Which customer problem has been validated?

• Under what operational conditions?

• Who controls the relevant budget?

• What must happen between the pilot and repeat procurement?

• What production rate will the customer ultimately require?

• What does the system cost per mission or engagement?

• Which part of the product improves through deployment data?

• What evidence would cause the customer to expand—or cancel—the program?

These questions are not obstacles to the founder’s vision.

They are the route by which the vision becomes a durable company.

Founders should be equally precise about the purpose of each financing round. “Scaling the company” is not enough.

The clearer question is: Which material risk should be measurably lower by the time this capital has been deployed?

What investors need to underwrite

The same discipline applies to investors.

Rising government budgets establish market attractiveness. They do not establish company quality.

A serious defense underwriting process should distinguish among at least five risks:

• Technical performance

• Operational validation

• Procurement conversion

• Production execution

• Unit economics

Investors should then ask which of these risks have already been retired, which the new capital is expected to retire and which remain largely outside the company’s control.

This should also shape follow-on decisions.

A higher valuation is not evidence that the underlying risks have declined. Nor is a large new investor necessarily better informed about every part of the business.

Reserves should follow the retirement of risk, not simply the momentum of valuation.

Abundant capital changes company behavior

When capital is scarce, companies are forced to confront customer evidence early.

When capital is plentiful, they have more choices. They can pursue several products, enter multiple geographies, build manufacturing capacity ahead of demand and hire for programs that may still be uncertain.

Sometimes that ambition creates the platform.

Sometimes it creates an impressive collection of unresolved risks.

The distinction is especially important for early-stage investors. They cannot—and should not—compete with billion-dollar growth rounds on capital alone.

Their advantage must come from recognizing which teams are converting technical learning into customer trust faster than the market appreciates.

That may be a founder who understands the budget pathway as deeply as the technology.

It may be a company designing around cost per engagement rather than maximum technical performance.

It may be a software and autonomy layer that improves across several hardware platforms.

Or it may be a manufacturing process that creates a compounding data advantage with every deployed system.

These are more durable signals than the size of the next financing.

Helsing shows why the answer is not simple

Helsing’s development illustrates the nuance.

The company began with defense AI software and has expanded into strike drones, underwater systems, aircraft applications and manufacturing. It has now announced a West Virginia facility designed to produce more than 2,000 HX-2 drones per month.

That could create a powerful combination of software, hardware, operational data and industrial capacity.

It could also create a much larger execution surface.

The $1.8 billion financing gives Helsing the resources to pursue that ambition. The financing announcement alone cannot tell us whether each part of the strategy will produce durable value.

And that is precisely the point.

Capital is useful evidence of investor demand. It is not customer validation, manufacturing proof or confirmation of value accrual.

Strategic importance should raise the standard

Defense technology is becoming one of venture’s most important categories.

The founders building within it are addressing real and urgent problems. They should have access to serious capital.

But the strategic importance of the category should raise the standard of underwriting—not lower it.

In defense tech, capital can buy time, talent and industrial capacity. It cannot buy procurement truth.

Quantum System’s $1.2 billion Series D brought together Blackstone and Airbus.

Same round. Same wire date. Same valuation. But not the same signal.

That distinction matters. Treat Blackstone and Airbus as the same kind of investor, and you will misread not only this round, but the next dozen growth rounds in defense, AI-enabled drug discovery, robotics, and physical AI.

This is not an isolated pattern

The same financing pattern is showing up across capital-intensive technology markets:

• CHAOS Industries raised $510 million in a round led by Valor Equity Partners - financial sponsor capital.

• Earendil Labs raised $787 million in a round co-led by Sanofi and the Pfizer-Hillhouse Biotech Development Fund - strategic capital.

• Apptronik brought in fresh capital from John Deere alongside its existing venture syndicate - corporate balance sheet capital sitting next to venture capital.

Same financing market. Very different money.

The common thread is that these companies need more than most traditional venture capital can usually provide on its own. Defense, drug discovery, and robotics all require capital, but they also require patience, specialized access, and operating context.

Defense companies need investors who understand certification timelines, procurement cycles, export controls, and the reality that government adoption rarely moves at software speed.

AI-enabled drug discovery companies need capital that can hold through clinical and regulatory timelines measured in years, not quarters.

Robotics and physical AI companies need funding that can support manufacturing, hardware iteration, supply chains, deployment, access to relevant data sets and compute capacity - not just go-to-market spend.

That is why private equity and strategics are stepping in. But the fact that they are entering the same rounds does not mean they should be underwritten the same way.

The easy read is the wrong read

The easy interpretation is that this is unambiguously good news for early-stage investors. More capital at Series C and D should mean fewer portfolio companies getting stranded, stronger markups, and more buyers for the next round.

That is partly true. But it is incomplete.

The more important question is not how much capital showed up. It is who showed up, why they showed up, and whether they are likely to show up again.

A large round led by a financial sponsor does not send the same signal as a large round led only by a corporate strategic. Both may validate the company. Only one may suggest durable, return-mandated follow-on capital.

Blackstone and Airbus are not the same kind of money

Blackstone runs on a financial return mandate. It has LPs, fund lives, underwriting discipline, portfolio construction logic, and a clear obligation to keep deploying capital if the return case continues to hold.

Airbus runs on Airbus’s own corporate priorities. Its investment appetite may be shaped by defense budgets, procurement roadmaps, internal R&D priorities, board-level capital allocation, and strategic relevance to the core business.

Those priorities can change for reasons that have very little to do with how Quantum Systems is performing.

That does not make strategic capital bad. In many sectors, it may be essential. A strategic investor can bring procurement access, technical credibility, manufacturing capability, clinical relationships, or commercial distribution that pure financial capital cannot easily replicate.

But strategic capital is not the same signal as financial sponsor capital. Its value may be high, but its durability may be more conditional.

History supports the distinction

Growth equity itself began as private equity moved down-market into high-growth minority deals. Over time, it did not disappear; it hardened into a distinct asset class. The core logic was durable: apply institutional underwriting and return discipline to companies that were still growing quickly but needed larger checks than traditional early-stage venture could provide.

Corporate venture capital carries different scar tissue. When the dot-com bubble burst, corporate venture arms were among the first sources of capital to retreat. Total VC deployment collapsed from roughly $86 billion in 2000 to $6.9 billion by 2002, and corporates led much of that pullback because their checks were never independent of the parent company’s own fortunes.

That history does not mean corporate strategic capital should be avoided. It means it should be discounted differently. A corporate-led round may be strong validation, but it should not automatically be treated as the same durability signal as a PE-led round.

What investors should ask

For early-stage investors, the diligence question has to change.

It is no longer enough to ask whether a company raised a large Series C or D. The better questions are:

• Who is actually leading the round?

• Is the lead investor return-driven, strategic, or a hybrid of both?

• Would this investor follow on if the financing market tightened?

• Is the corporate parent’s own budget, R&D spend, or capex cycle supportive of continued involvement?

• Does this round expand the company’s optionality, or does it make the company look captured by one strategic agenda?

Pro-rata and information rights are the baseline, not the whole playbook. Investors should build more active financing discipline around rounds where strategics are involved.

• Secure MFN and information parity where possible. A small non-lead check still needs visibility into the real dynamics of the round.

• Help shape the syndicate early. If your board seat, network, or sector credibility allows you to introduce a PE fund before a strategic becomes the default lead, you are shaping the company’s future financing signal rather than reacting to it.

• Treat a corporate-only-led round as validation, not automatic continuity. It may be a good round, but it may also be a secondary window if the markup is driven by capital that may not be there in the next cycle.

• Watch the parent company, not just the portfolio company. Airbus’s defense exposure, Sanofi’s R&D appetite, John Deere’s capex priorities - those may be leading indicators of whether the strategic shows up again.

What founders should understand

For founders, the lead investor can shape future optionality more than the valuation does.

A PE-led or PE-plus-strategic structure may preserve flexibility. It tells the market that the company has financial durability, while still allowing it to benefit from strategic access.

A sole strategic lead can bring real commercial value, but it can also create real exposure. The risk is not simply dilution. The risk is that the company becomes perceived as tied to one corporate agenda.

That perception can matter in future financings, commercial partnerships, and M&A discussions.

Founders should take the strategic value where it is real - procurement access, clinical relationships, manufacturing capacity, distribution, technical validation - but they should be careful about the governance and commercial rights that often travel with that capital.

• Exclusivity clauses can quietly narrow the market.

• ROFRs on future M&A can chill other buyers.

• Governance rights can give a strategic influence that exceeds the economics of its check.

• A single strategic relationship can become a dependency if the company builds its go-to-market motion around one corporate partner.

The best founders will use strategic capital without becoming strategically captured. They will take the commercial advantage while preserving competitive tension, financing flexibility, and future buyer optionality.

The real underwriting question

The capital entering these sectors is real, and it is growing. That is the good news.

The harder truth is that not all capital is the same currency.

Some capital validates. Some capital compounds. Some capital disappears when the parent company’s priorities change.

The investors and founders who understand the difference will make better decisions than those who only look at the size of the round.

Stop pricing the round. Start pricing the lead.

That is not a philosophical position—it is empirically observable.

Despite a surge in AI deployment across portfolio companies and GPs, few firms can point to measurable, realized value creation from AI at scale. Advisory work from leading firms shows that while adoption is widespread, value capture remains concentrated in a small set of outliers .

The gap is not capability. It is operating model design.

This is the central mispricing in private markets today.

1. From Data Infrastructure to Execution Systems

Most firms still operate on what can only be described as data exhaust systems - platforms that aggregate, store, and visualize.

But the next phase is already underway: execution systems.

These systems do not just inform decisions - they initiate them:

• LP reporting that dynamically generates narratives from live fund and portfolio data

• Portfolio monitoring that flags deviations before quarterly reviews

• Capital allocation workflows triggered by predefined data thresholds

The critical shift is collapsing the latency between signal → decision → action.

Where this becomes economically relevant is in decision velocity.

Empirical evidence from multiple operational benchmarking studies (e.g., Bain, BCG operating models work) consistently shows that faster decision cycles correlate with superior capital deployment outcomes, particularly in volatile environments.

Yet most firms have not redesigned workflows to take advantage of this.

They are automating fragments, not systems.

The result: marginal efficiency gains without structural advantage.

Question:
If your decision cycle shortened by 30–50%, how would it change your portfolio outcomes - and do your current systems even allow for that possibility?

2. The Data Layer Is Now a Competitive Weapon

Across private markets, firms are splitting into two camps:

1. Those building integrated, decision-grade data layers

2. Those maintaining fragmented, reporting-oriented systems

Only one of these compounds.

The consequence is already visible in underwriting.

At the growth stage, companies are increasingly built with:

• Real-time data pipelines

• Embedded analytics across product and finance

• Dedicated data engineering resources early in lifecycle

By contrast, buyout firms continue to face:

• Post-acquisition integration complexity

• Legacy system fragmentation

• Data structured for reporting, not forward-looking insight

This creates a structural inversion:

The asset is often more data-mature than the owner.

That has direct implications:

• Slower post-acquisition value creation

• Delayed synergy realization

• Reduced ability to operationalize insights

At the same time, startups lack governance rigor—creating risk.

The winning model is not choosing between the two. It is combining them:
data-native infrastructure + institutional governance discipline

Firms that fail to converge these will face a widening capability gap.

Question:
When you diligence an investment today, do you explicitly underwrite its data architecture - and how does it integrates into your own?

3. Why Most AI Programs Fail (And Will Continue To)

The headline failure rate of AI initiatives—often cited in the 80–95% range for experiments that do not scale—is directionally correct, even if imprecise.

The underlying causes are not ambiguous:

• No workflow redesign → AI layered onto legacy processes

• Fragmented data environments → outputs lack credibility

• No clear ownership → initiatives stall in pilot mode

• Misaligned incentives → passive resistance within teams

• Tool-first thinking → no prioritization of high-value use cases

Critically, this aligns with observed patterns in portfolio companies:

Most firms remain in the “deploy” phase (tool distribution) rather than progressing to:

• Reshape (organizational redesign)

• Invent (AI-native business models)

This explains why adoption is high but impact is low.

AI does not fail because models underperform.

It fails because firms refuse to change how work gets done.

Question:
Where in your organization have you redesigned the workflow - not just added AI - and what P&L impact has it actually delivered?

4. The Myth of Efficiency as Cost Reduction

There is a persistent but flawed framing: efficiency equals cost-cutting.

That is wrong.

Efficiency, in high-performing firms, is a function of operating clarity.

The firms that scale consistently do three things:

1. Define core vs non-core explicitly
Outsourcing decisions are strategic, not reactive.

2. Align incentives across functions
Investment, finance, and IR operate against shared outcomes—not siloed KPIs.

3. Establish execution cadence
Weekly, monthly, and quarterly rhythms are institutionalized.

These are not soft factors—they determine operating leverage.

A critical shift here is the elevation of Investor Relations.

In a constrained fundraising environment:

• LP responsiveness

• Data transparency

• Consistency of communication

…are becoming quantifiable differentiators.

Firms that treat IR as reporting will underperform in capital formation.

Question:
If an LP evaluated your firm purely on responsiveness, transparency, and insight quality, would you rank in the top quartile?

5. Measuring Operational Excellence Correctly

Most firms still measure what is easy—not what matters.

Traditional metrics:

• Cost ratios

• Headcount efficiency

• Budget adherence

These are lagging indicators.

The leading indicators are:

• Decision Velocity: time from data availability to action

• Insight Quality: predictive vs retrospective analysis

• Execution Consistency: variance between plan and outcome

• LP Experience: responsiveness and clarity of communication

This reflects a broader shift across industries:
measurement is moving from outputs → decision systems

Firms that fail to update their metrics will misallocate capital—because they are optimizing for the wrong outcomes.

Question:
What percentage of your internal metrics actually measure decision quality rather than operational activity?

6. Liquidity Is Now an Active Strategy

The slower exit environment has forced a structural shift:

Liquidity is no longer episodic. It is managed.

CFOs are increasingly using:

• Secondary markets for partial realizations

• Structured liquidity solutions

• More disciplined capital recycling

At the same time, technology is improving visibility into:

• Fund-level liquidity

• Portfolio cash flows

• Forward capital needs

But the key change is behavioral.

Firms are shifting from:
reactive liquidity management → proactive capital strategy

This requires tighter integration across:

• Investment teams

• Finance functions

• Portfolio operators

Without that integration, visibility does not translate into action.

Question:
How often do you proactively generate liquidity versus waiting for exit opportunities- and what is the impact on DPI?

7. Talent Is Not a Hiring Problem—It’s a Design Problem

The industry continues to frame talent gaps incorrectly.

The issue is not a shortage of talent.

It is a mismatch between roles and operating models.

Emerging roles include:

• Data architects embedded in deal teams

• AI workflow designers within operations

• Hybrid finance-technology operators

At the same time, AI is disproportionately eliminating:
coordination-heavy roles, not judgment-intensive ones

This has two implications:

1. Firms must reskill—not just hire

2. Organizational structures must evolve

Firms that fail to redesign roles will see declining productivity - even if they adopt AI tools.

Question:
Which roles in your organization are fundamentally coordination-driven - and what is your plan to eliminate or redesign them?

8. The Only Operational Bet That Matters

There are many initiatives competing for attention.

Most are incremental.

One is structural:

A unified, AI-native data layer across investment, portfolio, and LP functions

If correct, it enables:

• Faster decisions

• Better capital allocation

• Differentiated LP experience

If incorrect, the cost is not just wasted capital - it is lost time in a compounding game.

Because operating models compound.

And lag compounds faster.

Question:
If you had to rebuild your firm today from scratch, would you design your data architecture the same way?

So….what are we really saying:

This is not about efficiency.

It is about who compounds faster.

• Faster decisions → better capital deployment

• Better data → stronger underwriting

• Better LP experience → superior fundraising

Private markets have historically competed on access, relationships, and judgment.

The next decade introduces a fourth variable:

Operating system design

Most firms are not prepared for that shift.

THE END!!