Welcome to the 102nd edition of Deep Tech Catalyst, the educational channel from The Scenarionist where science meets venture!
AI in healthcare is moving fast—from diagnostics and imaging to drug discovery, virtual care, and advanced therapies.
Alongside the technology, though, there is a quieter but decisive layer: the way capital is structured, which influences how long companies can stay in experimentation mode, how they absorb risk, and how they pace their path to scale.
Within that layer, different models coexist. Traditional closed-end funds work with defined fundraising, deployment, and exit cycles. Family offices and disease-focused foundations often use more flexible or evergreen structures. Each of these approaches creates its own constraints and opportunities for founders and investors working in healthcare.
To explore the patient capital model in more detail, we’re joined by Everett Kamin, Healthcare Investment Director at Gemini Capital Partners, Chief Investment Officer and Board Advisor to the Silverstein Family Office and its Dream Foundation, and Impact Investment and Board Advisor to Macmillan Cancer Support.
Key takeaways from the episode (TL;DR):
🧭 Family Offices Play a Different Game Than Funds
They are not constrained by a 7–10 year fund cycle and focus more on capital returned over time than on IRR, which changes how they underwrite risk in healthcare.
🌱 Evergreen Structures Match Healthcare Timelines
Without a forced exit clock, patient capital can stay with services, devices, and enabling technologies for as long as it takes for regulation, adoption, and reimbursement to catch up.
🤖 In Healthcare AI, Data Comes Before the Model
Investability hinges on data integrity, security, ownership, exclusivity, and access duration—only then does model defensibility and the “flywheel” effect really matter.
👥 Founders Need Both Depth and Versatility
Experienced, “Swiss army knife” teams that have built and launched products before, and that target real unmet needs rather than crowded categories, stand out to this type of investor.
🎗️ Foundations Bring Capital, Credibility, and Access
Mission-driven investors like the Silverstein Dream Foundation and Macmillan Cancer Support act like VCs on the cap table, while also opening doors to hospitals, ecosystems, and non-dilutive grants aligned with specific disease areas.
Deep Tech Monthly in Review - November 2025
Field notes from last month in Deep Tech startups and private markets — a strategic recap for Builders and Backers.
BEYOND THE CONVERSATION — STRATEGIC INSIGHTS FROM THE EPISODE
How Family Offices Differ from Traditional Funds
Closed-end funds are built on a specific logic:
They raise capital from external investors, usually over a defined fundraising period of one to three years.
They then have a defined deployment period, again often one to three years, during which they are expected to put that capital to work.
Finally, there is a harvesting or realization window of similar length, during which they are expected to exit investments and return capital with a profit.
The life cycle of a VC fund in this model is often seven to ten years; for private equity, it is usually five to seven.
There can be extensions, but the structure creates a clear cadence and a real sense of timing. Limited partners expect their money back, and managers plan exits with that horizon in mind.
Family offices mostly deploy their own capital and operate differently from other investment firms.
They generally fall into three main types:
A single family office revolves around one principal and that principal’s family, managing and deploying a pool of wealth into chosen strategies.
A multifamily office brings together multiple families, often to pool resources and share operational efficiencies while co-investing in strategies or individual deals.
Then there are more institutionalized multifamily offices, where several families are invested, but professional money managers run the platform day to day, with a chief investment officer holding real investment authority.
There is no external timetable forcing them to make an investment by a specific deadline or to sell an asset because a fund is approaching the end of its life. In practice, many family offices also invest in funds as limited partners and sometimes as general partners.
They may have a liquid book, investing in public markets; an illiquid book, allocating to private equity and venture capital funds; and a direct book, comprising operating companies or minority stakes in businesses.
Where a family office often diverges from the classic fund mindset is in how it evaluates success over time. In a fund, the primary performance yardstick is usually IRR, an annualized internal rate of return that builds in time-value-of-money assumptions and tends to reward faster exits and quicker return of capital.
In a family office context, the emphasis tends to shift more toward multiple on invested capital—how much capital has been returned versus how much was originally deployed—especially when an asset can be held for twenty years, collecting dividends and cash flows along the way.
IRR is still relevant, but in an evergreen or continuation structure, the compound effect of cash flows and capital returned can become the more intuitive way to think about value.
Why Evergreen Structures Can Fit Healthcare Timelines
An evergreen structure does not impose a fixed period within which capital must be deployed or harvested. There is no obligation to exit an investment purely because a vehicle is nearing the end of its life. Instead, the hold period can be adapted to the specifics of the company and the market.
Capital can stay with a business for a handful of years or for much longer, as long as the investment thesis remains intact.
For companies on the other side of the table, that creates a different type of relationship.
When they take capital from an evergreen, family-office-backed strategy, they are not tied to a predefined fund clock; equally, they are working with a partner whose decision to stay or exit is case by case rather than driven by a pooled LP base.
With a traditional fund, the timelines and expectations are more defined; with evergreen capital, the flexibility is greater, but the pacing is set more by shared judgment than by fund life.
In healthcare, where the development of services, devices, diagnostics, and enabling technologies rarely fits neatly into a seven-year cycle, this flexibility can be relevant.
In practice, this leads to an investor landscape where different models coexist. Some actors combine control and minority stakes, mature and early-stage companies, services, and enabling technologies, under an evergreen structure. Others operate through classic closed-end funds with defined horizons.
For founders and investors, the key is not to assume one model is inherently superior, but to understand how each structure interacts with the realities of healthcare innovation—and to choose partners whose time horizons, incentives, and constraints match the journey they are trying to undertake.
AI’s Role in Healthcare: An Investor’s Lens
One of the clearest areas where AI is already making a difference is diagnostics and imaging. The raw numbers may look modest—a few percentage points of accuracy gained, systems performing at roughly the level of human experts—but the context matters.
Clinicians work long shifts under constant pressure, and fatigue inevitably raises the risk of missing something. AI introduces a diagnostic layer that never gets tired, running in the background to double-check scans and test results with the same consistency at the end of a shift as at the beginning.
Even small gains in accuracy, applied continuously, translate into fewer misses and a more reliable baseline of care. Inside the lab, AI takes over repetitive tasks such as counting blood cells, handling them in seconds, and freeing technicians to focus on higher-value work.
Coupled with the ability to process very large datasets, this makes it possible to detect diseases earlier and more efficiently, using patterns that would be too time-consuming to track manually.
Accelerating Drug Discovery and Clinical Trials
The impact extends into drug discovery and clinical development.
AI can significantly shorten the discovery process, not through a single breakthrough algorithm, but by supporting many steps in the workflow: exploring chemical and biological space faster, prioritizing candidates, and revealing patterns in preclinical and clinical data that would be hard to see otherwise.
Around these core scientific uses, similar tools streamline the operational side of clinical trials—documentation, payments, data flows, coordination—reducing friction so that projects can move more smoothly from concept to clinic and from clinic to real-world use.
Automating Reimbursement and Administrative Work
AI is also starting to reshape the less visible but extremely consequential world of reimbursement and administration.
In many systems, staff spend substantial time on the phone with payers to confirm coverage and map procedures or devices to specific reimbursement codes.
Software can now query policies, interpret rules, and perform much of this mapping automatically, compressing what used to take multiple calls into minutes.
That shortens the time between a clinical decision and action, helps patients and providers get clarity sooner, and reduces the delay between delivering a service and getting paid.
From Wearables to Advanced Therapies and Production
On the patient-facing side, wearables and virtual care platforms allow continuous monitoring instead of occasional snapshots.
Signals that were once captured only during sporadic clinic visits can now be tracked throughout the day, with AI scanning the data stream for early signs of trouble.
At the other end of the spectrum, AI is influencing how advanced therapies are produced. CAR T-cell therapy is a concrete example: originally extremely expensive and slow to manufacture, it can now be produced in a fraction of the time and at a much lower cost when AI-enabled processes are integrated into the workflow.
That shift alters both the economics and the practical accessibility of such treatments.
4 Elements That Make an AI Healthcare Startup Investable
Building on Solid Data: Integrity, Security, Ownership, and Access
When an AI company operates in healthcare, the first questions an early-stage investor asks are not just about clever algorithms. They are about the data that sits underneath everything.
The integrity of that data and the way it is handled form the foundation on which the entire investment case rests.
Here are five important considerations to stress-test your startup project.
The starting point is understanding exactly what data is feeding the model.
How robust is it? How representative is it of the populations and use cases the company claims to serve? Is it based on a narrow cohort, or is it built on a sufficiently broad and reliable base to support the clinical claims being made?
The uniqueness of the data.
If anyone can access the same datasets in the same way, the advantage is fragile. If a company has built, curated, or secured access to something that others cannot easily replicate, that strengthens its position.
Security.
Healthcare data is sensitive, and the expectations around privacy and protection are only increasing. It is not enough to say that data is secure; a company has to show that the right protocols are in place and that they are being followed. In the current environment, that is not a side note—it is one of the big questions.
Ownership.
An investor needs to know who actually owns the data being used. Is it the company, a hospital, a third-party provider, or some combination of the three? Are there clear rights in place for the company to use that data in the way it intends? If the data is shared or licensed, what does that license look like? How long does it last?
This is where details such as data lockups and exclusivity periods become important. A company might have exclusivity over a dataset for a year, only to lose it in the second year. That changes the durability of its advantage.
The consistency of data updates.
Is the company working with static data that never changes, or with live data streams that are refreshed regularly? Will it have access to those streams over time, or is that access precarious?
Taken together, data integrity, security, ownership, lockups, exclusivity, and access duration form a picture of how solid the foundation really is. Only once that picture is clear does it make sense to move on to the model itself.
Designing Models with Real Longevity and Defensible Moats
In AI, there is an uncomfortable but necessary assumption: a better model will always be built. Even if a company has the best model today, it is almost inevitable that someone, somewhere, will come up with something stronger tomorrow.
That reality shapes how investors think about model longevity and defensibility.
The question is not whether the model is perfect. It is whether it can be defended long enough to become the dominant, default choice in its niche—whether that niche is a specific indication, a particular workflow, or a clearly defined discovery process.
The goal is to become the de facto service for that use case, so that when people think about solving that problem, they automatically turn to this company’s solution.
One way to build that defensibility is through a data and usage flywheel.
If the model performs well, more users will be attracted to it. As more users come, they generate more data. As more data flows into the system, the model can be retrained, refined, and improved. That, in turn, makes it even more attractive to users.
Over time, this creates a situation where it is not just the model architecture that matters, but the accumulated data and experience built into it.
The moat then comes from scale and embeddedness.
A competitor may build a technically interesting model, but catching up to a system that has already become the go-to in its space is far from trivial, especially once that system is fed by ongoing real-world usage.
This is why many healthcare AI companies are choosing a specific spot and trying to dominate it, rather than spreading themselves too thin. The priority is to own a clearly defined segment and build a defensible position there, rather than being vaguely present in many areas without real depth anywhere.
The Core Team: Experience, Versatility, and Market Insight
Beyond data and models, the composition of the founding team is crucial. Investors are often candid about this: they do not want to pay for other people’s learnings.
Every new founder will make mistakes, but a team that has built and scaled companies before will skip a long list of avoidable errors. That experience compresses time and reduces risk.
In practical terms, that means looking for founders who have already formed companies, created products, navigated regulatory and legal structures, and dealt with issues such as intellectual property.
They know how to file patents where they make sense, and how to think about defensibility even in spaces where patents are not the main tool.
They understand how to move from a concept to a product, and from a product to a sale.
They know what it means to get something into a market and then expand into adjacent markets.
In the earliest stages—seed and pre-seed—teams are small, and resources are limited.
There is no fully built-out C-suite, no large roster of senior executives to lean on. A founder may find themselves acting as de facto CEO, COO, and CFO in the same week, while also speaking to customers and working with engineers.
That is why investors look for “Swiss army knife” profiles: people who can wear multiple hats, who have a broad set of skills, and who can draw on personal networks and resources without always needing to spend cash.
What matters is not only what the founders know, but how they behave. A team that has the energy and drive to run through walls, combined with enough experience to avoid the most common traps, is far more likely to execute.
When that execution is focused on a product that answers a real gap in the market, and when the team can articulate why this gap has been overlooked or poorly served until now, the investment case strengthens.
Solving A Clear Unmet Need
Finally, there is the question of what the company is actually trying to achieve. Investors are not looking for yet another version of what everyone else is already doing. They are drawn to propositions that address unmet needs in a way that can be defended and scaled.
The ideal is a company that is not simply competing in an existing crowded category, but defining or creating its own category—a space where it can grow without immediately facing a wall of indistinguishable rivals.
That does not mean chasing novelty for its own sake. It means identifying a problem that is real, important, and currently underserved, then designing a solution that fits that problem tightly.
In an AI healthcare context, that might be a specific diagnostic gap, a particular bottleneck in clinical workflows, a neglected patient population, or an overlooked part of the drug development process.
Whatever it is, the key is that the solution is not just another generic tool, but something that clearly answers a need that others have not addressed properly.
When a company brings together solid data foundations, a model with a credible path to defensibility, a team with real experience and versatility, and a proposition that tackles a genuine unmet need, it stands out.
It moves from being one more AI healthcare startup among many to being a candidate for a long-term partnership—one that is worth supporting not just because it is fashionable, but because it has the ingredients to build something durable.
Partnering with Mission-Driven Foundations in Healthcare
Foundations as Strategic Partners
If you are building a company around a clearly defined disease area, a foundation aligned with that indication can become a specific kind of partner.
Their starting point is often the impact a technology might have on the patient group they care about.
That lens influences which projects they look at, how they evaluate relevance, and how they think about the potential long-term role of a company in a given therapeutic area. The capital they deploy is still managed with discipline, but the analysis includes clinical and societal dimensions alongside financial ones.
Networks, Signaling, and Access to Ecosystems
Another dimension highlighted in the discussion is the network that often surrounds these organizations. Foundations like Macmillan tend to be connected to hospitals, patient support services, nonprofits, and other stakeholders in oncology. The Silverstein Dream Foundation plays a similar role within the diabetes ecosystem.
For a founder, this can translate into easier access to certain conversations.
Being backed by a foundation that is already embedded in a clinical and philanthropic network can help with introductions to clinicians, hospital systems, and partner organizations that are familiar with that foundation’s work.
It does not automatically guarantee adoption or success, but it can shorten the time it takes to reach some of the people you need to speak to.
There is also a signaling effect to consider. When a foundation with a clear mission and established reputation decides to invest in a startup, other investors can read that as a sign that the company is relevant to a specific disease area and aligned with the foundation’s understanding of the problem.
These organizations are usually selective.
Their presence on the cap table suggests that a company has passed through an additional filter tied to medical and mission fit, not only to financial potential.
Because they are active in the philanthropic world, these foundations also tend to know the landscape of grants and non-dilutive funding connected to their indication.
Once they are involved, they can sometimes help a team navigate toward programs and funders whose missions are aligned. Outcomes are never guaranteed, but they can make it easier to map opportunities that would otherwise be harder to identify.
💸 Seeds pile into construction AI, defense autonomy & battery optimization; mature capital anchors space surveillance & clinical networks | Deep Tech Capital Movements 47
The Week’s State of Deep Tech Capital: who’s Raising, who’s Betting, and why.
