Investor And Venture Outlook On AI | Takeaways For Founders And Product Leaders
A grounded lens on where AI value will compound, which risks matter, and why execution discipline beats hype.
TLDR: This blog shares what investors really think about AI in 2025. The big idea: AI is still in its early days, even if it doesn’t feel that way. Just because everyone in tech is talking about AI doesn’t mean businesses are actually using it yet. Real adoption shows up in budgets, not just experiments. Many industries have barely started. The core message for founders and investors: the AI opportunity is just getting started, not winding down.
Introduction
Founder Intro: Investor and Venture Outlook on AI in 2025
There’s no shortage of opinions about AI’s future. What’s far rarer is clarity about what actually matters right now. For founders, it is about building companies; for investors, about deciding where conviction belongs.
Panel 5 was designed to cut through that noise. Rather than speculate about distant futures or abstract breakthroughs, we wanted to anchor the conversation in the realities shaping AI businesses in 2025: adoption curves, economics, org design, governance, and where durable value is actually accruing.
To do that, we brought together investors who are actively underwriting these questions across different stages, geographies, and market structures:
Lukas Linemayr, Partner at Streamlined Ventures.
Rak Gard, Partner at Bain Capital Ventures.
Zao Chen, Investor at Craft Ventures.
What emerged was a surprisingly grounded picture of the AI landscape. Yes, the market is early, but it is not empty. Yes, capital investment is massive, but revenue realization takes time. Yes, platform risk is real, but applications still capture value. And perhaps most importantly: AI has expanded the outcome space for founders rather than narrowing it.
This panel wasn’t about predicting AGI timelines or chasing the next hype cycle. It was about understanding constraints, making realistic bets, and recognizing where opportunity still hides — often in overlooked markets, unglamorous workflows, and human-heavy industries that software never fully reached.
Across the discussion, one theme stood out:
“AI changes what’s possible — not what’s required to build a real business.”
Durable companies are still built on trust, usage, distribution, and judgment. The tools are new. The fundamentals are not. The sections that follow break down how investors are thinking about value capture, revenue quality, founder profiles, governance, and scale — not as theory, but as underwriting criteria today.
If you’re building in AI and trying to decide what kind of company to build, whether venture is the right path, or where the next decade of opportunity actually lies, this panel offers a clear place to start.
1. The Market Is Early — But Not Empty
One of the most consistent refrains across the panel was a corrective to a common misconception:
AI adoption feels saturated inside tech circles — but it isn’t saturated in the real economy.
What looks crowded from within Silicon Valley looks very different when viewed across industries, geographies, and buyer maturity curves.
Inside the Bubble vs Outside the Market
Within technology ecosystems, AI can feel ubiquitous. Models are improving rapidly. New products launch weekly. Capital is flowing aggressively.
But as multiple panelists emphasized, this perspective is deeply skewed.
Outside of tech-forward companies:
Most enterprises are still experimenting.
Deployments are limited to pilots or narrow workflows.
Leadership teams are cautious.
Organizational readiness lags technical capability.
As Lukas Linemayr, Partner at Streamlined Ventures, noted, exposure should not be confused with adoption. Awareness is high. Actual usage at scale is not.
Budgets Tell the Real Story
Several panelists pointed to a simple reality check: budget allocation.
Despite the attention AI receives, AI spend remains a small fraction of overall enterprise budgets. In most organizations, it competes with:
Legacy software commitments.
Infrastructure modernization.
Security and compliance spend.
Headcount and services.
As Rak Gard, Partner at Bain Capital Ventures, emphasized, real adoption shows up in sustained budget line items — not experimentation funds. By that measure, most enterprises are still in early innings.
Consumer Adoption Is Uneven, Not Universal
The panel also pushed back on the idea that consumer AI adoption is “done.”
While some products have achieved massive usage, adoption remains:
Uneven across geographies.
Concentrated among power users.
Fragmented by use case.
Highly sensitive to trust and clarity.
As Tiger Gao, Investor at Apax Digital, pointed out, consumer behavior varies dramatically outside of early-adopter markets. What feels mainstream in one region can be niche in another.
This unevenness suggests opportunity — not saturation.
Entire Industries Are Barely Started
Perhaps the most important insight was how many sectors have barely begun meaningful AI deployment. Industries like healthcare, manufacturing, logistics, financial operations, and regulated services face constraints that slow down,
Adoption.
Compliance requirements.
Legacy systems.
Data fragmentation.
Cultural resistance.
As Zao Chen, Investor at Craft Ventures, noted, these constraints don’t eliminate opportunity; they delay it. And delayed markets often end up being the largest ones.
Capital ≠ Product-Market Fit
A key clarification from the panel was that capital investment should not be mistaken for market maturity.
Yes, enormous amounts of capital have flowed into AI. No, that does not mean product-market fit is solved.
At-scale PMF:
Is still forming.
Looks different by industry.
Requires integration, not just intelligence.
Unfolds over years, not quarters.
Many AI products are still searching for repeatable, durable deployment patterns.
Diffusion Has Just Begun
This led to the panel’s core takeaway:
Today’s traction does not represent peak penetration.
It represents the beginning of diffusion.
We are early in the curve where:
Workflows are being discovered.
Buyers are learning how to buy.
Organizations are learning how to deploy.
Trust is still being earned.
For founders and investors alike, this reframes the opportunity.
The market isn’t empty. But it’s far from full.
The Practical Takeaway
AI may feel late-stage if you only look at demos, headlines, and funding rounds.
But if you look at:
Real usage.
Real budgets.
Real deployment.
Real behavior.
The conclusion is clear: we’re still at the beginning of adoption, not the end.
For companies that can survive the experimentation phase and earn trust at scale, the next wave of growth is still ahead.
2. AGI Debates Matter Less Than Near-Term Constraints
AGI and superintelligence inevitably came up during the panel, but notably, they were treated as context, not catalysts.
The investors were aligned on a simple point:
AGI debates are intellectually interesting. And that near-term constraints determine outcomes.
AGI Is a Moving Target
One of the first issues raised was definitional.
As Lukas Linemayr, Partner at Streamlined Ventures, noted, there is no stable, shared definition of AGI. What qualifies as “general” varies by speaker, by benchmark, and by moment in time.
This makes AGI a poor anchor for:
Investment decisions.
Company strategy.
Product roadmaps.
If the goalposts keep moving, progress becomes impossible to evaluate meaningfully.
Reasoning Exists — But Only Inside Boxes
The panel acknowledged real advances in multi-step reasoning.
Models today can:
Chain logic.
Follow structured plans.
Solve complex problems within constrained domains.
But that constraint is doing the real work.
As Rak Gard, Partner at Bain Capital Ventures, emphasized, reasoning degrades rapidly once systems leave controlled environments. Outside of well-scoped tasks, models struggle with ambiguity, long-horizon execution, and accountability.
This gap matters far more than abstract intelligence scores.
Autonomy Is Bottlenecked by the World, Not Models
Another key insight was that autonomy isn’t limited by model capability alone.
It’s bottlenecked by:
Messy real-world environments,
Poor or fragmented data,
Limited feedback loops,
Immature reinforcement learning systems.
As Tiger Gao, Investor at Apax Digital, pointed out, intelligence without grounding doesn’t scale. The world is not a clean API. Until systems can reliably sense, act, and learn in open environments, autonomy will remain constrained regardless of model improvements.
Timelines Are Longer Than the Discourse Suggests
The panel was notably conservative on timelines.
Not pessimistic, rather realistic.
Breakthroughs will happen.
Capabilities will improve.
New classes of applications will emerge.
But as Zao Chen, Investor at Craft Ventures, noted, the gap between lab demos and reliable deployment is often measured in years, not months. Overestimating timelines is one of the fastest ways to make bad bets.
Investors Underwrite Constraints, Not Possibility
This led to a shared investment posture.
While AGI-level outcomes may shape long-term narratives, investors operating today underwrite constraints:
Where models fail.
Where workflows break.
Where adoption stalls.
Where economics don’t pencil.
Near-term success depends on navigating these limitations and not assuming they’ll disappear.
Founders who build as if constraints are permanent often outperform those betting on imminent breakthroughs.
The Practical Takeaway
AGI debates will continue — and they matter for long-term vision.
But in 2025:
Constraints drive outcomes.
Environments matter more than intelligence.
Deployment beats demos.
Realism beats speculation.
For builders and investors alike, the message was clear:
The next wave of value won’t come from waiting for AGI. It will come from building durable businesses inside today’s limits and also expanding those limits over time.
3. Massive CapEx Does Not Automatically Equal Massive Revenue
One of the most candid discussions on the panel centered around a growing tension in the AI ecosystem:
Infrastructure spending has exploded, but revenue realization is still catching up.
This disconnect is real, and it matters.
Infrastructure Spend Is Front-Loaded by Design
The panel acknowledged the obvious headline: AI has triggered one of the largest infrastructure buildouts in modern tech history.
Compute.
Data centers.
Specialized hardware.
Energy commitments.
As Rak Gard, Partner at Bain Capital Ventures, noted, this level of CapEx is unprecedented outside of telecom or cloud hyperscalers. But unlike traditional software, AI infrastructure must be built ahead of demand.
This makes early financials look distorted — not broken.
Revenue Exists — Just Not in Proportion Yet
A key nuance the panel emphasized was that AI revenue is real and growing quickly.
Some AI applications are:
Growing faster than any prior software category.
Achieving meaningful ARR at early stages.
Demonstrating strong willingness to pay.
As Lukas Linemayr, Partner at Streamlined Ventures, pointed out, aggregate AI ARR across the ecosystem is already substantial.
What it is not yet is proportional to the infrastructure being built to support future demand.
That gap is expected and temporary.
Monetization Lags Capability
Another consistent insight was that monetization always lags technical capability.
Models improve first.
Use cases emerge next.
Business models stabilize last.
As Tiger Gao, Investor at Apax Digital, explained, AI creates value before it captures value. It takes time for:
Buyers need to understand ROI.
Pricing models to normalize.
Procurement processes to adapt.
Budgets to shift meaningfully.
This lag is not unique to AI, but the scale makes it more visible.
CapEx Absorption Takes Time
The panel converged on a clear expectation:
CapEx absorption will take years, not quarters.
Infrastructure will be amortized over long time horizons.
Revenue will arrive unevenly.
Some segments will monetize faster than others.
As Zao Chen, Investor at Craft Ventures, emphasized, this doesn’t imply poor returns — it implies patience. Investors expecting immediate proportionality between spend and revenue are misreading the cycle.
Uneven Returns Are a Feature, Not a Bug
Another important point was that returns will not be distributed evenly.
Some layers will:
Capture outsized value early.
Show strong unit economics.
Justify spending quickly.
Others will:
Struggle to monetize.
Remain infrastructure-heavy.
Consolidate over time.
This unevenness is characteristic of platform shifts, not a sign of failure.
The Practical Takeaway
Massive CapEx is not proof of massive revenue, yet.
But it is a prerequisite for it.
The panel’s consensus was grounded but optimistic:
Revenue is coming.
Monetization is forming.
Timelines are longer than hype suggests.
For investors and founders alike, the message was clear:
Don’t confuse delayed returns with absent returns.
The AI buildout is early — and uneven by design.
4. Value Accrues to Applications, Not Foundations
One of the strongest points of alignment across the panel was a lesson the industry has learned repeatedly:
Platforms enable value.
Applications capture it.
AI does not break that pattern; it reinforces it.
History Rhymes — Even When Technology Changes
The panel situated AI within a familiar historical arc.
In prior platform shifts:
Operating systems enabled software companies.
Cloud infrastructure enabled SaaS.
Mobile platforms enabled app ecosystems.
In each case, the enabling layer was essential — but the enduring value accrued to the application layer.
As Rak Gard, Partner at Bain Capital Ventures, emphasized, AI follows the same economic logic. Infrastructure makes new behavior possible. Applications turn that possibility into revenue.
Foundations Are Necessary — and Brutal
The panel was clear-eyed about the difficulty of foundation-layer businesses.
Chips, models, and infrastructure are:
Capital-intensive.
Technically complex.
Strategically critical.
But they are also:
Highly competitive.
Subject to commoditization.
Constrained by margin pressure.
As Lukas Linemayr, Partner at Streamlined Ventures, noted, the model layer increasingly resembles cloud infrastructure wars — massive scale advantages, few winners, and brutal economics for everyone else.
These businesses matter — but they are structurally hard to own as long-term value capture plays.
Applications Control the Customer
What applications uniquely possess is the user relationship.
Applications own:
Workflow integration.
Daily usage.
Customer trust.
Switching costs.
As Tiger Gao, Investor at Apax Digital, pointed out, this control translates directly into pricing power. Users pay for outcomes, not for abstractions.
When models improve, applications benefit without having to rebuild trust from scratch.
Differentiation Lives Above the Model
Another key point was that models converge faster than experiences.
Model performance gaps compress.
APIs standardize.
Capabilities diffuse.
Applications differentiate by:
Domain expertise.
Workflow design.
Data context.
User experience.
Operational integration.
As Zao Chen, Investor at Craft Ventures, emphasized, durable defensibility emerges from how AI is applied — not from the intelligence itself.
Margins Expand Up the Stack
The panel also highlighted a familiar economic pattern:
Margins expand as you move closer to the user.
Infrastructure margins are constrained by cost curves.
Model margins are pressured by competition.
Application margins grow through differentiation and pricing power.
This doesn’t diminish the importance of foundational layers — but it clarifies where sustained value capture occurs.
The Practical Takeaway
AI infrastructure enables the future.
Applications monetize it.
For founders, this means:
Obsessing over workflows, not models.
Owning user trust and integration.
Building differentiation above the foundation.
For investors, it reinforces a familiar truth:
The largest, most durable outcomes are still built at the application layer, even in an AI-first world.
5. Platform Risk Is Real — But Not Fatal
The panel didn’t avoid one of the most sensitive topics in AI investing:
platform risk is real.
Model providers are moving downstream.
APIs are evolving.
Feature parity is increasing.
But the consensus view was notably pragmatic — not alarmist.
Tension Is Inevitable in Platform Shifts
As platforms mature, they naturally look for ways to monetize.
That often means:
Expanding feature sets.
Offering more opinionated tools.
Encroaching on application territory.
As Rak Gard, Partner at Bain Capital Ventures, noted, this tension is not unique to AI. It showed up in cloud, mobile, and SaaS before.
Platforms and applications coexist — sometimes uneasily — because they serve different economic roles.
API Risk Is a Known Variable
Several panelists acknowledged legitimate concerns around:
Access changes.
Pricing shifts.
Deprecations.
Policy updates.
As Lukas Linemayr, Partner at Streamlined Ventures, pointed out, APIs are dependencies — not guarantees. Smart teams model this risk explicitly rather than pretending it doesn’t exist.
Platform risk becomes fatal only when it’s ignored.
Differentiation Isn’t in the Model
The panel repeatedly returned to where applications actually win.
Apps differentiate through:
Workflow design.
Domain expertise.
Product taste.
Brand and trust.
Customer relationships.
As Tiger Gao, Investor at Apax Digital, emphasized, platforms optimize for breadth. Applications win through depth.
That depth is hard to replicate — even for the platform itself.
Competition Reshapes Opportunity
One of the more grounded insights was that competition doesn’t eliminate opportunity; it reshapes it.
When platforms move downstream:
They validate demand.
They educate the market.
They raise baseline expectations.
This often creates new whitespace for more specialized, higher-quality applications.
As Zao Chen, Investor at Craft Ventures, noted, many successful SaaS companies were built after platforms entered adjacent spaces — not before.
Risk Is a Pricing Input, Not a Stop Signal
The panel ultimately framed platform risk the same way investors do:
As a factor to price in, not a reason to walk away.
Founders who understand their dependency surface, design for portability, own the customer relationship, and build real differentiation can survive — and even benefit from — platform competition.
The Practical Takeaway
Platform risk in AI is real.
But it’s not new.
It’s not fatal.
And it’s not a reason to avoid building.
The companies that win:
Acknowledge the risk.
Design around it.
Differentiate beyond the platform.
Move faster than incumbents.
In AI, as in every platform shift before it, value accrues to teams that build where platforms can’t — not where they can.
6. “Quality of Revenue” Now Matters at Seed
One of the clearest shifts highlighted by investors was the earlier evaluation of revenue.
In prior cycles, seed revenue was rare and often enough on its own.
In AI, revenue shows up earlier.
That changes the bar.
Revenue Is Easier to Generate — and Easier to Misread
AI has dramatically compressed time-to-revenue.
Teams can:
Ship quickly.
Demo convincingly.
Monetize early interest.
Close initial contracts faster than ever.
But as multiple panelists emphasized, early revenue is no longer synonymous with a real business.
As Lukas Linemayr, Partner at Streamlined Ventures, noted, the question is no longer “Do you have revenue?” — it’s “What kind of revenue is this?”
The New Questions Investors Ask
Across the panel, investors described a sharper line of inquiry at seed and Series A.
They want to understand:
Durability: Does usage persist after novelty fades?
Depth: Are customers relying on the product, or just experimenting?
Repeatability: Does demand recur, or is it opportunistic?
Expansion: Is there a credible path from $10M to $100M to public markets?
As Rak Gard, Partner at Bain Capital Ventures, emphasized, investors are increasingly underwriting trajectory, not just traction.
Novelty Masks Weak Signals
Several panelists warned that AI novelty can distort early metrics.
Short-term spikes may reflect:
Curiosity.
Experimentation budgets.
Executive mandates.
Fear of missing out.
As Tiger Gao, Investor at Apax Digital, pointed out, these signals look strong in dashboards — but decay quickly if the product doesn’t earn its place in a workflow.
Retention, not activation, tells the real story.
Usage Reveals Business Reality
A recurring theme was that usage behavior is more informative than revenue timing.
Investors look closely at:
Frequency of use.
Depth of engagement.
Reliance during critical moments.
Behavior when the product fails.
As Zao Chen, Investor at Craft Ventures, noted, strong businesses show resilience. Customers return even when things break. Weak ones disappear quietly.
Revenue without usage conviction is fragile.
Scale Tests Everything
Another important point was that scaling reveals quality quickly.
Many AI products can reach $1–5M in ARR through:
Founder-led sales.
Bespoke deployments.
Heavy services.
Early adopter enthusiasm.
The real question is whether the business can:
Standardize delivery.
Reduce marginal cost.
Survive broader scrutiny.
Scale distribution without collapsing economics.
As the panel emphasized, the path from $10M to $100M remains the true test—and AI has not shortened it.
Time-to-Business Maturity Hasn’t Changed
This led to one of the panel’s most grounded conclusions:
AI has compressed time-to-revenue.
It has not compressed time-to-business maturity.
Trust still takes time.
Habits still take time.
Markets still take time.
No model shortcut changes that.
The Practical Takeaway
Revenue is necessary — but no longer sufficient.
For founders:
Focus on usage durability, not just monetization.
Optimize for reliance, not novelty.
Build businesses that survive attention decay.
For investors:
Early revenue is a starting point for diligence, not the end.
In an AI-first world, the quality of revenue matters earlier because it’s easier than ever to get the wrong kind.
7. Taste, Brand, and Community Are Emerging Moats
One of the more surprising — and strongly aligned — themes across the panel was how much intangible moats now matter in AI.
In fact, the investors suggested they may matter more than in traditional SaaS.
Feature Parity Is the New Default
As models converge and capabilities diffuse, feature parity arrives faster than teams expect.
What once felt differentiated — reasoning quality, speed, and output polish — now quickly becomes the baseline.
As Lukas Linemayr, Partner at Streamlined Ventures, noted, when technical advantages compress, competition shifts up the stack — toward how products feel, not just what they do.
Taste Creates Coherence
The panel framed taste not as aesthetics, but as coherence.
Taste shows up in:
Which problems are chosen?
Which features are excluded?
How are workflows structured?
How does the product behave under stress?
As Rak Gard, Partner at Bain Capital Ventures, emphasized, taste is what makes a product feel intentional rather than accidental. In AI products, where outputs are probabilistic, that sense of intention is deeply reassuring.
Coherence builds confidence.
Confidence builds habit.
Brand Is a Trust Shortcut
Brand also took on a more functional meaning in the discussion.
In AI, brand is not about awareness — it’s about trust compression.
As Tiger Gao, Investor at Apax Digital, pointed out, when users don’t fully understand how a system works, they rely on signals. Brand becomes a shortcut for:
Reliability.
Alignment.
Safety.
Intent.
In uncertain environments, trusted brands reduce friction in adoption and forgiveness in the face of failure.
Community Multiplies Distribution and Retention
Community was discussed not as engagement, but as leverage.
Strong communities:
Normalize uncertainty.
Spread best practices.
Reinforce identity.
Accelerate onboarding.
As Zao Chen, Investor at Craft Ventures, noted, community transforms products from tools into shared experiences. That shift increases retention and turns users into distributors.
Community doesn’t lock users in technically — it locks them in emotionally.
Switching Costs Are Becoming Emotional
Perhaps the most important reframe was around switching costs.
In AI, switching costs are often low technically:
Data can be exported.
Integrations are portable.
Models are interchangeable.
But switching costs are high emotionally.
People stick with products they:
Trust.
Identify with.
Feel understood by.
Have invested in learning.
As the panel emphasized, these costs aren’t enforced — they’re felt.
Moats You Can’t Diagram
The panel acknowledged that taste, brand, and community are harder to quantify than traditional moats.
But that doesn’t make them weaker.
In fact, they’re often:
Slower to build.
Harder to copy.
More durable over time.
As one investor summarized, competitors can clone features in months. They can’t clone trust, coherence, or belonging on the same timeline.
The Practical Takeaway
In an AI world defined by rapid convergence, the strongest moats are increasingly human.
They live in:
Product judgment.
Emotional resonance.
Shared identity.
Trust is built over time.
For founders, this means:
Investing in coherence early.
Treating brand as infrastructure.
Designing community intentionally.
For investors, it reframes defensibility.
The most durable moats may no longer be enforced by code; they’re earned through experience.
8. Founder Profiles Are Expanding, Not Narrowing
One of the most encouraging conclusions from the panel was the extent to which the founder archetype is expanding in the AI era. Rather than narrowing the set of who can build venture-scale companies, AI is expanding it.
The Old Pattern Is Breaking
Historically, venture-backed success clustered around a familiar profile:
Elite technical pedigree.
Prior big-tech experience.
Access to capital and networks.
Long lead times to build.
The panel agreed that this pattern is weakening.
As Lukas Linemayr, Partner at Streamlined Ventures, noted, AI dramatically lowers the cost of experimentation. Founders no longer need massive teams or years of infrastructure work to reach meaningful traction.
This opens the door to a much broader set of builders.
Younger Founders Are Succeeding Earlier
Several investors pointed out that founders are reaching real scale earlier in their careers.
AI allows:
Faster iteration.
Quicker feedback from the market.
Earlier revenue.
More compressed learning cycles.
As Rak Gard, Partner at Bain Capital Ventures, emphasized, velocity now matters more than a resume. Teams that learn quickly often outperform those with deeper credentials but slower adaptation.
Domain Expertise Is Rising in Importance
Another major shift discussed was the increasing value of deep domain knowledge.
In many AI categories:
The hard part isn’t building intelligence.
It’s understanding the workflow.
Navigating edge cases.
Earning trust in complex environments.
As Tiger Gao, Investor at Apax Digital, pointed out, founders with lived experience in a problem domain often have sharper product intuition than technically elite generalists.
Knowing what shouldn’t be automated is often more valuable than knowing how to automate everything.
Adaptability Is the New Core Skill
The panel was unified on one point: AI rewards founders who adapt continuously.
Successful founders today must:
Navigate constant model changes.
Reassess architectural decisions regularly.
Update mental models frequently.
Make decisions with incomplete information.
As Zao Chen, Investor at Craft Ventures, noted, the ability to revise beliefs quickly has become a defining trait. Rigid thinkers struggle in environments where assumptions expire every quarter.
Opinionated Thinking Matters More Than Credentials
Another subtle but important theme was the value of opinionated judgment.
With so many tools, models, and paths available, founders who have clear points of view, make decisive tradeoffs, resist chasing every trend, and articulate why they believe something tend to move faster and build more coherent companies.
Pedigree may open doors, but judgment keeps companies alive.
The Founder Archetype Is Broadening
Taken together, the panel painted a clear picture:
There is no single “ideal” AI founder.
Instead, the market rewards:
Speed over seniority.
Learning over lineage.
Judgment over credentials.
Adaptability over perfection.
This is a structural shift — not a temporary one.
The Practical Takeaway
AI is not concentrating on opportunity. It’s distributing it.
For founders, this is a call to lean into:
Lived experience.
Clear thinking.
Fast learning.
Strong opinions.
For investors, it means expanding pattern recognition — not narrowing it.
In the AI era, the founders who win won’t all look the same and that’s a feature, not a bug.
9. Venture-Backed Is a Choice — Not a Default
One of the most refreshingly candid moments in the panel came when the conversation turned to founder paths.
The investors were aligned on a point that’s often left unsaid:
Not every great AI business should be venture-backed.
And that’s not a failure — it’s a feature of the moment we’re in.
AI Has Changed the Economics of Building
AI has dramatically lowered the cost of starting companies.
Founders can now:
Build sophisticated products with small teams.
Reach customers directly.
Generate revenue early.
Operate profitably at smaller scales.
As Lukas Linemayr, Partner at Streamlined Ventures, noted, this fundamentally expands the set of viable outcomes. Venture is no longer the only path to building something meaningful — or enduring.
Niche, Profitable Businesses Are More Viable Than Ever
Several panelists highlighted how AI enables high-quality, niche businesses.
These companies:
Serve specific audiences deeply.
Operate with strong margins.
Grow sustainably.
Don’t require hypergrowth.
As Tiger Gao, Investor at Apax Digital, pointed out, many of these businesses would have struggled to exist a decade ago. Today, they can thrive — and founders can own more of the upside.
Scale isn’t the only measure of success.
Community Enables Profitable Distribution
Another enabling factor discussed was the rise of community-driven distribution.
Strong communities allow companies to:
Reach users directly.
Reduce CAC dramatically.
Build trust faster.
Monetize without heavy spend.
As Zao Chen, Investor at Craft Ventures, noted, community doesn’t just support growth — it supports profitability. For many AI products, that changes the calculus entirely.
Venture Comes With Constraints
The panel was also clear about what venture capital demands.
Venture-backed paths require:
Chasing very large markets.
Tolerating higher risk.
Optimizing for scale over stability.
Committing to outcomes that justify dilution.
As Rak Gard, Partner at Bain Capital Ventures, emphasized, venture is best suited for companies willing to pursue problems that are structurally large — often adjacent to, but not dependent on, AGI-level breakthroughs.
It’s a powerful tool — but it narrows the problem space.
Choosing Venture Means Choosing the Problem
One of the most important reframes was that venture is not just a financing choice — it’s a product choice.
It implicitly commits founders to:
A certain growth rate.
A certain market size.
A certain risk profile.
Founders who don’t want those constraints shouldn’t feel compelled to accept them.
As the panel underscored, opting out of venture isn’t opting out of ambition — it’s opting into a different kind of ambition.
AI Expands the Outcome Space
The broader conclusion was optimistic.
AI doesn’t funnel founders into a single path. It multiplies the paths available.
Some companies should:
Raise aggressively.
Chase massive markets.
Take on existential risk.
Others should:
Stay small and profitable.
Serve communities deeply.
Compound quietly over time.
Both are valid. Both can be impactful.
The Practical Takeaway
AI lowers the cost of building — but it doesn’t dictate how you should build.
Venture-backed is no longer the default. It’s a choice.
The best founders don’t ask:
“Can this raise venture?”
They ask:
“What kind of company do I want to build — and what path best supports that?”
In an AI-first world, freedom of choice is one of the most powerful new advantages founders have.
10. Huge Markets Remain Underserved
Despite how crowded parts of the AI landscape appear, the panel was emphatic on one point: Many of the largest opportunities aren’t crowded at all. They’re simply overlooked.
Silicon Valley Sees a Narrow Slice of the Economy
The panel highlighted a structural blind spot in how markets are perceived.
Inside tech ecosystems, attention clusters around:
Developer tools.
Knowledge work productivity.
Media and content.
Obvious white-collar workflows.
But as Zao Chen, Investor at Craft Ventures, noted, these categories represent a small fraction of global economic activity.
Outside that bubble sit enormous industries that are:
Operationally complex.
Heavily manual.
Under-softwared.
Resistant to prior automation.
These sectors don’t appear on demo days, but they dominate real GDP.
Service Industries Are Still Software-Poor
Several investors emphasized how many service-heavy industries remain untouched by modern software.
Examples discussed included:
Field services.
Logistics coordination.
Healthcare operations.
Compliance-heavy workflows.
Back-office functions in regulated industries.
As Rak Gard, Partner at Bain Capital Ventures, pointed out, many of these markets were poor fits for traditional SaaS. The workflows were too fragmented, too judgment-heavy, or too expensive to automate manually.
AI changes that calculus.
AI Enables Automation Where Software Never Reached
The panel stressed that AI’s most powerful impact may not be where software already exists — but where it never could.
AI can:
Handle ambiguity.
Adapt to messy inputs.
Support human judgment.
Operate across inconsistent processes.
As Tiger Gao, Investor at Apax Digital, explained, this opens entirely new categories. Work that was previously uneconomical to software-enable suddenly becomes tractable.
The opportunity isn’t a marginal improvement. It’s first-time automation.
Visibility, Not Ideation, Is the Bottleneck
Another important reframing was around innovation itself.
The panel rejected the idea that success requires discovering a “new” idea. Instead, it requires:
Seeing existing problems clearly.
Understanding how work actually happens.
Recognizing where human labor is trapped by process.
As Lukas Linemayr, Partner at Streamlined Ventures, noted, many of the biggest AI companies of the next decade won’t feel novel to insiders. They’ll feel obvious — once someone finally builds them.
Underserved Markets Often Look Unattractive Early
One reason these markets remain open is that they rarely look attractive at first glance. They:
Lack clean APIs.
Involve legacy systems.
Require domain expertise.
Don’t fit standard growth narratives.
But as the panel emphasized, these same traits often signal durability. Once solved, these problems create:
High switching costs.
Deep customer reliance.
Long-term contracts.
Real economic impact.
The Practical Takeaway
AI opportunity isn’t concentrated only where attention is loudest. It’s often hiding in:
Invisible workflows.
Neglected industries.
Unglamorous services.
Problems people stopped trying to solve.
The panel’s closing reframe was simple but powerful:
The opportunity is not finding a new idea, it’s seeing an old problem clearly for the first time.
For founders willing to look beyond the obvious, the AI market is still wide open.
11. Hiring and Org Design Are Still Bottlenecks
One of the most pragmatic points the panel made was also one of the least glamorous: AI does not eliminatea eliminate organizational bottlenecks. It often exposes them.
Despite dramatic gains in technical capability, the fundamentals of building and scaling companies remain stubbornly human.
AI Doesn’t Replace Go-To-Market Reality
The panel was explicit that AI does not remove the need for:
Selling.
Onboarding.
Change management.
Domain translation.
Forward-deployed work.
As Rak Gard, Partner at Bain Capital Ventures, noted, many AI companies underestimate how much of the work happens outside the model. Especially in enterprise and regulated markets, trust must still be earned person by person.
Models don’t close deals. People do.
Non-Technical Roles Matter More Than Expected
A recurring surprise for many founders is how critical non-coding roles remain. They become essential when:
Sales cycles are long.
Buyers are non-technical.
Workflows are entrenched.
Adoption requires behavior change.
As Zao Chen, Investor at Craft Ventures, emphasized, AI products often increase the need for translation — not reduce it. Someone still has to explain what the system does, where it works, where it doesn’t, and how to integrate it safely.
That work doesn’t disappear. It shifts.
Forward-Deployed Humans Are Often the Unlock
Several panelists pointed out that forward-deployed teams are not a sign of weakness — they’re often a sign of realism.
In complex environments, humans:
Adapt to messy workflows.
Handle exceptions.
Earn trust in high-stakes settings.
Surface product gaps quickly.
As Lukas Linemayr, Partner at Streamlined Ventures, noted, many successful AI companies scale through forward-deployed work before they scale away from it. The mistake is treating these roles as temporary hacks instead of strategic leverage.
Org Design Determines Where AI Actually Scales
Another key insight was that organizational design determines where AI leverage shows up.
Teams that struggle often:
Over-index on engineers.
Under-invest in GTM and enablement.
Assume automation replaces coordination.
Delay hiring for customer-facing roles.
As Tiger Gao, Investor at Apax Digital, pointed out, this creates a mismatch: powerful technology paired with insufficient human scaffolding. Adoption stalls not because the product is weak — but because the org can’t support it.
Leverage Comes From Deploying Humans Intentionally
The panel emphasized that winning teams don’t eliminate humans; they deploy them strategically. They:
Put humans where judgment matters most.
Automate where repetition dominates.
Keep humans close to customers early.
Pull them back only once patterns stabilize.
This isn’t inefficient. It’s how learning compounds.
The Practical Takeaway
AI changes what humans do not whether they’re needed.
The companies that win:
Design orgs around real-world adoption.
Hire for translation, trust, and judgment.
Accept that some work cannot be automated early.
Deploy humans where leverage is highest.
In an AI-first world, technology scales fastest when organizations are designed to support it.
Ignoring hiring and org design doesn’t make them go away. It just turns them into silent bottlenecks.
12. Governance Will Emerge Bottom-Up, Not Top-Down
When the conversation turned to regulation and governance, the panel aligned around a view that was notably pragmatic:
Governance will not arrive first through policy.
It will emerge through products.
This isn’t ideological — it’s observational.
Regulation Will Always Lag Innovation
The panel was clear that regulation inevitably trails technology.
AI is moving too quickly for:
Comprehensive legislation.
Globally consistent standards.
Real-time regulatory oversight.
As Lukas Linemayr, Partner at Streamlined Ventures, noted, this lag is not a failure of regulators — it’s a structural reality. By the time rules are written, the underlying technology has already shifted.
Waiting for regulation to define governance is therefore unrealistic.
Governance Will Be Built, Not Declared
Instead, governance is emerging bottom-up, through tooling and infrastructure.
The panel emphasized that real governance is operational, not philosophical.
It shows up as:
Auditability.
Observability.
Access controls.
Permissions.
Rollback mechanisms.
Monitoring and logging.
As Rak Gard, Partner at Bain Capital Ventures, explained, these capabilities allow organizations to manage risk before regulation requires it. They become de facto standards because they work — not because they’re mandated.
Trust Is Earned Through Control, Not Promises
Another recurring theme was that trust cannot be asserted.
In AI systems, trust is earned when:
Behavior is observable.
Decisions can be inspected.
Failures are traceable.
Systems can be constrained.
As Tiger Gao, Investor at Apax Digital, pointed out, customers don’t want assurances — they want mechanisms. Products that offer real control are adopted faster than those that simply claim safety.
Compliance Will Be Solved Inside Products
The panel also reframed compliance as a product problem.
Rather than external enforcement, compliance will increasingly be achieved through:
Built-in controls.
Clear boundaries.
Configurable policies.
Embedded audit trails.
As Zao Chen, Investor at Craft Ventures, noted, the most successful AI products treat compliance as an enabling feature — not an afterthought. When compliance is integrated, adoption accelerates instead of slowing.
Tooling Creates De Facto Standards
Over time, the panel expects governance norms to crystallize around what works in practice.
Tools that reduce risk, improve transparency, and support accountability will spread organically across companies, industries, and geographies.
These tools become standards not because they’re required, but because they’re indispensable.
The Final Takeaway
AI governance won’t arrive as a single policy moment.
It will emerge gradually, through:
Observability layers.
Control systems.
Audit tooling.
Product-level constraints.
Trust, safety, and compliance will be built into systems, not bolted on by regulators after the fact.
In the AI era, the companies that define governance will be the ones that operationalize it first — long before anyone tells them they have to.