In its 2025 Annual Retail Report, global fintech platform Adyen reveals a sharp rise in AI adoption among Asia Pacific (APAC) consumers and retailers, highlighting a significant shift in shopping behaviors and business strategies across the region.

The report, based on a survey of 41,000 consumers across 28 markets including Singapore, Australia, Hong Kong, India, Japan, and Malaysia, underscores how the tech is reshaping the retail experience and signaling broader trends for the region’s digital economy.

According to Adyen, over a third (38 per cent) of APAC consumers now use AI to assist with shopping—a 39 per cent increase from 2024. Notably, more than one in ten APAC consumers (11 per cent) tried AI-powered shopping for the first time in the past year.

The appeal of AI lies largely in its ability to offer fresh inspiration and personalised recommendations. Nearly two-thirds (63 per cent) of AI-using consumers said it helps them discover new choices for everything from outfits to meals faster than any human assistant could.

Additionally, 62 per cent expressed interest in using AI to find unique brands and shopping experiences, opening doors for retailers to drive sales through partnerships and cross-selling strategies.

Cross-generational adoption

While younger generations remain at the forefront of AI shopping, older cohorts are quickly catching up. In Malaysia and Hong Kong, Gen Z adoption stands at 74 per cent and 64 per cent, respectively.

Also Read: Ecosystem Roundup: AI’s capital frenzy, bolttech’s US$147M funding, and Southeast Asia’s VC crunch

Meanwhile, Singapore’s Generation X and Millennials have shown substantial growth, with AI shopping adoption increasing by 45 per cent and 28 per cent respectively over the past year. Even among consumers aged 60 and above, nearly a third (30 per cent) reported using AI to assist with purchases.

“The introduction of AI in shopping has created new shopper journeys that are more exciting than ever,” said Warren Hayashi, President of Asia Pacific at Adyen. “For retailers, embracing AI isn’t just about staying current; it’s about meeting evolving consumer expectations and staying competitive in a fast-changing retail landscape”.

On the business side, more than a third (34 per cent) of APAC retailers plan to increase their AI investments in the coming year to enhance sales, marketing, product innovation, and security. Payments data — a largely untapped resource — presents significant potential for AI-driven optimisation.

While AI garners attention, the report also points to gaps in omnichannel capabilities. Less than half (46 per cent) of APAC retailers currently support seamless cross-channel shopping, though another 19 per cent plan to enable this within the next year.

Consumer expectations are evolving quickly: 46 per cent want businesses to offer integrated experiences across online platforms, social media, and physical stores. Despite the rise of digital commerce, 42 per cent of consumers still value in-store shopping equally alongside online channels.

As the region’s retail landscape continues to digitise, AI is emerging not only as a tool for personalisation and convenience but also as a strategic differentiator for retailers navigating an increasingly competitive market.

Also Read: The art of artificial intelligence: How Hagia Labs is reimagining creativity

Balancing innovation with security concerns

Despite enthusiasm for AI, concerns around fraud persist. About 26 per cent of consumers expressed heightened worries about scams, and 20 per cent avoid storing payment details on devices for security reasons. Currently, 40 per cent of APAC retailers are leveraging AI to combat fraud by detecting anomalies and predicting fraudulent activities using their transaction data.

“Besides optimising revenue, AI could aid in the fraud-fighting efforts of retailers,” Hayashi noted. “It can spot anomalies, identify patterns, and predict fraud attempts – ultimately ensuring consumer trust and protecting retailers’ hard-earned revenue”.

—

Image Credit: Mike Petrucci on Unsplash

The post AI shopping adoption surges 39 per cent in APAC, fueling retail tech investments appeared first on e27.

In 2024, venture capital across Asia-Pacific sank to its lowest level since the 2021 peak. In Southeast Asia, startup funding dropped by 42 per cent, with investors either pulling back entirely or doubling down only on high-traction, near-profitable, or already profitable startups. At Spacely AI, we had no choice but to rethink everything.

Early on, we made a decision: build a product people would pay for, and structure our growth around revenue—not runway. We raised a modest pre-seed round, stretched every dollar, and aimed for profitability from day one. We didn’t call it seed-strapping at the time—but that’s exactly what it was.

We haven’t reached profitability yet. But this approach extended our runway far beyond projections. It allowed us to keep our team small—under 10 full-time employees. We avoided layoffs. And unexpectedly, it gave us something most founders struggle to find during turbulent times: leverage, clarity, and freedom.

The VC model is breaking in Southeast Asia

The region’s venture landscape is facing serious headwinds. Billion-dollar exits are few and far between. And without reliable exits, LPs are more cautious, which makes it harder for VC funds to raise capital. Many are quietly failing to raise their third or fourth fund. It’s not because they can’t find good startups, but because the math no longer adds up.

The silver lining? There’s still capital out there—but it’s more selective than ever. It’s reserved for companies showing real traction and a clear path to profitability. The bar has shifted. The days when “potential” alone could raise millions are over. That’s the new reality: many VCs simply can’t invest, not because they don’t believe in you—but because they’re trying to survive, too.

Also Read: Building future sustainable business: The role of rural commerce platforms

The rise of seed-strapping

This is why the smartest founders I know are shifting to the “seed-strapping” model. Seed-strapping is quickly becoming the new startup playbook—raise just enough capital to reach cash-flow positive, then let revenue take you the rest of the way. You don’t need a massive seed round. You need just enough to reach profitability.

We stayed lean with fewer than 10 FTEs, automated as much as possible with AI, and focused entirely on finding product-market fit. We didn’t grow through expensive paid campaigns. Our customer base and revenue were built through organic acquisition. That forced us to stay disciplined. No distractions. Just sell, build, test, repeat.

Let me be clear: this path is not easy. At one point, we reduced salaries across the entire company by 50 per cent. It was painful, but necessary. It wasn’t about bravado. It was about survival. And in the midst of this, we found focus. That constraint gave us perspective. And it opened our eyes to the real power of AI—not just as a product, but as a company-building force.

AI is the deflationary force changing everything

One of the biggest tailwinds behind seed-strapping is AI. Not just because we’re an AI company—but because it changed how we work, scale, and think about cost.

Every founder faces the same three levers: raise money, cut costs, or grow revenue. And AI can supercharge all three. We’ve trained ourselves to ask: “Can we AI this before hiring for it?” For example, at Spacely AI, we run all our growth channels (social media, blog, and SEO) with one growth analyst. That analyst is empowered with the right AI tools, templates, and workflows to do the job of an entire team. The result? Lower cost, more output, and better quality.

AI didn’t just help us survive. It helped us operate better. Founders who understand this dynamic—who treat AI as a margin engine, not just a product feature—are going to win.

Revenue is the best funding you can get

Cutting costs and increasing productivity only get you so far. The other side of survival is revenue. That’s where real leverage lives.

VC money is useful. But customer money is better. Revenue is non-dilutive. It’s fast. It’s proof that you’re solving a real problem. And every US$10,000 in MRR buys you more than just another month of runway—it gives you proof.

Most startup advice focuses on perfecting your pitch. But what if you pitched less and sold more? What if you built your business around the customers you’re trying to serve—not the investors you’re trying to impress?

There’s a quote I read recently that feels especially true in this climate: “Profits solve all problems.” Reflecting on our journey, I couldn’t agree more.

Also Read: Turning intimidation into innovation: Embracing sustainability’s new opportunities

The new playbook: PMF, margin, and discipline

If you’re building a startup in Southeast Asia right now, I’d challenge you to adopt this lens. The old “growth at all costs” mentality doesn’t fit the current market. The new playbook looks like this:

• PMF first: Lock in one clear use case. Prove it. Then scale.
• Profitable unit economics: 70 per cent+ gross margin, 12-month payback period or better.
• Lean teams, AI-enabled: Ten high-performers with AI > 50 without.

We didn’t invent this strategy. We adopted it out of necessity. But it’s made us sharper and more resilient.

Profitability is the ultimate leverage

There’s a saying: “The best time to raise money is when you don’t need it.” Every founder loves that phrase—and for good reason. Once you approach or hit profitability, the entire game changes.

You get clarity—on what to build, who to build for, and what it takes to scale. You gain options. Not just the option to raise or not raise, but the power to choose who to raise from—and who to walk away from.

Let’s be honest: fundraising takes time. Some say six months. Lately, I’ve heard twelve. Profitability—or even a credible path to it—gives you the endurance to survive those cycles. More importantly, it keeps you in control.

The founder’s checklist for surviving the funding winter

If you’re navigating this market, here’s what I recommend:

• Solve a painful problem customers will pay for
• Use AI to increase productivity and stay lean
• Focus on leading indicators of PMF—activation, engagement, referrals (for SaaS)
• Track your burn rate, CAC, and LTV
• Rally your team around cash-flow positive as a company-wide goal
• Treat VC funding as optional fuel, not oxygen

Final thoughts

We’re in a funding winter. But winters don’t last forever—and they often produce the most resilient companies. If you can build around customers, automate smartly, and seed-strap your way forward, you’ll emerge stronger, faster, and freer.

At Spacely AI, we chose to seed-strap because we didn’t want to depend on a volatile capital market. AI helped us stay lean. Our focus on customer value gave us breathing room. And our users—real people paying real money—turned out to be the best VCs we could ever ask for.

If you’re building in 2025, don’t wait for a term sheet to start acting like a real company. The new playbook is clear: sell first, build something people want, and spend your customers’ money wisely.

—

Editor’s note: e27 aims to foster thought leadership by publishing views from the community. Share your opinion by submitting an article, video, podcast, or infographic.

Join us on Instagram, Facebook, X, LinkedIn, and our WA community to stay connected.

Image courtesy of the author.

The post AI, seed-strapping, and the new playbook: Why customers are the best VCs appeared first on e27.

At a recent AI vs Human keynote showdown, someone in the audience threw me a question many founders quietly ask: “But AI hallucinates. Isn’t that dangerous?”

My reply was simple, but it caught a few off guard: “Yes. But humans hallucinate, too. And often, it’s far more dangerous.”

The debate isn’t whether AI makes mistakes — we know it does. The real problem is who we choose to trust when confidence meets uncertainty. As founders, that’s where the true risk lives.

What is hallucination, really?

Let’s start by demystifying the term.

AI hallucination happens when large language models (LLMs) like GPT generate responses that are factually incorrect but sound completely plausible. They aren’t lying. They’re simply predicting text based on probability patterns.

Public examples prove this risk. Sky News’ Sam Coates confronted ChatGPT live for generating false podcast transcripts. OpenAI’s own testing data shows significant hallucination rates:

• 33 per cent false information rate for its o3 model.
• 48 per cent for its o4-mini model.

AI can sound extremely confident even while being wrong, and that’s precisely what triggers automation bias, when humans trust machine outputs simply because they “sound right.”

But here’s the uncomfortable truth: Humans hallucinate too, and we rarely catch ourselves doing it.

The human hallucination problem: Narratives we build

AI hallucinates through prediction. Humans hallucinate through narrative.

We build impressions. Those impressions become judgments. Judgments turn into stories. And those stories drive our business decisions.

• We overestimate market size based on a handful of customer interviews.
• We assume product-market fit based on early interest.
• We hire poorly because of a great interview.
• We raise funding on projections fuelled more by hope than data.

These aren’t rare. They are startup norms.

In many cases, founders hallucinate entire business models with full conviction. The difference? There’s rarely a system that alerts us when we’re slipping into narrative-driven delusion.

Also Read: AI adoption in SEA e-commerce: The clock is ticking for sellers

The confidence trap: Why founders trust the wrong things

Both AI and humans share one dangerous similarity: They deliver outputs with confidence, whether right or wrong.

That confidence triggers trust. And trust, unchecked, leads to bad decisions.

• AI: “The answer is definitely X.”
• Founder brain: “We’ll definitely 10x next year.”

The issue isn’t hallucination itself, it’s how quickly we surrender our skepticism when something sounds certain.

The true founder risk isn’t just AI hallucination. It’s our reflex to accept confidence as truth.

My operator view: How I design around hallucination

Across my ventures, I’ve built AI into daily workflows. But I never outsource my thinking.

Here’s my personal system design:

• Separate generation from verification: AI helps structure thoughts, draft options, and synthesise. But facts get independently verified.
• Build multi-step logic chains: I don’t ask for one-shot answers. I design prompts that generate reasoning, assumptions, counterpoints, and validations.
• Cross-check everything: Whether it’s market data, analysis, or competitor signals, I verify across multiple sources.
• Use AI as augmentation, not authority: Seraphina AI, my personal assistant, mirrors my thought process because it was trained to follow how I already operate. It amplifies my logic but doesn’t replace it.

The meta-moment: While writing this article

Even while drafting this article with AI assistance, I actively ask: “Is the AI hallucinating here?”

The answer? No, because I’m not asking it to invent facts. I’m using it to structure my thinking, arrange arguments, and explore narrative flows. The core reasoning remains mine, AI simply amplifies and organises.

AI is my logic partner, not my fact source. That distinction is where most founders struggle: they surrender too much authority too quickly.

The founder’s three guardrails against hallucination

Here’s the framework I live by and recommend to every founder:

• Separate generation from verification: Never let AI verify its own outputs. Always layer external data and checks.
• Build multi-step prompts: Don’t chase immediate answers. Build prompt chains that explore reasoning, objections, and edge cases.
• Treat AI like a team member: You wouldn’t trust a junior hire’s first draft without review. Apply the same discipline to your AI assistant.

Also Read: Startups, is your email strategy driving growth, or just gathering dust?

The harder truth: Human hallucination is more dangerous

The brutal reality? We can engineer systems to reduce AI hallucinations. But human hallucination, especially founder hallucination, is far more difficult to catch.

• Ego pushes us to double down on flawed assumptions.
• Investor pressure accelerates premature scaling.
• Team echo chambers reinforce dangerous narratives.
• Emotional attachment clouds product decisions.

Human hallucination isn’t probabilistic — it’s emotional. And emotions rarely fit into predictable guardrails. That’s why many startups fail — not from AI errors, but from founders’ unchecked certainty.

AI hallucination is mechanical. Human hallucination is narrative.

The founder advantage today isn’t about trusting AI more or less. It’s about developing the cognitive discipline to manage both AI and human fallibility simultaneously.

The hybrid founder edge

The founders who thrive in this AI-powered era won’t be those who fear hallucination.

They’ll be the ones who:

• Build operating systems that minimise blind spots.
• Maintain cognitive sovereignty over both algorithms and their own internal narratives.
• Use AI to amplify clear thinking, not replace it.

AI doesn’t replace thinking. It exposes who never learned how to think systematically in the first place. And in this new landscape, that, not hallucination itself, will define who scales and who fails.

—

Editor’s note: e27 aims to foster thought leadership by publishing views from the community. Share your opinion by submitting an article, video, podcast, or infographic.

Join us on Instagram, Facebook, X, LinkedIn, and our WA community to stay connected.

Image credit: DALL-E

The post Why founders should fear their own narratives more than AI’s mistakes appeared first on e27.

As we reach the midpoint of 2024, it’s an ideal time to reflect on emerging trends that have shaped our perspectives. For me, it’s Multimodal Large Language Models (MLLMs).

2023 was a game-changer for AI, no thanks to ChatGPT. We saw a surge in large language models (LLMs) and generative AI, which made everything from chatting with bots to getting content way faster and better.

I won’t lie — I wasn’t very fond of the AI hype. Seeing everyone generate low-quality stock images for their posts and slides and being wowed by trivial advancements was honestly quite frustrating.

While the generative AI hype still prevails, I do have to admit it is maturing. Slowly.

These advancements and the growing consumer adoption of AI technology have paved the way for what we’re seeing in 2024: the emergence of multimodal AI models (MLLMs).

‘Multimodality’ is a somewhat new term for an old concept, i.e. the way humans have always learned about things. People have always gathered information through various senses like sight, sound, and touch. Then, our brains merge these different types of ‘data’ to create our understanding of reality.

So basically, multimodal language models are advanced AI systems that can process and understand multiple types of data, like text, images, audio, and video, all at once. I know, shocking, right?

Ironically, many people have interacted with aspects of multimodal AI without even realising it.

I didn’t even realise I was building my startup, LFG, around this technology and its concept. Learning about multimodal AI has completely shifted my views on its implications and potential.

The rise of visual commerce in retail

This is an emerging trend that has stood out for me this year

So far, the exciting trend I’ve seen in 2024 is the rise of visual commerce, especially in the fashion and beauty sector. Multimodal AI is making waves here by enabling consumers to use natural language, images, and videos to search and buy, transforming how we shop for clothes, accessories, and beauty products.

Also Read: From mining engineer to travel tech visionary: Darryl Han transforms trip discovery

In the US, startups focusing on multimodal search have received significant funding and support, like Daydream (US$50M seed funding) and Lumona (YCW24), underscoring the growing importance of this technology.

With ViSenze (a Singapore tech company at the forefront of multi-search), for example, you can upload a photo of a dress you love (even from a social media post), and their AI-powered search will find similar styles available for purchase. This makes shopping more engaging and personalised, and it’s clear that visual content is becoming a major player in retail decisions.

The shift towards personalised travel experiences

Insights gained about the direction of the travel industry

While this technology is being experimented with and refined in the fashion and beauty sectors, I believe its potential impact on the travel industry is even more profound. Travel encompasses a whole range of services and experiences, from flights and hotels to tours and local attractions. You can already see how big this sector is on its own.

There is a shift in consumer mindsets that personalisation is no longer optional — it’s essential. Multimodal AI can simplify and personalise these offerings by analysing a combination of text, images, and videos, making it easier for travellers to discover exactly what they’re looking for and enhancing their overall experience.

For instance, a traveller searching for “hidden, speakeasy late bars in Kuala Lumpur” can benefit from an AI that not only processes the textual description but also analyses images and videos to find the perfect match. This leads to more precise and personalised recommendations, enhancing user satisfaction.

Also Read: Into the metaverse: How to extract real business value from the hype?

Some further implications of multimodal AI for travel that I’m eager to build and see developed include the following:

• Dynamic pricing: Adjusting prices and offers in real-time based on market trends and user behaviour, maximising revenue and satisfaction.
• Streamlined bookings: Understanding natural language queries and providing instant booking assistance and results, improving user experience.
• Smart assistants: Offering real-time support with voice commands, travel document and location analysis, and instant translations, making travel easier and more enjoyable.

Embracing multimodal AI for future growth

Lesson learned that will shape my approach for the rest of 2024

As multimodal AI continues to advance, it will undoubtedly shape the future of any form of commerce, driving growth and enhancing the overall travel and shopping experience for consumers worldwide.

For travel startups like ours, we’ll definitely be exploring and leveraging these multimodal applications to redefine how travellers search, discover and experience the world — bringing a more intuitive and enjoyable journey before the trip begins.

—

Editor’s note: e27 aims to foster thought leadership by publishing views from the community. Share your opinion by submitting an article, video, podcast, or infographic.

Enjoyed this read? Don’t miss out on the next insight. Join our WhatsApp channel for real-time drops.

Image credit: Microsoft

 

The post Multimodal AI: Reshaping search and discovery in retail and travel appeared first on e27.

Seedefy, a Singapore-registered company, began with a proposition that resonated with the region’s startup ecosystem. It positioned itself as a bridge between early-stage founders and capital, promising structure, access, and a clearer path through the fragmented world of seed funding. In a market where many first-time founders struggle to navigate investors, accelerators, and legal complexity, the narrative proved compelling.

For a period, that positioning gained traction. Seedefy attracted attention, partnerships, and a growing community of founders who viewed it as an entry point into the funding landscape. Its pitch aligned with Southeast Asia’s appetite for platforms that simplify complexity and lower barriers to entry. Momentum followed, as is often the case in early-stage ventures.

That trajectory later shifted, and the shift was comparatively rapid.

When momentum masks fragility

As with many young startups, Seedefy’s perceived growth appeared to advance more quickly than its underlying foundations. Expansion of scope progressed faster than the development of internal controls. Governance arrangements, incentive structures, and execution capacity became increasingly difficult for external stakeholders to evaluate, even as visibility increased.

This sequence is familiar. Early traction tends to foster confidence, which in turn reduces scrutiny. Investors and partners extrapolate short-term signals into assumptions of durability. In Seedefy’s case, the narrative retained credibility even as signs of structural strain became more apparent.

When doubts began to surface more broadly, they translated into erosion rather than prolonged decline. Confidence weakened. Relationships became strained. The business model showed limited resilience under closer examination. What followed was not a dramatic collapse, but a relatively swift loss of relevance. The platform receded from the centre of the ecosystem it had sought to organise.

Also Read: Why due diligence matters, especially when investing in early-stage startups

Due diligence rarely fails loudly

Seedefy’s trajectory appears less rooted in overt misconduct than in layered assumptions. Investors anticipated governance would mature over time. Founders expected scale to address early gaps. Partners inferred alignment where documentation, incentives, and regulatory positioning were not always clearly articulated.

One element proved particularly consequential. Certain aspects of the model operated close to regulated financial and crypto-adjacent activities, in a context where licensing requirements and regulatory boundaries were not always clearly delineated. While this point rarely dominated discussion, it increased exposure once scrutiny intensified.

Due diligence often falters in this understated manner. The issue is not missing documentation, but deferred questions. Who ultimately holds decision-making authority? How conflicts are resolved. How regulatory considerations are managed as models evolve. How downside scenarios are addressed in practice.

In early-stage investing, such questions are frequently postponed. Speed, access, and fear of missing out tend to prevail. The cost typically emerges later.

The Founder remains the central risk

What Seedefy illustrates is a reality many investors acknowledge privately but underweight in practice. The founder remains the most influential variable in any early-stage company.

Markets evolve. Products pivot. Strategies adjust. Behavioural patterns, decision-making styles, and approaches to accountability tend to display greater continuity.

Effective due diligence, therefore, extends beyond pitch decks and data rooms. It involves examining a founder’s operating history with care, including discussions with former colleagues, employees, contractors, and prior investors. The objective is not to seek consensus, but to identify consistency in how pressure was handled, conflicts addressed, and responsibility assumed when outcomes disappointed.

Such conversations rarely yield definitive judgments. They do, however, add depth. They transform narrative into context. For investors, that additional dimension is often decisive.

Integrity, responsibility, and how founders exit matters

The contrast becomes clearer when viewed alongside how other founders have handled comparable outcomes. In recent years, a number of early-stage founders have chosen a more explicit approach when ventures failed to meet expectations. They communicated directly with investors and partners, acknowledged misjudgements, and remained accessible after operations ceased.

A recent example shared publicly by a founder illustrates this approach clearly, documenting the decision to wind down a startup with transparency, personal accountability, and continued engagement with stakeholders even after commercial prospects had ended.

Also Read: ‘Due diligence is like dating before the long-term marriage’: Accion Venture Lab’s Paolo Limcaoco

In some cases, founders publish post-mortems. In others, they remain reachable long after operations have stopped. These actions rarely alter financial outcomes, but they materially shape trust. Investors are left informed rather than uncertain. Employees and partners receive closure rather than silence. The failure of the venture does not extend into a failure of responsibility.

This distinction matters. In early-stage companies, failure itself is rarely disqualifying. How founders behave once momentum breaks often proves more revealing than how they behave during periods of growth. Transparency under pressure, willingness to engage when prospects dim, and respect for stakeholder relationships tend to persist into future ventures.

From a due diligence perspective, this underscores the value of examining not only how founders build, but how they unwind. Past shutdowns, difficult chapters, and public accountability often provide more signal than success stories alone.

The asymmetry of startup risk

Early-stage investing remains defined by asymmetry. Upside attracts attention. Downside determines outcomes.

Seedefy demonstrates how non-financial risks can shape results. Governance risk. Execution risk. Regulatory exposure. Alignment risk. These factors resist simple modelling, yet they frequently influence survival.

For angel investors and early backers, narrative proximity can feel reassuring. Distance and scepticism tend to offer greater protection.

A broader warning for the ecosystem

Southeast Asia’s startup ecosystem continues to mature. Capital is more accessible. Structures appear more sophisticated. The fundamentals of risk remain unchanged.

Platforms such as Seedefy emerge because they address genuine pain points. Their failure does not negate those problems. It reinforces the importance of discipline on the investment side of the table.

Notably, Seedefy did not conclude with a clearly communicated endpoint. Activity diminished. Public communication subsided. The company appeared to wind down without a formal announcement or resolution. For investors, that absence of closure carried its own implications.

Due diligence is not an administrative exercise. It sits at the core of early-stage investing. When treated as secondary, outcomes tend to converge toward disappointment.

What investors should take away

The rise and fall of Seedefy offers a restrained but instructive reminder. Early momentum does not ensure durability. Visibility does not guarantee governance. Access does not confer protection.

Investors who remain active over time tend to develop a preference for structure, context, and downside analysis. They ask fewer aspirational questions and more operational ones. In early-stage investing, those questions often separate informed risk from avoidable loss.

The cost of overlooking them is rarely immediate. When it materialises, it is usually conclusive.

—

The post A study on what the rise and fall of Seedefy reveals about due diligence in early-stage investing appeared first on e27.