The Counterintuitive Innovation Principle Behind 70% of Projects Being Canceled
“Seventy percent of projects are destined to be canceled.”
A few years ago, this would have sounded like a demoralizing statement. But amid today’s AI boom, it has become a hard survival rule for some cutting-edge AI labs. They’ve even instituted an almost brutal staffing threshold: “Over four people—and the project graduates.” Once a team exceeds four members, the project must either achieve independent viability—or be disbanded on the spot.
This completely defies conventional management intuition.
In the past, large tech companies placed faith in massive, coordinated teams and long-term planning cycles. Today, however, foundation models iterate weekly, intelligent work agents are updated daily, and the hierarchical reporting structures of yesteryear have become friction points that stifle innovation. Writing code now takes seconds; what determines a project’s life or death is no longer R&D capacity—but speed of experimentation.
The winds have shifted. Heavy internal “horse-racing” mechanisms are receding. Instead, organizations are converging—unanimously—on an incubator model built around small teams, high attrition rates, and rapid iteration.
Many breakout applications emerge from the rubble of dozens-strong “failed teams.” When projects are canceled, people naturally feel disheartened—but the insights gained through trial and error must remain. This high-mobility logic forces organizations to function like living systems: constantly dissolving boundaries and minimizing communication overhead.
Here arises the question: When “reorganization at any time” becomes the norm, what digital foundation can enterprises rely on to support such extreme fluidity?
Piling up a suite of disconnected tools only makes the system increasingly bloated. A truly effective agile foundation must seamlessly integrate everyday capabilities—real-time messaging, documents, calendars, video conferencing, and multi-dimensional spreadsheets. Information naturally settles as it flows; collaboration feels frictionless.
Innovation resembles plant breeding: placing bets, incubating ideas, graduating successful ones, and eliminating failures—a complete cycle. The soil supporting all this cannot be compacted, hardened earth.
Breaking Down Organizational Boundaries: High-Mobility Collaboration in Ad Hoc Teams
Compacted, hardened earth cannot grow agile root systems. Since “70% of projects are destined to be canceled” has become common ground, assembling innovation teams is no longer about methodical deployment—it’s a series of frequent, high-frequency “ad hoc raids.”
A typical experimentation cycle unfolds like this: pull three to five key contributors from different business units to form a temporary task force. There’s no lengthy, cumbersome project initiation process—just a quick alignment session where everyone gets on the same page.
If the tools are difficult to use at this stage, the spark of innovation dies instantly.
On Lark’s digital foundation, this high-mobility collaboration is reduced to the most intuitive—and purest—actions possible. The team lead creates a real-time messaging group and pins core documents and multi-dimensional spreadsheets to the top; opens the calendar, checks teammates’ availability, and locks in a two-hour window for the upcoming review; and if cross-regional collaboration arises, a one-click video conference becomes the decisive forum—where the team decides whether to continue, pivot, or cancel the project.
“No need to struggle learning a new system—just focus on the problem itself.” This is the most authentic user experience reported by frontline innovators.
When these everyday capabilities combine, ad hoc teams achieve true “plug-and-play” functionality. If a project fails validation, the chat group is archived, calendar slots are released, and meeting notes automatically settle as digital assets. Team members disband smoothly—and stand ready to regroup for the next “bet.”
There’s no tedious permission handover, nor visible information silos.
Throughout this process, organizational boundaries aren’t violently shattered—they quietly dissolve through the free flow of information. Real agility isn’t about forcing individuals to run faster; it’s about ensuring teams feel no weight from the system when reorganizing.
Turning Failure into Assets: Digitally Capturing Lessons from Experimentation
That 70% of projects get canceled looks, under traditional management thinking, like pure sunk cost—wasted resources. Yet in an agile organization built on high-mobility experimentation, those 70% represent the most valuable digital assets.
Cancelling a project isn’t an endpoint—honest, structured retrospection extracts the nutrients needed for innovation.
Once a decision is made, teams avoid getting mired in blame-shifting. Instead, they open a document and conduct a structured retrospective. This captures not only the evolution of business logic and discrepancies in data validation—but also the raw, unfiltered emotions of frontline innovators.
“When a project is canceled, people naturally feel disappointed,” reflects a research & development leader who has experienced multiple “graduations.” In the document’s margin space, the pain of failure transforms into searchable, reusable experience.
If documents serve as the “container” for capturing experience, then multi-dimensional spreadsheets act as the “sandbox” that oversees the entire landscape.
Every project receiving a “bet” has a clearly defined lifecycle within the multi-dimensional spreadsheet—from initial inspiration and prototype validation, through data tracking and status transitions, all the way to final “graduation” or “cancellation.” Every milestone is visible at a glance. It’s long since ceased being merely a progress tracker—it’s now a dynamic innovation asset repository.
When launching a new project, teams don’t start from scratch, blindly feeling their way forward. They filter past validation data for similar initiatives in the multi-dimensional spreadsheet; they browse earlier documentation to see which pitfalls others already encountered—and suddenly, confidence emerges.
Failure is no longer hidden shame—it’s structured nourishment.
Each “bet” leaves a traceable record; each “cancellation” is transparent and explicit. When lessons from experimentation and validated data truly become shared digital assets across the organization, the next wave of innovation can confidently build upon predecessors’ “failures.”
Eliminating Tool Friction: Seamless AI Collaboration Within an Open Ecosystem
Innovation pace in the AI era is measured in weeks—not months or years. Yet in reality, team members’ time is often fragmented beyond recognition.
Writing documents, reviewing data, aligning on design mockups, running code… every switch between applications drains creative flow. Tools grow increasingly unwieldy; systems become ever more fragmented—silent killers of agility.
“If I have to toggle across three windows just to locate a file, my enthusiasm for innovation cools down before the page even finishes loading.”
Lark’s solution is reduction—not expansion. Rather than attempting to dominate every vertical domain with deep, specialized functionality, Lark steps back to serve as a unified collaboration entry point.
Built upon an open ecosystem foundation, Lark seamlessly integrates everyday capabilities—real-time messaging, documents, calendars, and video conferencing—with diverse third-party AI assistants and intelligent work agents.
This is not simply stacking APIs—it’s deeply weaving capabilities into context.
While structuring product logic inside a document, users directly link AI-powered office tools to generate analytical frameworks; while discussing requirements in a real-time messaging group, intelligent work agents intervene instantly—organizing scattered conversations into coherent structure.
No need to memorize complex commands, nor jump repeatedly across disparate systems.
As boundaries blur among design, data, coding, and day-to-day collaboration, tools cease to be stumbling blocks—and instead become extensions of human capability. Cognitive load drops to a minimum; operational friction disappears entirely.
Tools must adapt to people—not the other way around. Only when systems stay lightweight can teams channel full attention toward scenarios demanding deep, uniquely human insight.
Beyond Technical Equity: Domain-Specific Depth Is Innovation’s Final Barrier
Foundation models have lowered the barrier to code generation to near zero. And a harsh truth surfaces alongside it: technical equity does not equal innovation equity.
AI can generate elegant architectures in seconds—but it cannot replace your ability to detect users’ subtle, hidden pain points within a specific business context. Intelligent work agents can automate data pipelines—but they cannot define a product’s aesthetic sensibility or commercial intuition.
Beneath the AI acceleration, foundational development capabilities are rapidly leveling out. The real barrier has long since shifted—to deep domain expertise and relentless, granular engagement with real-world scenarios.
This is precisely the ultimate purpose of an agile collaboration foundation.
By interweaving everyday capabilities—real-time messaging, documents, multi-dimensional spreadsheets, and video conferencing—Lark handles universal internal connections and information flow. It functions like a silent nervous system, reliably absorbing the entropy generated by high-frequency experimentation.
The system handles the underlying flow—so people can return to creation.
Innovation teams can thus dedicate their full energy to business validation and scenario-specific depth. They track gray-scale feedback in multi-dimensional spreadsheets, capture failure insights in documents, and maintain the delicate balance between compliance and flexibility within approval workflows.
It’s not frightening that 70% of projects will inevitably be canceled.
What’s truly alarming is wasting away our acute, scenario-specific intuition in inefficient collaboration. When the foundation minimizes the cost of experimentation—and when organizational resilience continuously reshapes itself through fluidity—we must ask ourselves one final question:
In an era where AI accelerates everything—has your team prepared its “blunt-force perseverance” in deep scenario engagement to meet the next innovation cycle?