Wisk Aero and the Decade-Long Lesson the eVTOL Industry Is Still Learning
What ten years of autonomous air taxi development reveals about certification, safety cases, and the hard limits of moving fast in aviation.
There is a version of the eVTOL story that gets told constantly: a bold startup, a compressed timeline, a vision of urban air mobility that will transform cities within five years. The investors love it. The press releases are immaculate. And then the aircraft hits the FAA certification process, and the timeline quietly doubles.
Wisk Aero has lived the longer version. Backed by Boeing and emerging from Google co-founder Larry Page’s Kitty Hawk project, Wisk has been developing autonomous electric vertical takeoff and landing aircraft since roughly 2015 — longer than almost every other active eVTOL program. Their Cora aircraft has logged more autonomous flight hours than any competitor platform. Their engineering organization has iterated through multiple aircraft generations, multiple regulatory frameworks, and multiple honest reckonings with what autonomous aviation actually requires.
That history is not a liability. It is the only thing in the eVTOL space that cannot be bought or shortcut: demonstrated time in the problem.
The Autonomous Difference
To understand what Wisk has learned, you have to start with what makes autonomous aviation structurally different from piloted aircraft certification — and why that difference compounds in ways that are not obvious at the outset.
In piloted aircraft certification, the human pilot is part of the safety system. The certification case acknowledges that a trained professional is in the loop, able to respond to anomalies, exercise judgment, and intervene. This does not make certification simple — it is never simple — but it provides a known reference point. Decades of NTSB data, simulator testing standards, and airman certification frameworks give the FAA a shared vocabulary with applicants.
Remove the pilot, and that vocabulary disappears. Every question that was answered by “the pilot will handle it” must now be answered by a software function, a sensor suite, a communications link, or a redundant hardware architecture. The safety case must account for failure modes that in piloted aircraft are mitigated by human judgment. The requirements document for an autonomous aircraft is not longer than a piloted aircraft’s requirements document. It is categorically different.
Wisk encountered this early, and their public statements from the 2018-2021 period reflect an organization working through what it actually means to build a certification case without a pilot. Their approach shifted from attempting to adapt existing FAA Part 23 and Part 135 frameworks to engaging in what the FAA now calls “issue paper” development — a process of co-defining the regulatory framework alongside the agency, rather than submitting to a pre-existing one.
That shift cost time. It also means Wisk is now operating inside a regulatory framework they helped design, while competitors arriving later are operating inside frameworks they inherited without context.
Certification as a Moving Target That Eventually Stopped Moving
Between 2018 and 2022, the FAA’s approach to advanced air mobility certification evolved faster than most applicants could track. MOSAIC rulemaking, the Part 23 rewrite, AC 21.17-1 for special class aircraft, the BEYOND program, the Advanced Air Mobility Implementation Plan — the regulatory landscape was not just uncertain, it was actively being constructed in real time.
Companies that entered eVTOL development in 2019 or 2020 often built their initial certification plans around frameworks that were superseded before the aircraft reached flight testing. Wisk, having been in the space since 2015, had the uncomfortable advantage of having already built certification plans around superseded frameworks — and having already rebuilt them.
The practical lesson Wisk embeds in their engineering culture is one the rest of the industry is still absorbing: certification planning is not a one-time activity that happens before development and after development. It is a continuous systems engineering function that must be maintained in parallel with the hardware program. When the regulatory framework shifts, the requirements baseline must shift with it, and every downstream artifact — design documents, verification plans, test procedures, hazard analyses — must be traced and updated coherently.
This is not a tool problem, exactly. But it is a systems engineering problem that tools either make tractable or make worse. Programs that manage requirements in documents — Word files, PDFs, spreadsheet-based RTMs — discover that a regulatory framework change triggers a manual update cascade that takes months and introduces errors. Programs that manage requirements in connected, traceable models can propagate a framework change and immediately identify which requirements are affected, which verification activities need to be revisited, and where gaps have opened.
What a Safety Case Actually Looks Like at Scale
Wisk’s Cora safety case is not public in detail, but the broad architecture is knowable from their published materials, FAA working group participation, and industry conference presentations. It is, by any measure, one of the most complex safety engineering artifacts in aviation — and it has been rebuilt multiple times.
An autonomous aircraft safety case must address several layers that piloted aircraft safety cases handle differently or not at all:
Command and control link integrity. If the ground control station is involved in any flight function, the communications link is part of the safety-critical system. Failure modes for that link — latency, loss of signal, spoofing, interference — must be analyzed and mitigated. For Wisk’s fully autonomous operations, the aircraft must be capable of executing a complete safe flight termination or continued mission execution without any ground input. That requires a self-contained decision architecture that must be certified on its own terms.
Perception system reliability. Autonomous aircraft detect and avoid other aircraft and obstacles using sensor fusion — cameras, lidar, radar, ADS-B. The reliability of that perception system must be quantified to support the safety case. This is not like certifying a pitot tube. There is no established failure rate database for machine learning-based perception systems. Wisk has been building that database for a decade.
Software-intensive system assurance. DO-178C governs avionics software certification. Applying it to the scale of software in an autonomous aircraft — where the software is not just managing flight control laws but making operational decisions — requires a rigor and documentation burden that grows nonlinearly with software complexity. The Wisk engineering organization has had to develop internal processes for managing this at a scale that did not exist in aviation five years ago.
Operational hazard analysis at the system-of-systems level. An autonomous air taxi does not operate in isolation. It operates in airspace shared with other aircraft, above ground infrastructure, in weather, in proximity to people. The hazard analysis must model the full operational environment, not just the aircraft itself. Wisk’s New Zealand flight testing — conducted with regulatory permission in an environment with real operational variability — was partly about accumulating the data needed to populate that analysis with real numbers rather than conservative assumptions.
The companies now entering eVTOL certification with piloted aircraft first are, in some cases, making a deliberate choice to defer this complexity. Joby, Archer, and Lilium (before its restructuring) pursued type certification for piloted aircraft as a stepping stone. This is a legitimate strategy. But it means the autonomous safety case remains ahead of them, unsolved. Wisk has been solving it continuously for a decade, under increasing regulatory scrutiny and with increasing data.
What the Engineering Organization Learned
A decade of eVTOL development does something to an engineering organization that cannot be replicated by hiring talent from aerospace primes. It builds scar tissue — institutional knowledge of what goes wrong, when, and why.
Wisk’s engineering organization, by the time of Boeing’s full acquisition, had been through multiple aircraft generations. Each generation was not a product refresh. Each was a rearchitecture driven by what the previous generation revealed about the gap between design intent and flight reality.
Several patterns emerge from watching this organization evolve:
Requirements must be owned, not just documented. In early-stage programs, requirements documents are often written to satisfy a deliverable rather than to drive design. Someone writes the system requirements specification, it gets reviewed and baselined, and then the engineers build what they were already building. Wisk’s iteration cycles produced an organization where requirements are actively maintained artifacts with owners who are accountable for their currency and traceability. When a test result reveals that a requirement is wrong or incomplete, the process for updating it — and propagating that update — is well-defined and fast.
Verification planning cannot trail design. In programs that move fast, verification planning often happens after design is complete, which means it discovers late that certain design choices are not verifiable by any available method. Wisk’s mature program treats verification planning as a co-equal activity with design, not a downstream one.
Safety analysis is not a phase. The safety case is not something you build at the end of development to hand to the FAA. It is a living artifact that is updated as design decisions are made, as test results arrive, and as the operational concept is refined. The engineering cost of maintaining a living safety case is significant. The cost of not maintaining it — and having to reconstruct the argument from scratch when the FAA asks a question — is larger.
Autonomous systems requirements are not software requirements. There is a persistent tendency in technology-company-flavored aerospace programs to treat autonomous system development as primarily a software problem. Wisk’s experience is that the requirements for an autonomous aviation system are systems-level requirements — they involve hardware, software, communications, environment, operations, and human factors in combination. Managing them requires systems engineering discipline, not agile software development discipline alone.
What the Rest of the Industry Is Still Learning
The eVTOL companies that raised capital in 2020 and 2021 at valuations that implied imminent commercial operation are now, in 2026, working through the certification realities that Wisk encountered between 2018 and 2022. Some of them will navigate it. Some will not.
The distinguishing factor is usually not technology. The technology across leading eVTOL programs is more similar than the marketing suggests — distributed electric propulsion, battery management systems, fly-by-wire flight control, some variation of detect-and-avoid. The distinguishing factor is systems engineering maturity: the ability to manage a complex, interconnected set of requirements, hazard analyses, verification plans, and design artifacts across a multi-year program without losing coherence.
This is where tooling matters in ways that are not glamorous but are decisive. Programs that manage their requirements in disconnected documents accumulate technical debt in their certification artifacts the same way software programs accumulate technical debt in their codebases. The debt is invisible until a regulatory review or a design change forces you to reconcile everything, at which point you discover that your safety case is based on requirements that were updated six months ago without the hazard analysis being revisited.
Modern systems engineering platforms — tools like Flow Engineering that treat requirements as nodes in a connected graph rather than lines in a document — make this kind of coherence maintainable at scale. When a requirement changes, the impact propagates visibly through the traceability graph: which downstream requirements are affected, which verification activities reference this requirement, which hazard analyses depend on this assumption. In a program with thousands of requirements and a certification timeline measured in years, that visibility is not a convenience. It is the difference between a safety case that holds together under regulatory scrutiny and one that does not.
Wisk did not build its program with modern AI-native tooling — they predate it. They built the discipline through painful iteration. The companies following them have the option of building the discipline faster, with better tools. Whether they exercise that option is a separate question.
Honest Assessment
Wisk Aero is not guaranteed to win the autonomous air taxi market. Boeing’s backing provides financial durability, but commercial operation remains ahead, and the path from certification to scaled operations introduces its own set of challenges — infrastructure, public acceptance, air traffic management integration — that are distinct from the engineering problems Wisk has spent a decade solving.
What Wisk has that cannot be replicated quickly is demonstrated learning. They know where autonomous aviation safety cases break down. They know what the FAA asks when it is not satisfied with a failure mode analysis. They know what a generation of flight testing reveals about the gap between requirements and reality. They have built an engineering organization that treats that knowledge as an asset and manages it systematically.
The eVTOL industry is full of companies that are smart, well-funded, and technically capable. The ones that reach certification will be the ones that also developed systems engineering rigor early enough to use it. Wisk’s decade is not a cautionary tale about moving slowly. It is evidence of what it actually takes to certify an autonomous aircraft, delivered in the most credible form available: flight hours and regulatory engagement, accumulated over time, with the scars to show for it.
The rest of the industry is still learning what Wisk already knows. The question is whether they have enough time to learn it before the capital runs out.