Drone & Aerial
Percepto Axion Opens Drone Inspection AI to Any Hardware
Percepto launched Axion on October 27, 2026: a hardware-agnostic platform with 160+ AI models for drone inspection data. What it means for video teams.

TL;DR
Percepto launched Axion, a software platform that analyzes inspection data from third-party drones, sensors and archives with 160+ AI models. Here is what was announced, what is still unproven, and how it differs from searching the raw footage.
On October 27, 2026, Austin-based Percepto launched Axion, which it describes as a hardware-agnostic "Physical AI" platform for operators of mission-critical infrastructure, according to DroneLife and the company's PR Newswire release. The short version of the Percepto Axion news: a company that built its name on its own autonomous drones and docking stations is now selling its inspection AI as software that works on data from other vendors' drones, sensors and old archives.
For anyone who manages drone footage, that matters because the bottleneck in aerial inspection has moved. When regulators open longer offshore routes, such as the Skyways FAA Gulf BVLOS authorization in November 2026, capture range grows faster than review habits. Capturing video is cheap; deciding what is in it, across thousands of flights, is the expensive part. This post explains what Axion is, what Percepto has and has not said, and where searching the footage itself still fits.
What did Percepto announce with Axion?
Percepto announced Axion as a software-as-a-service product that analyzes inspection data from existing or new drone fleets, third-party sensors and other data sources. DroneLife reports the platform is available now and that it follows the company's next-generation inspection software introduction at InnovateEnergy Week in June 2026.
The launch details reported by DroneLife:
- Scale of the AI: more than 160 specialized AI models.
- Training data: more than 100,000 hours of autonomous inspection data collected over more than a decade.
- Organization: data is organized by asset, combining AI output with asset context, historical trends and expert-backed workflows.
- Inputs: supported third-party drone platforms, fixed and mobile sensors, and existing inspection archives.
- Plain-language questions: users can ask about assets and findings in plain language and receive answers and recommendations drawn from platform data.
Chief executive Julien Bellanger, who leads the software-first expansion, is quoted: "Axion is the first step in this next chapter, opening our technology to more operators." He adds: "From day one, a team can bring the inspection data it already has and see which assets need attention first."
Why does a hardware-agnostic inspection platform matter?
It matters because most utilities, pipeline operators and industrial sites already own several kinds of drones and sensors, and vendor-locked analytics forces them to run parallel workflows. When one platform can read all of it, an inspection team can compare an asset across years of flights, regardless of who flew them.
Percepto's own description of the logic is that findings should be judged in context. DroneLife gives two examples from the announcement: identifying a downed wire from its relationship to nearby components, and judging a thermal anomaly against expected asset behavior and environmental conditions. That is a different job from generic object detection. A hot spot on a transformer is only a problem relative to load, weather and history.
The deployment path Percepto describes is gradual. Operators start with data they already have, then optionally add Percepto's Ranger drone portfolio and Harbor autonomous station, plus managed remote operations, for more frequent coverage. The company says it operates under a nationwide FAA beyond visual line of sight (BVLOS) waiver, which is why the hardware option remains part of the pitch.
What is still unproven about Axion?
The performance claims are the company's own, and nothing in the launch coverage includes accuracy figures, false-alarm rates, pricing or a named customer for Axion itself. The "160+ models" and "100,000 hours" numbers describe inputs, not results. A buyer should treat them as a starting point for a pilot, not as evidence of detection quality on their own assets.
Three questions are worth asking any vendor in this category:
- Which third-party platforms are supported today? "Supported third-party drone platforms" is a list, and your fleet is either on it or not.
- What happens to the raw video? Platforms organized around assets and findings may keep imagery as supporting evidence, not as a searchable record of everything the camera saw.
- How are findings validated? Percepto says each finding is backed by inspection history and expert validation. Ask what that costs and how long it takes.
How is asset-centric inspection different from video search?
Asset-centric inspection answers "which asset needs attention," while video search answers "show me every moment where something specific appears." They overlap but are not the same product, and many inspection programs end up needing both.
An asset platform like Axion starts from a structured model of the site: towers, lines, tanks, panels. Findings attach to those assets and trend over time. Video search starts from the footage: you type a description and get the matching moments, with no pre-built asset model. That is useful when the question was not anticipated, for example "every clip where a crane is near the east fence," "the flight where the pilot mentions a bird nest," or a sign number visible on screen.
Our earlier coverage of drone programs shows the same split. In the Evolver DFR launch, the hardware and dispatch software were bundled, and the open question was reviewing the recorded video afterward. In the SkyeBrowse Timelines and roof damage analysis update, drone capture became measurable 3D models and storm-screening reports, but the original footage still needed its own search. And in the NGED transformer inspection work, AI flagged candidates that humans then reviewed. Oil and gas dock programs echo the same gap in FlytBase Octane: asset workflows scale; raw flight video still needs search.
| Question | Asset-centric platform (Axion-style) | Video search (Ceptory-style) |
|---|---|---|
| Starting point | A model of the asset | The recorded footage |
| Typical query | Which assets need attention first? | Show every moment matching this description |
| Strength | Trends per asset over time | Unanticipated questions, speech and on-screen text |
| Needs an asset model | Yes | No |
| Source of video | Supported drones, sensors, archives | Any video you upload or add by URL |
Where does Ceptory fit alongside a platform like Axion?
Ceptory fits as the search layer over recorded video, not as an inspection analytics product. Ceptory indexes video you upload or add by URL through its API, lets teams search it in natural language across scenes, speech and on-screen text, and runs multimodal analysis from the dashboard or REST API. See the drone monitoring feature and the search feature.
The honest caveats:
- Ceptory does not build asset models, trend findings per tower, or plan flights. If you need that, a platform like Axion is the closer fit.
- Ceptory has no integration with Axion. You would export or reference footage yourself.
- We publish no accuracy figures for Ceptory, so run a pilot on your own footage.
- Ceptory sells no drones or cameras and needs no hardware change.
On cost, indexing is $0.042 per minute plus $0.0015 per minute of infrastructure, search is $4 per 1,000 queries, and the first 10 minutes of indexing are free. See pricing. To estimate the review workload you would be replacing, the free video review cost calculator takes hours of footage and reviewer rates.
For the broader picture of searching recorded video in natural language, see enterprise video search in natural language, and for how video platforms differ from camera management systems, video intelligence platforms vs traditional VMS.
What should an inspection team do now?
Treat the launch as a prompt to audit your own data, not as a reason to switch tools this week. A practical sequence:
- Inventory your sources. List every drone model, fixed camera and sensor that produces inspection data, and where the files live.
- Separate the two questions. Write down which of your recurring questions are asset-status questions and which are "find the moment" questions.
- Ask for a pilot on your archive. Axion's pitch is that you can bring existing data on day one. Test it with a past incident you already know the answer to, and note what it missed.
- Ask for the numbers. Request detection results on your asset types, false-alarm rates, pricing and the list of supported third-party platforms in writing.
- Keep the raw video searchable. Whatever platform you pick, retain the original footage somewhere you can query it. A free Ceptory trial covers the first 10 minutes of indexing.
- Plan the human step. Percepto stresses expert validation. Decide who signs off on a finding before it becomes a work order.
Who is likely to benefit first from Axion?
Operators with large, mixed inspection archives stand to benefit first. A utility that has flown transmission lines with two or three different drone types over several years has data that no single vendor's tool has ever read together. Axion's stated promise, that a team can bring existing data and see which assets need attention first, targets exactly that backlog. Percepto also says a leading U.S. energy company awarded it a multi-year transmission line inspection contract covering aerial data collection and AI analytics, though that contract predates Axion and the coverage does not say Axion is what that customer uses.
Smaller teams with a handful of flights per year may find the asset-centric approach heavier than they need. For them, the practical problem is often simpler: finding the right clip from last spring's flight, or checking what a pilot said in the audio. That is a search problem, and it does not require an asset model at all.
Also keep in mind that "physical AI" is a broad label. In the Axion announcement it describes AI that reasons about real-world assets using context and history. It is not a claim that the system controls robots or aircraft on its own, and the launch coverage does not say Axion changes how flights are planned or flown.
What does this say about the market?
It says analytics is decoupling from hardware. Percepto's Ranger drones and Harbor station remain in the lineup, but the headline is software that accepts data from rivals. If more inspection vendors follow, buyers get a freer choice of drones and a harder choice of where intelligence lives. Our view is that the durable asset is the footage itself: keep it in a form you can search in plain language, and any analytics layer, Axion or otherwise, becomes easier to evaluate and to replace.
That is the case for pairing an inspection platform with a video search layer. One tells you which asset to visit; the other lets you ask the footage anything afterward. Learn more at Ceptory. For the regulatory side of drone operations, see the FCC's vote on drones in the 800 MHz band. For how the U.S. military side is organizing counter-UAS testing and clearinghouses, see UND and JIATF-401.
Related reading: Axon's Skydio investment covers how public safety drone channels are consolidating in parallel.
Frequently asked questions
What is Percepto Axion?
Axion is a hardware-agnostic, software-as-a-service platform from Percepto that analyzes infrastructure inspection data. Percepto launched it on October 27, 2026, and says it works with existing or new drone fleets, third-party sensors and other data sources.
How many AI models does Axion use?
Percepto says Axion uses more than 160 specialized AI models, trained on more than 100,000 hours of autonomous inspection data collected over more than a decade. These are the company's own figures, and no independent accuracy results were published with the launch.
Does Axion require Percepto drones?
No. Axion accepts data from supported third-party drone platforms, fixed and mobile sensors and existing inspection archives. Operators can later add Percepto's Ranger drones and Harbor station for more frequent coverage.
Does Axion replace searching inspection video?
Not by itself. Axion organizes findings by asset. Teams that also need to search raw footage for a scene, a spoken remark or visible text across many sources need a video search layer, which is what Ceptory provides.
Does Ceptory integrate with Axion?
No. Ceptory has no integration with Axion. Ceptory indexes recorded video you upload or add by URL through its API, so inspection footage can be searched in natural language independently of any inspection platform.
Ceptory
Search your own footage in plain language. The free tier indexes 60 minutes of video.
Related reading


