AeroInsite

Auckland, New Zealand

A new perspective.Better decisions.

AeroInsite is an early-stage venture exploring how drones and AI can help organisations understand the places and assets they are responsible for — and see what has changed since they last looked.

Explore our approach The venture
Plan view of an aerial survey A survey area outlined above a road corridor and terrain contours. Flight lines sweep back and forth across the area, a drone sits on one pass with its image footprint marked beneath it, and three building outlines are flagged as features of interest.
Early-stage venture Drones, aerial data and applied AI Choosing our first problem

Most organisations are responsible for more ground than they can regularly go and look at.

  • Aerial observation
  • Applied AI
  • Decision support

The opportunity

Seeing a place is not the same as understanding it.

A roading engineer, a facilities manager and a land owner have the same problem in different forms. The thing they are accountable for is large, outdoors and slowly changing, and the usual way to know its condition is to send somebody to look. Looking is expensive, so it happens rarely, so problems are found late.

A drone can cover that ground quickly, repeatedly, and from angles nobody can reach on foot. What it produces is imagery — and imagery on its own is not an answer. The valuable step is the one after it: turning a picture of a place into a description of it, comparing that description with the last one, and putting the difference in front of the person who has to decide something.

Observe, describe, compare, decide. That sequence is what we intend to build a business around. Where it is worth the most, and to whom, is what we are working out now.

We have not chosen our first application. Choosing it on customer evidence rather than on assumption is the work directly in front of us.

From imagery to a decision Three stacked plan-view layers. The lowest is a grid of captured imagery. The middle traces the outlines and points found in it. The top marks one item as changed and carries it out to a decision.
The step that matters. Imagery is the input, not the product. The intended product is a described, comparable view of a place that a person can act on. This diagram shows our proposed model, not a working system.

Potential applications

Three directions worth testing first.

These are areas we are exploring, not services we sell. Each one starts with a question somebody would have to answer anyway, and an idea of what a useful answer would look like. We expect customer conversations to change this list.

  1. 01

    A bridge span with observation points along it

    Infrastructure and built assets

    Bridges, roofs, retaining walls, poles, tanks — long-lived structures that are inspected on a cycle rather than when something actually changes.

    The question
    Which of these needs attention before the next inspection round, and why?
    The kind of output we are considering
    A ranked short list of assets with the imagery that justifies each entry, and a record of what is different from last time.
    What we would need to learn first
    How inspection decisions are actually made and funded today, what evidence would be accepted, and what accuracy is genuinely required.

  2. 02

    Four land parcels with one marked as changed

    Cities, land and development

    Sites, corridors and open space that change continuously while the official record of them updates slowly.

    The question
    What has actually changed here since we last looked, and does it matter to us?
    The kind of output we are considering
    A dated comparison of a site or corridor — surfaces, structures, vegetation, access — with the differences marked and easy to review.
    What we would need to learn first
    Which changes people care enough about to pay to detect, how often, and how such a comparison would sit alongside the records they already keep.

  3. 03

    A shoreline with contour bands and sample points

    Environmental change

    Vegetation, erosion, waterway margins and restoration work — slow change across areas too large to walk and too varied to sample well from the ground.

    The question
    Is this getting better or worse, where exactly, and how confident can we be?
    The kind of output we are considering
    A repeatable description of extent and condition that can be flown again later and compared honestly, including where the method is weak.
    What we would need to learn first
    Who funds this monitoring, what standard of evidence they must meet, and whether aerial capture can meet it at a cost that works.

The proposed approach

How we intend to work.

Four steps, in order, with a person accountable at each one. This is the model we intend to test — not a description of a system in operation.

01

Understand the question

Start from a decision somebody already has to make — a budget, an inspection, a consent, a response. If aerial data would not change that decision, we should say so rather than fly.

02

Obtain suitable aerial data

Choose the flight, the sensor, the resolution and the conditions that would actually answer the question, within the rules that apply to the airspace and the site. Plan for the flight to be repeatable, because comparison is the point.

03

Analyse and review

Use AI to find and describe what matters in the imagery — outlines, surfaces, conditions, differences from last time — and have a person check the result before anyone relies on it.

04

Communicate useful findings

Deliver something a busy person can act on: what changed, where, how certain we are, what we would suggest doing, and what we could not determine.

Proposed sequence Steps 01 – 04
The proposed sequence over one site One plan view built up in four layers: the area of interest and the decision point; the flight lines and image footprints; the outlines and points found by analysis with a review check; and a short summary of findings passed onward.
The proposed sequence for a single site, shown in full. Nothing here runs a model or reports a measured result.

Building AeroInsite

What we are doing next.

Planned work, in the order we intend to do it. None of it is finished, and we would rather revise it against real customer evidence than defend it.

  1. STEP 01

    Choose the initial problem

    Talk to the people who would have to buy this and the people who would have to use it, and pick the one problem where an aerial answer is clearly better than the alternative.

  2. STEP 02

    Understand the requirement

    Establish what evidence, accuracy, frequency and format would make the result genuinely useful — and what operating rules and permissions apply to getting it.

  3. STEP 03

    Scope a pilot

    Design one small, lawful, real piece of work with a written scope, agreed data rights and acceptance criteria set before we start rather than after.

  4. STEP 04

    Evaluate it honestly

    Did it change a decision? What did it cost to deliver? Would they pay for it again? Report the limitations as clearly as the results.

AeroInsite is at the beginning. We have the idea, a drone and the intention to test this properly. Everything above is planned work, described as such.

An invitation

If this is your problem, we would like to hear about it.

The most useful thing anyone can give us right now is an honest description of a decision they struggle to make well. We are looking for those conversations before we commit to a product.

AeroInsite

Read how we are thinking about the venture, what we intend to validate next, and the kinds of conversations that would help shape it.

The venture

You have the problem

You manage assets, land or an environment you cannot inspect as often as you should. Tell us what you actually need to know, and how the decision gets made today.

You have built this before

You have worked in aerial survey, geospatial analysis, computer vision or drone operations, and you know where these projects usually come unstuck.

You invest early

You back ventures at the idea stage and want to understand the problem, the plan and the evidence we intend to gather before committing.