How can AI assistants be used in spatial computing?
In recent months we have seen interesting developments: YouTube has appeared in 53% of Google’s AI Overviews for searches on vitamins and supplements, Axis Robotics has released AXIS, a browser-based data engine with over 200 robotic manipulation activities, and MG Ship has integrated AI-based route optimisation to speed up logistics. All these use cases revolve around a common thread: interacting with spatial data in a natural and intelligent way. In this guide, we will explore how to combine LLM-based assistants with spatial computing to create practical workflows, from prompt engineering to building concrete applications.
Why spatial computing is important today
Spatial computing is not just about augmented reality (AR) or virtual reality (VR). It is any system that considers position, orientation, and geometry as an integral part of the decision-making process. In 2026, we see this in:
- Robotics: robots that learn from thousands of manipulation trajectories (like AXIS).
- Logistics: ships that calculate optimal routes in real time (MG Ship).
- Training and inspection: AI guides that overlay 3D instructions onto the physical world.
- Product design and development: generating 3D prototypes from text descriptions.
What kind of prompt works best for spatial assistants
An effective prompt for a spatial AI assistant must contain:
- Spatial context:coordinates, reference system, scale.
- User goal:desired action (e.g., generate a map, optimise a path, label an object).
- Constraints:available resources, required formats, time frames.
- Output style:interactive 3D, JSON, textual instructions, etc.
A good prompt model is therefore structured as a step-by-step instruction.
Practical example: a warehouse mapping assistant
Imagine you need to create an interactive 3D map of a warehouse for a new robotic picking system. With an AI assistant you can get the result in seconds.
Prompt:
<prompt>Generate an interactive 3D map of the warehouse showing me the exact location of each shelf, the handling routes, and suggest me the optimal route to collect items A, B, and C.</prompt>The assistant can:
- Import existing CAD data or floor plans.
- Place shelves in 3D space.
- Calculate safe routes using a built-in optimisation algorithm (similar to MG Ship but at the warehouse level).
- Expose the model as a glTF model or Unity scene for further testing.
How to leverage AXIS for robotic manipulation training
AXIS offers over 200 robotic manipulation activities and 50,000+ trajectories accessible directly from the browser. You can use an AI assistant to:
- Analyse a specific dataset and generate a custom training sheet.
- Create VR simulations that show the robot how to grasp different objects.
- Generate robotic programming scripts (e.g., ROS) from a textual description.
Example prompt for AXIS:
<prompt>Give me a step-by-step procedure to teach a robotic arm to stack 5 cubes using AXIS data. Include control parameters and a short Python code to import the trajectories.</prompt>Route optimisation with AI: the MG Ship case
MG Ship has added an AI route optimisation module that automatically calculates the most economical and fastest route between multiple ports. The same principle can be applied to internal warehouses or production plants.
Prompt pattern:
<prompt>Optimise the logistics route to deliver cargo C7 from Milan to Paris, avoiding the Panama Canal and taking into account the loading capacity constraints of each ship.</prompt>The assistant returns a detailed itinerary, with cost estimates and a 3D diagram of the route.
Quick checklist for creating spatial workflows with AI assistants
- Define the coordinate system:metres, latitude/longitude, local frame.
- Collect data:CAD, point clouds, robotic datasets (AXIS), route history.
- Write a structured prompt:context + goal + constraints + output style.
- Execute the prompt:use LLM compatible with spatial context (e.g., GPT-4 Turbo with 3D plugin).
- Verify and iterate:compare output with the real world; refine input data or prompt.
Conclusion: why this combination works in 2026
Spatial computing provides the geometric and positional structure, while AI assistants provide language, reasoning, and personalisation. Together, they transform raw data into useful actions, whether it’s moving a robot, optimising a shipment, or training an employee on a 3D environment. Mastering the art of prompting for these spatial scenarios unlocks efficiency, speed, and new creative possibilities in any industry.