The Allsolve Agent is here: 5 ways it can make your life easier

Authors

Juha Riippi
CEO & Co-founder, Quanscient

Think about how much of a simulation engineer's day is actually spent engineering.

You build the geometry. You track down the right materials. You define the physics, set the boundary conditions, mesh the model, launch the run, and when it fails you go digging through logs to find out why. Then you do it all again for the next variation.

None of that is the reason you became an engineer. The part that matters, understanding the problem and finding a better design, gets whatever time is left after the software has taken its cut.

That is backwards. And across the industry, the same idea keeps surfacing to fix it: the AI engineer. A way of working where the mechanical setup is automated, where agents handle the busywork in the background and in parallel, and where the human is there for the part that needs a human, the judgment, the hard problems, the thinking.

We are building toward that, and today, the next step is here.

Allsolve Agent is an AI assistant built directly into Allsolve. You describe the simulation you want in plain language, and it builds it: geometry, materials, physics, meshing, and the solve itself. No code to write, no separate environment to set up, no external LLM subscription. It is already there when you open a project, and it is available to all Allsolve customers today.

The Allsolve Agent on the right as seen in the Quanscient Allsolve UI.

Below, let's look at five ways it makes your day-to-day simulation work easier.

1. Build a simulation just by describing it

Build geometries and run end-to-end multiphysics simulations from plain-language prompts.

Ask it to build a CMUT cell for ultrasound imaging and find its resonant frequency, and it does exactly that. It creates the membrane and the vacuum gap, pulls the right materials from the library, sets up the coupled physics with the bias voltage, meshes the model, runs an eigenmode analysis, and shows you the mode shape.

From an empty project to a solved result in the time it takes to describe what you wanted.

Its also not just filling in a template. It reasons about the whole setup: which physics to couple, where the boundary conditions belong, which analysis type fits the question you asked. And you can dial how much reasoning it puts into a task, quick answers for a simple lookup, deeper thinking for a full build.

2. Catch mistakes before they cost you

Catch model errors and fix failed runs by describing the problem.

An assistant that only builds things is a liability. In engineering, the result has to be trustworthy, and the assistant has to help you get there. So the Agent is built to check its work and yours.

Tell it the resonant frequency does not match what you expect, and it inspects the model, compares it against the design intent, and finds the problem: the membrane that is ten times too thick, the unit that got converted wrong.

When a run fails outright, it reads the solver logs, explains in plain language what went wrong, and offers the fix. It will even read through the customizations you have made to the underlying script and flag what might be causing the issue.

The kind of debugging that used to cost you an afternoon now takes a sentence.

3. Know your result is real and not a mesh artifact

Run mesh convergence studies to confirm a result is physical.

When you need to be sure a result is physical and not an artifact of the mesh, ask the Agent for a convergence study. It identifies the region that actually drives the result, re-runs the model at increasing refinement, and shows you the answer settling down as the mesh gets finer. It lays the runs out side by side so you can see the result converge for yourself. You walk away knowing the answer holds, with the evidence to back it.

4. Turn the docs into your personal tutor

Search the docs and run any tutorial, guided step by step.

The Agent knows our documentation. Ask which tutorials are available and it searches the docs and hands you the direct links. Point it at one and it will replicate the entire workflow, geometry through solve, without you touching a menu.

And if you are still learning Allsolve, ask it to go step by step: it pauses after each stage, explains what it did and why, and lets you ask questions before it moves on. It is a fast way to learn the tool by doing, not just by reading.

5. Let it work while you do something else

Kick off long-running simulations and let the Agent work in the background.

The Agent does not need you watching over its shoulder. By default it asks before it changes anything, so you stay in control. Once you trust it on a task, let it run on auto-accept, switch to another project, and come back to finished results. It does the work in the background and keeps track of the jobs while you focus on the next thing.

It also clears out the repetitive setup that would otherwise eat into your day. Instead of copying a single unit cell a hundred times to build an array, you ask for the array and it lays out the whole grid in one step. You describe the intent. It handles the mechanics.

More ways to use it

These are five ways to put Allsolve Agent to work today, but they are not the only ones. Engineers are already using it for things that did not make this list, and every project it touches is a little different from the last. The more you lean on it, the more uses you find.

It also keeps getting better as we do. The Agent is something we are actively developing, and every step forward adds to what it can take off your hands.

Your data stays yours

For most engineering teams, the first question about any AI feature is about IP. That is the right question to ask.

We keep data residency where it belongs: EU data in the EU, US data in the US. The model providers we use do not retain your data. And we do not use your chat history to improve anything without your explicit agreement. We are also working to bring more of this inside our own cloud over time, so that even fewer third parties are involved. Security and privacy are settled before anything ships to you, not after.

See it live

Allsolve Agent is one step in a direction we are committed to: giving engineers back the part of the job they actually trained for. There is more coming behind it.

The best way to understand what it can do is to watch it work. Join our live launch webinar, where we will build, validate, and run real simulations from plain language in front of you, and answer your questions at the end.

Allsolve Agent is available to all Allsolve customers today. If you are not a customer yet and want to try it, get in touch and we will sort out access.

Register for the webinar here →