We are happy to announce that AI-Assisted Development is now available for everyone, in Service Studio beta.



We took the EAP to improve this feature that boosts your productivity when creating flows by providing guidance on each next step. This was made possible with hundreds of iterations and the latest research on machine learning applied to software code from our outsystems.ai team.

What is AI-assisted development all about?

  • You receive inline next step suggestions in any flow, anywhere in the flow – get guidance in web, mobile, BPT, service actions, you name it!

  • Between 1 and 6 suggestions – our AI assistant is able to assess each suggestion’s confidence level. If it’s high, it will show you just that one, and all you need to do is press ENTER to add it.

  • Suggestions specific to your business logic – this feature is able to deeply understand the context of what you’re developing and gives you suggestions specific to your business logic. As an example, it doesn’t tell you you need to call a Server Action and then lets you choose from all possible options. It actually tells you what is the likely action you need, such as a specific CRUD Wrapper to update an entity.

  • Attempts to auto-fill node properties – to make the experience as frictionless as possible and accelerate it, the right properties for the added nodes are inferred. This means that often-times it will completely configure the node allowing to focus on the next step.

If you were at NextStep Denver, you already saw the latest capabilities live and learned about the secret sauce behind it. Otherwise, we’re also doing another presentation at NextStep Amsterdam later this month, so be sure you don’t miss it!

In any case, you can already download the latest Service Studio beta and start getting AI-powered assistance now. We’re planning general availability for November and you'll see improvements coming in every week!

You can also learn how to use AI-Assisted Development in our documentation.

Looking forward to give it a try! Looks brilliantly from what I have seen

At ODC was also demonstrated how AI could assist in designing an entity model. Is there a roadmap where we can see what to expect in the near future?

Excelent! Using in 3, 2, 1... :D

Daniël Kuhlmann wrote:

At ODC was also demonstrated how AI could assist in designing an entity model. Is there a roadmap where we can see what to expect in the near future?

I can tell you that you already could see some of that capability in Workflow Builder (for those who were at NextStep), suggesting the attributes in the Case/Form.

We have more planned, it's coming but can't share more details right now :)

Looking forward to the Workflow Builder, a lot has been shared but on this one didn't come across any information 

Given that the AI implementation is based on machine learning, can we expect it to give (for the time being) better suggestions for traditional web apps than for reactive web apps, given that the Reactive Web App template is just introduced?

Thanks all for your responses. i was having similar questions and your answers were helpful to me.

Daniël Kuhlmann wrote:

Given that the AI implementation is based on machine learning, can we expect it to give (for the time being) better suggestions for traditional web apps than for reactive web apps, given that the Reactive Web App template is just introduced?

On Traditional Web, since its been around for a long time, there are more consolidated patterns that are learned and suggested, which should make the assistant more consistent. Reactive Web will benefit from learnings from Mobile patterns as well, though the community will definitely start adopting new practices with the new app kind. Since we're adding new features and the ML models are retraining with the new data, you'll see it improve over time for the more emerging patterns.

In that sense, do send us feedback to ai@outsystems.com from your experience.

Great news, congrats!

Who's the actual AI-assistant behind this feature? Is Neo's mind being used for this? :-D

Nordin Ahdi wrote:

Great news, congrats!

Who's the actual AI-assistant behind this feature? Is Neo's mind being used for this? :-D

Rumor has it, it's called Maria...

Antonio Alegria wrote:

Daniël Kuhlmann wrote:

Given that the AI implementation is based on machine learning, can we expect it to give (for the time being) better suggestions for traditional web apps than for reactive web apps, given that the Reactive Web App template is just introduced?

On Traditional Web, since its been around for a long time, there are more consolidated patterns that are learned and suggested, which should make the assistant more consistent. Reactive Web will benefit from learnings from Mobile patterns as well, though the community will definitely start adopting new practices with the new app kind. Since we're adding new features and the ML models are retraining with the new data, you'll see it improve over time for the more emerging patterns.

In that sense, do send us feedback to ai@outsystems.com from your experience.

How does OutSystems obtain reactive web app practises from community? 


Antonio Alegria wrote:

Nordin Ahdi wrote:

Great news, congrats!

Who's the actual AI-assistant behind this feature? Is Neo's mind being used for this? :-D

Rumor has it, it's called Maria...

Looking forward to work with it... uh with Maria!


Daniël Kuhlmann wrote:

Antonio Alegria wrote:

Daniël Kuhlmann wrote:

Given that the AI implementation is based on machine learning, can we expect it to give (for the time being) better suggestions for traditional web apps than for reactive web apps, given that the Reactive Web App template is just introduced?

On Traditional Web, since its been around for a long time, there are more consolidated patterns that are learned and suggested, which should make the assistant more consistent. Reactive Web will benefit from learnings from Mobile patterns as well, though the community will definitely start adopting new practices with the new app kind. Since we're adding new features and the ML models are retraining with the new data, you'll see it improve over time for the more emerging patterns.

In that sense, do send us feedback to ai@outsystems.com from your experience.

How does OutSystems obtain reactive web app practises from community? 


I meant that as more and more people start using Reactive Web, some new patterns might naturally emerge and as they emerge they should be learned and suggested.


I just tried this for the first time on some CRUD wrappers and where creating these first took me around 5 minutes of time and lot of mouse movement it now was completed in a moment because the AI guessed exactly what I wanted to do! Impressive :D. 

Great news, congrats! I will try it

It looks awesome! Will definitely try this! 

I'm really getting used to this quick way of developing an app. I do however encounter the following a lot and AI doesn't seem to pick that I want something else suggested, namely "Use Server Action".

I have encountered this several times now and I manually drag the Server Action widget to the flow and connect it with the If widget but this behavior doesn't get added to the suggestions. 

Vincent Koning wrote:

I'm really getting used to this quick way of developing an app. I do however encounter the following a lot and AI doesn't seem to pick that I want something else suggested, namely "Use Server Action".

I have encountered this several times now and I manually drag the Server Action widget to the flow and connect it with the If widget but this behavior doesn't get added to the suggestions. 

Hi Vincent, thank you for reaching out -- one thing we're not doing yet is real-time adaptation of the suggestions. 

I'd like to better understand these particular scenarios you're describing in more detail -- I will reach out directly.