AURA Assistants

Using automation to accelerate work has always been around. IDEs, libraries, open source packages, patterns and templates have aided designers and coders since many years. Large language models too have been around for more than 3 decades. So what has changed ?

For the first time, a financially viable, internet scale, trained model has become possible . The trio of computing power, storage and network speeds have reached the tipping point to enable this.

This has changed the fundamental way tech products are going to get built, and there is no going back.

Until now, the game has been for humans to learn programming languages and design concepts well, and apply them smartly. Now, the game is to train assistants who are good and this, and for humans to get really good at using these assistants.

At AURA, as we build and leverage several assistants we observe three key things for this to work:

1. Teaching the assistant the right things to do and imparting your wisdom to it so that it's output is inline with what your process needs

2. Continuously re-evaluating and being aware of the assistant's strengths and limitations. Be patient with it. When it makes mistakes allow it to learn from it and grow

3. Have the right mix of human and machine intelligence. Each of them have to recognize their roles and find the right synergy

4. We do not need generalist assistants, since humans are good at that role. Each assistant must be a specialist who never get tired, never forget, and never miss out on corners in its arena

Introducing our machine friends

This writeup is to give a general introduction to each of our assistants. Over the coming months, we plan to release detailed breakdown about each of the assistants and our key partners in the space.

PO Assistant

(Element: Earth | Phases: Discovery & Prototyping |Divamian Operator: Product Owner)

The key goal of the assistant is to interact with its user, understand and define what needs to be built and why. Once it understands the crux of what needs to be built, it then proceeds to generating the high level functionality & the overall product components: key apps & modules and interactions between them, features within each module prioritised by Divami’s own proprietary labelling and instructions matrix and so on.

The Divamian PO then utilises where the assistant left off, to further work with various stakeholders, understand real world limitations, define the right business priorities etc.

UX Assistant

(Element: Water | Phases: Productionise & Prototyping | Divamian Operator: UX Designer)

This buddy of ours, understand the product specification and generates user flows and user personas. Once the designer aids it and confirms its understanding to be on the right track, it proceeds to generate different Information Architecture possibilities.

The model is trained to leverage several design patterns used to solve complex design problems over the last 15 years of divami journey. Based on this history, along with data triangulation methods, the assistant generates high level wireframes for key screens along with navigations and connections.

The amount of work put in by the Divamian is almost always equal or more than the assistant. The UX expert brings their gut feel and experience, balances the practicality of timelines and budgets, and operates within real world constraints to take the experience to the next level.

The assistant, being tireless by nature, does heavy lifting on other time consuming activities which are automated, making for a perfect partnership

Front end Engg Assistant

(Element: Water | Phases: Productionise & Prototyping | Divamian Operator: Front end Engineer)

Designers at Divami are translated to working front end code using this assistant. Presently the assistant is built to work on top of figma designs. Several steps that the assistant undertakes are as follows:

• Figma design analysis

• Optimising grouping and analysing layers

• Separating out assets and downloading

• Creating a design theme

• Identify individual UI components

• Separating out reusable components

• Finalising design for code generation

• Generating HTML, CSS library and react components

• Invoking automation to create an environment and deploy to sandbox

In each of the step, the front end engineer operating the assistant, brings in his/her expertise and keeps guiding the work being done. This ensures that we get the right output.

The advantages this assistant gives us are tremendous:

1. Great time saver.

2. Reduces the amount of design bugs by more than 80%

3. Allows the front end engineer to concentrate on complex component logics more leading to higher quality code

4. Forces the design work to be very structured, thus bringing in a thinking about product structure and implementation in tandem while the designs are being created itself

5. Maintains a clean and manageable design system without duplicate copies for different screen form factors etc.

There are however several complex scenarios where the assistant cannot help. For e.g. in complex hub and spoke design patterns, involving several widgets with drag and drop etc. we have seen that it is better off only to invoke the assistant in parts and not on the whole. And this is totally understood. Working with the machine's intelligence in the most optimal manner, respecting it's strengths and limitations is what AURA is all about.

Work under the wraps

• UI Assistant (Element: Water | Phases: Productionise & Prototyping | Divamian Operator: UI Designer)

• Project Management Assistant (Element: Water | Phases: Productionise | Divamian Operator: Project Manager)

  
  

A closing note?

We have already seen a lot of impact that these assistants are having on our company. People are starting to embrace the help they get. We are also seeing over-reliance at times, and we help each other fight it back at times too.

The world of design and engineering has changed. From us using tools, its going to be machines using tools with our guidance. And we recognise that we have begun to scratch the surface.

Our roadmap ahead is going to take us into automating parts of application security, deployment cycles, automatic maintanence & bug fixing and much much more.

But some fundamentals will never change. Success is good people, spending time in building great products.