Commenda — Weaving email threads into a real product

What

worked

The chat window
The more it was used, the less customers called agents until eventually, agent’s personal numbers were not given out at all.
Revealing the process upfront
There were much fewer questions of “what’s going on?” or “where are we stuck?”

A centralised system
Documents and information could be found inside the Commenda platform. Goodbye email threads!

What

worked?

kind of

The layout
It got the job done, but it was hard for customer and agent to understand what tasks unlocked for them and when they were unlocked

Copy being the only way to understand task ownership
Agents and customers needed to read out the copy each time in order to understand who the step was for. This was just subpar UX.

What

didn’t work

Rigidity of the process
We found that some agents were still doing things the manual way when they hit a snag.

The assumption that processes were
step-by-step
The interface assumed that every step in the process took place in a linear fashion when really, some steps could happen in parallel.

Understanding progress
While it helped that the full service was laid out, one could not quickly infer how far along they were.

My Role

As the sole designer on this project, I did the following:

Gathered requirements to design V1 in 3 days
In the next 4 months, collected Hotjar recordings, bug reports, and user feedback.
For V2, looped in engineering to understand technical restrictions and used this to scope out features and improvements for V2. Designed V2 in 2 weeks

Aparna Bala

Designer of things

Saas

Commenda

0 → 1 → 10

Weaving email threads into a real product

How I turned email-based services into a product that Commenda’s customers and the service delivery agents used daily.

V1 in 2 weeks, V2 in a month with 1 Manager, 1 Engineer and 1 Designer (timelines include development)

For years, Commenda’s services operated over email  

Commenda helps businesses set up global entities and remain compliant.In-house service delivery agents exchanged information and documents with customers and got just about anything done this way.The user experience with email was functional, but suboptimal with the potential to seriously affect business. But convincing agents to switch out of their inboxes to a brand new tool wasn’t going to be easy.

But email-first wasn’t going to scale

For Customers:

Lack of visibility over the entire process
Important documents and information scattered across email threads
Sensitive information shared across email

For Agents:

Frequent phone calls from customers to get updates
No zoomed-out view of who was handling which client, or what stage each service was at
Important documents and information scattered across email threads and hard to access for future services with the same customers

“

I’m juggling emails and phone calls all day, but I always get the job done.”

~ the Agent

“

I’m finally expanding to the US. I really hope this pays off.”

~ the customer

Poor service delivery had the potential to hurt business

The incorporation service was the start of Commenda’s “flywheel” business model. It wasn’t incorporations that made money for the business, but the direct and indirect tax package upsells that happened afterward, as these other services were recurring and high ticket in nature.

So a speedy product experience (v1) was designed in just 3 days

And it was fully built in 2 weeks on top of many assumptions. The idea was to pilot just one service and understand if this path could solve some problems.We chose the service “Indian LLP Incorporation”; Commenda’s most popular. It was designed to be stupidly easy to understand and build.

Services were delivered this way for 4 months.

I worked on a bunch of other things during this period, but V2 was always hanging in the back of my mind. Watching people hit bumps because of something you designed is humbling.

Agents came to me with problems.

I also sat with agents and realised that the product was being used (and not used) in ways that could only happen when things didn’t work the way they were intended to.

I watched customer HotJar recordings almost everyday.

Some customers complained too. I noted it all down.

V1 validated the concept, but there needed to be improvements

new & improved

I designed the version our users needed

The back and forth between agents remained, but I improved some key aspects of that experience.

 for EVERYONE

How far along are we and who is the service currently stuck on?

Earlier, tasks were nested inside task groups and it was tricky for one to tell when there was an open task assigned to them. We brought this information to the surface

Where do I find the NIC code?

Users often needed documents and information from one service to help carry out others. Earlier, finding these was tricky- one had to go inside the service > Task group > task to find what they needed. Now, I brought it all to the surface.

Where can I have my random queries resolved?

The all new chat now came with some much needed improvements such as AI rephrasing, individual message replies, pinning, search and more.

 a little something extra for AGENTS

This task wasn’t done correctly. Can I reopen it?

There were times when rules changed and steps had to be redone as a result. To enable this, agents could now reopen steps for themselves and the customer.

We don’t need to do this task anymore. Can I skip past it?

Jira was originally used for this purpose, with agents and admins shuttling from app to app. Building this into the app cut costs for the business and also put these conversations inside the right context space.

I’m having issues with this service/ customer. How do I escalate to admin?

Agents could tag admins in the “internal chat” to escalate cases on time. This was previously done on Jira, and eliminating this tool reduced costs.

Can I control when the service is formally closed?

Earlier, services were auto closed when all tasks were marked complete. And with it, the chat window was also disabled.Agents found that certain steps were added or needed to be redone within a service and this auto-closure didn’t allow for this flexibility.To accommodate this, services were now manually closed by the agent.

The customer starts the first task in the service

Every service starts with the status “Not started” for both agent and customer. In this case, the customer is assigned the first step and will be the one to kick off things.

How           worked:

the pilot

The first step is completed

The status of the service has now changed to “In progress” and the task inside the task group “Company formation details” is now marked complete.The next task opens up now, and it’s for the agent.

On the agent’s side, a task now opens up

In this case, the customer fills a form and the agent now needs to approve of all the details.

This back and forth continues until the service is complete

The service is divided into task groups containing tasks. And new steps unlock as the process goes on. Agent and customer go back and forth until all steps are completed.

Note:In the interest of time, we attached the chat window from the previous design system.

And a few explorations later..

Skip to final screens

Case study written and illustrated by me

Here’s a little
something just
for you!

I’m fishing for new work!

If you’ve made it this far, I’d love to get in touch.Here’s where you can find me:

More case studies:

Browse:

Messages about the project: a post on how the agent app changed work, and a note from the Services Lead

Outcomes

A departure from emails

Incorporations were now handled on the Commenda platform for the 50+ countries serviced at the time.No important information hidden in email threads!

Reduced anxiety in customers

Agents stopped being bombarded by phone calls from anxious customers. Agents no longer needed to provide their personal phone numbers to customers.

4 platforms merged into 1

Shuttling between Slack, Jira, ChatGPT and Commenda to carry out work was dizzying and this problem was finally solved.

Tighter communication

Agents were averaging reply times of over 24 hours. Over the first few months of use, this dipped to just around 8 hours for many agents.

Less waiting between steps

Steps that don’t depend on each other can now run in parallel. This one change removed unnecessary blockers; reducing service delivery times.

A personal outcome from this project

From the Services Lead working with Commenda’s agents

Take a trip down memory lane with me

A video call with teammates across time zones

This is how we discussed page structures.
There was no surface left un-doodled.

More glass windows called for more ideating on windows.

We were a global team working across time zones. This made things extra challenging, but also extra rewarding

A lot of work fuelled by a lot of pretty sunsets.
This is one of the many things I miss!

Service summary

A summary on the right that answered: How far along are we? Where are we in the service? Who is doing this service/ which teammates are assigned to this?

A new “tasks” view

A tunnel visioned view without the context of where it belongs in the service

Task highlighting

An explicit highlight for them task groups contain open tasks for the user to complete, visible even when task groups are collapsed

Surfacing data and docs

Dedicated tabs where documents and information were automatically added as the service proceeded.

New message actions

Replying, pinning and searching individual messages

Better browsing

Increasing the height of the chat window to make long messages more readable

Updated design

New visuals to match the design system