Intranet Digital Assistant

UX + Conversation designer

Digital Assistant for Employees

Employees have more or less the same questions when it comes to internal company tools and services. This assistant's purpose is being that go-to resource.

What's the challenge?

▸ CONTEXT

I joined an ongoing project involving an HR chatbot that lived in the intranet portal and on Microsoft Teams of a tech consulting company.

The bot's main purpose was for employees to have easy access to company tools. It could help them set up their PTO, information on co-workers, bring up knowledge-base articles, etc.

By the time I joined, the research was done. Personality was defined. Most of the skills this bot had were already built and running.

What could I possibly bring to the table?
‍

▸ THE CHALLENGE

How might we assure quality conversations are built for this assistant, no matter who's in charge?

▸ WHO'S WE?

I was part of the design team, along with two other designers. I initiated the standardization of best practices of this bot.

▸ PROJECT DURATION

February 2019 - Jul 2019 (5 months)

Getting familiar

For the first couple of months, I learned about the bot and the way the team worked. The tools were Lucidchart for the flow diagrams, Word templates for the scripts and a dev environment to put the bot to test. I created 4 skills based off of what the team was already doing.

Designer and user feedback- We tested on the environment and the dev team collected the feedback and inputs from the bot in production. We, designers, reviewed comment by comment and added suggestions for improvement.

Lucidchart and Word scripts - We built out diagrams on Lucidchart, an extremely helpful output for the developers. The scripts provided us with more clarity on the conversation's voice, tone, and rhythm.

There were often inconsistencies from script to script, or even from script to flow.

However, we found ourselves writing in three varying styles. Our lack of formal design and writing standards was making us do a disservice to the previous work on bot personality.

▸ PROBLEMS

There was some disconnection between skills, and even phrases within the same skill: some were too friendly, others too robotic, too corporate, etc.
‍
Long and thorough peer reviews for each script and diagram. Instead of splitting the work three ways, we would sometimes end up tripling-down on the same work.
‍‍‍‍‍
Back and forth conversations around the best use of certain components, words and logic for consistency, without any written record.

Standardization process

I realized we needed guidelines and talked to the team about this idea. Then I started combining desktop research, benchmarking other bots and doing trainings in new disciplines to fill in our knowledge gaps.

Personality analysis - I took the defined personality to work on variations of voice and tone to set as example.

Desktop research - Curated around 15 sources of best practices for writing and conversational interaction patterns.

UX Writing Training - I set myself to learn about the UX Writing and conversational design disciplines, by doing online and in-person training.

Ready?

Here's the result

I built a four-pillar guidelines document to make the bot consistent both internally and externally with industry best practices.

Each standard includes examples/do's-don'ts, justifications, and technical specifications if applicable

We were able to load these guidelines on a documentation hub. Everything bot-related was accessible for more company areas to start building their own. This was an existing practice over which we had no control, but at least they had this for guidance.

Outcomes + impact

I left behind a bot that had proper communication and consistency, along with a manual to keep it that way.

THE IMPACT

> The existing skills were iterated based on the guidelines for consistency before moving into new skills.
‍
> Sadly, the project was placed on hold because of budget issues, so our design team was rolled-off.
‍
> On the other hand, if it ever comes back to life, the new team that takes on this bot already has guidelines to get it back on track with best practices in mind.

Takeaways

▸ THINGS WE DID RIGHT

> Setting apart the time to build the guidelines first, and once finished, implement them backwards, practicing on the newest skills.

> Restructuring our repository to be coherent on our mission for consistency. It made our work more efficient.

> Communicating closely with the dev team to document the technical constraints, allowing others to pick up where we left off in our technical learning curve.
‍

▸ THINGS WE LEARNED

> Script-diagram consistency was complicated. It was easy to modify one on the fly for dev purposes and leave the other unchanged. We could have reframed that workflow.

> Have each team member checking after a different aspect of the bot was much more efficient. Should have done that sooner.

More projects

Conversational journaling experience

Team engagement analysis

Digital Assistant for Employees

Ready?

Let's talk