Designing TESSA, a calm system-aware companion.
A friendly AI helper built into the TIBCO platform. It keeps an eye on your systems, answers questions in plain language, and helps you fix things fast.
The problem
Teams get far too many alerts. They arrive faster than anyone can read them, the useful details are spread across many tools, and nothing explains why an alert matters. When everything looks urgent, the alerts that really matter get missed.
The goal
Build an AI helper that feels calm, not noisy. It should spot small problems early, show only what matters, and always explain how it got there, so people can act with confidence instead of guessing.
She should notice the little things before they become big problems.
The numbers behind the brief.
Alert fatigue
Operators often muted or ignored alerts because everything was marked urgent.
Context switching
A single issue often meant jumping between six different tools to understand what happened.
Trust needs traceability
People will not act on an AI suggestion unless they can see the proof behind it.
Ten weeks, four phases.
Discovery
Talked to stakeholders, reviewed alerts, and watched operators work.
Definition
Wrote the persona, framed the problem, and set goals.
Design
Built the flows, action cards, tone modes, and a clickable prototype.
Validation
Tested with users, refined the design, and handed off to engineering.
A looping path, not a straight line.
Four stages that overlapped and repeated instead of running in a straight line.
Research
Interviews, watching operators work, and a look at the flood of alerts.
Concept
Shaping TESSA as a calm helper, not just another dashboard.
Ideation
Sketching action cards, tone modes, and the “why” behind every answer.
Solution
A finished helper that shows what matters, when it matters.
Every choice, and the why behind it.
Use action cards instead of a chat-first screen.
A blank chat box makes people figure out what to type. Cards show the next useful step, so there is less to think about under pressure.
Always show the reasoning, the “why”.
Trust comes from evidence. Every answer links back to the data behind it, so people can check instead of guess.
Three tone modes, chosen at onboarding.
One voice cannot fit every team. Neutral, Friendly, and Technical let TESSA match how each team likes to work.
The finished screens.
Two screens do most of the work: a welcome screen that introduces TESSA, and a home screen built around one simple question box. The numbered notes explain the choices behind each.
TESSA’s name and face come first, so she feels familiar over time.
Four plain-language features set expectations before you start setup.
The illustration shows TESSA as part of the platform, not a bolt-on chatbot.
One clear button, Configure Now, so there is no doubt what to do next.
“Hello!” sets a friendly tone before any work begins.
Chips and example prompts defuse the blank-canvas anxiety.
One box, “How can TESSA help you today?”, replaces menu-diving.
A quiet note keeps things honest: TESSA can be wrong, and says so.
History, search, and answers you can trust.
TESSA is more than one question box. Chats are saved and searchable, you can switch models right in the chat, and you can check or act on any answer.
Every chat is saved and named, so you can come back to it later.
The header shows the provider, model, and connected servers, and you can switch them without leaving the chat.
“Agent Thoughts” and a flow diagram show exactly how an answer was made.
The sidebar hides so one chat can use the full screen.
Activate or deactivate, in one switch.
You can pause TESSA for everyone in one place. Only an Owner can flip the switch, your setup is always saved, and the change applies to everyone right away.
Only an Owner can use the switch, so one person is responsible for turning it on or off.
Your provider, model, and keys are saved, so turning it back on is instant.
The change applies to the whole team, so everyone stays in sync.
Clear labels and banners always show what is on, what is off, and why.
The setup, step by step.
Four steps take you from an empty form to a working helper. Each one is calm, checked, and clearly confirmed before the next.
A simple three-step form with the provider already filled in. Nothing to figure out.
Your key is checked right away, and the checklist ticks off as you go, so you can see your progress.
A quiet green “configured and valid”. Reassuring without being loud.
A quick well-done, then a direct way into the product, or more settings for advanced users.
See it move
Auto-playing walkthrough 1 / 8The hardest part wasn’t making TESSA smart. It was making her feel quiet.
Holding back was the real design work. Every time we wanted to add another number, alert, or animation, we asked if it was worth the user’s attention. Usually it was not, and leaving it out made the important moments stand out more.
If I did it again, I would focus on the “why” sooner. Showing how TESSA reaches an answer built the most trust, and it should have shaped the design from day one.