← All work

// Case study 03 · Conversation design · Social impact

A career coach that lives
where the jobseekers already are

TESSA — the Training & Employment Support Services Assistant — is a Facebook Messenger chatbot built by Accenture for Plan International Philippines. It coaches young Filipino jobseekers, particularly young women from marginalized communities, toward employment through . I designed the engine underneath it.

My role
Conversation / CUI DesignerEnd-to-end dialogue architecture
Client
Plan International PHvia Accenture
Platform
Facebook MessengerNo app install required
Year
2018Recognised at WEF 2018
01 · The situation

The users who need guidance most are the ones job portals serve worst.

Traditional job platforms assume a lot: a polished CV, confident English, reliable data access, and the self-assurance to navigate filters and application forms alone. For young Filipinos from marginalized communities — Plan International's audience — those assumptions fail quietly and constantly.

The gap wasn't job listings. It was guidance: someone to ask "what am I even qualified for?" without embarrassment.

Barrier 01
No coach, no map
Career counseling barely exists for this audience. First-generation jobseekers navigate hiring blind.
Barrier 02
Portals assume readiness
Forms, filters, and CV uploads presume a user who already knows how hiring works.
Barrier 03
New apps are a cost
Data is expensive and trust is scarce. Asking this audience to install something new is asking them to leave.
02 · The system move

Don't build a destination. Build a conversation.

The core decision was to meet users inside Facebook Messenger — already installed, already trusted, already free-rated on most Philippine mobile plans. No new app, no new account, no new interface to learn. The interface is literally the one they use to talk to friends.

If the user already knows how to chat, the entire UI problem becomes a conversation-design problem.
The reframe that set the scope

That reframing is where my work lived. I designed the main engine — the conversational logic that drove the whole bot: the intent structure, the decision flows, and the dialogue architecture that turned a user's messy, human answers into a coherent coaching path. Not a scripted FAQ — a structured decision engine that maps a person's situation (skills, availability, goals) to actionable next steps, entirely through natural back-and-forth.

Interaction model · illustrative
Hi! I'm TESSA 👋 I can help you get ready for work. Ano'ng gusto mong gawin — find a job, or build your skills first?
di ko alam eh... wala pa akong experience
That's okay — everyone starts there. Let's find out what you're already good at. Tell me: have you ever helped in a family business, sari-sari store, or taken care of siblings?

Reconstructed to show the engine's behaviour: no dead ends, no shame, every answer routes somewhere useful. Production dialogue belongs to the client.

03 · The engine

Rules for a bot that talks to people with everything at stake.

A conversational UI for a vulnerable audience has almost no margin for confusion. A confusing form annoys a user; a confusing coach convinces them the problem is them. The architecture followed from that.

01
Every input routes somewhere
The decision tree had to absorb real human answers — uncertain, misspelled, off-script — and still move the user forward. "I don't know" is a valid answer with its own branch, not an error state.
02
Coach, don't interrogate
The same data a job portal collects through a form gets gathered through dialogue — a few questions at a time, each one earning the next. The structure is an interview; the experience is a chat with someone on your side.
03
Structure carries the empathy
Warmth isn't a coat of friendly copy on top — it's architectural. Branch design decides whether a user with no work experience feels guided or judged. The kindest thing a system can do is always have a next step.
04
End at action, not information
Every path terminates in something the user can do — a skill to build, a preparation step, a concrete lead — because the goal was employment readiness, not a well-informed dead end.

A visual UI shows all its options at once; a conversation reveals one turn at a time. Every plausible situation has to be mapped, branched, and resolved before the first user ever types "hi."

04 · The outcome

Recognised on the world stage.

TESSA's impact on youth employment was significant enough to be featured at the World Economic Forum in 2018 — a social-impact chatbot from the Philippines, alongside global initiatives. Plan International Philippines and Accenture publicly featured testimonials from young Filipinas who found employment with TESSA's guidance: the outcome the entire engine was designed to route toward.

01 Dialogue architecture map
(reconstructed · client dialogue withheld)

Outcomes limited to publicly documented facts. Detailed usage metrics belong to the client and are not reproduced here.

05 · Reflection

TESSA is the same discipline I use in banking and pharma work, pointed at a different stake: take a user with low context and zero tolerance for friction, and design a system that simply cannot strand them. Enterprise software users are trapped by their jobs; TESSA's users would just leave — which makes conversational architecture for them the more honest test of whether a flow actually works.

The model generating the words matters less than the architecture deciding where the conversation is allowed to go.
Djoaniel Hernandez · System note № 02
See all work
Talk to me about systems work (opens your email app)