← All work

Case study 02 · AC3E

◐ In discovery✦ AI

AC3E: from a reporting form to a management intranet

GESTIN, a research center's internal platform, existed to report to ANID, Chile's national research agency. Researchers entered data and got nothing back, while finance, development and technology transfer ran on spreadsheets beside it. I interviewed six groups and reframed the project: from a form redesign into a management intranet, starting with a billing dashboard for the director.

LINE 1LINE 2LINE 3LINE 4LINE 5LINE 6LINE 7AREA AAREA BAREA CNO 'AREA'IN GESTININSIDE THE CENTERIN THE ANID REPORTAC3E · GESTIN · 7 LINES → 3 AREAS
The missing entity. The center works in 7 lines; ANID evaluates 3 areas. The platform had no 'area', so the mapping was done by hand.SHEET 01
Client
AC3E, a Chilean research center
Years
2025 – 2026
Role
UX/UI lead, consultant
Team
Me, with AC3E's direction and development teams
Methods
Interviews, roles and permissions, IA, data viz
7interviews with six user groups, in two weeksDiscovery report
2 → 10user profiles, once interviews showed who really does whatDiscovery report
5tools used beside the platformExcel, Monday, Teams, Redmine, Rex+
7 ≠ 3research lines vs the areas ANID evaluatesDiscovery report
01

Context

A platform built to report upward, not to help the people using it

AC3E is a Chilean research center in electrical and electronic engineering, funded by ANID and by contracts with industry. GESTIN, its internal platform, collects what the center must report to ANID: publications, theses, projects, outreach.

Over time the center grew. Direction now tracks a target for industry billing, operations manages core funding on a financial year that starts in December, development runs projects with monthly costs and income, and technology transfer runs a sales pipeline. None of that lives in GESTIN.

02

Role and team

Discovery lead: research, structure, and the first designs

I lead the UX and UI of the redesign: interviews, the diagnosis, user profiles, the roles and permissions model, and the first designs. Earlier, I redesigned the center's public website, built with AI from my research and specs.

  • Interviews
  • Roles and permissions
  • IA
  • Data viz
  • ✦ AI-assisted
03

Research

Six groups, one platform, very different jobs

Over two weeks I ran 7 interviews with direction; operations, finance and people; development; technology transfer; communications; and researchers at three levels of seniority.

GESTIN had two researcher profiles: lead and associate. Interviews showed three levels with different jobs: associates who only see their own records; line leads who check their whole line every month; and area coordinators who build the story for ANID across two or three lines.

A page of the discovery report, in Spanish: the executive summary, the project strategy and the systemic breaks found in the interviews.
The diagnosis. The executive summary of my discovery report, written after the first round of interviews (in Spanish).FIG. 01
04

Findings

Four breaks in how the center works

F1 · Reciprocity

Entering data for someone else

Researchers fill in what ANID requires, but get nothing back: they can't see their available funds, their productivity status, or export their CV.

F2 · Structure

Seven lines, three areas, no 'area'

The center is organised in 7 research lines; ANID evaluates 3 areas and cross-line work. GESTIN has no area, so coordinators reconcile it in Excel.

F3 · Fragmentation

Work happens in five other tools

Sales prospects in Monday and Teams, billing and costs in separate spreadsheets, purchases and hours in Redmine, holidays in Rex+.

F4 · Data loss

Forms that lose work

No draft autosave, generic errors shown only at the end, and required fields people don't have at hand, like a student's ID number. People give up, or hand the task to someone else.

05

Reframe

Not one redesign: two products

GESTIN only served researchers and communications. So the project became two fronts at once:

Front 1 · Usability and retentionFor researchers and communications: fix the forms that cause errors, lost work and abandonment, and give people something back for what they enter.
Front 2 · ConsolidationFor direction, operations, development and transfer: bring the spreadsheets and side tools into one place, with indicators they can trust.

That meant going from 2 user profiles to 10, each with its own scope, permissions and outputs:

RoleSeesCan changeGets out
Super administratorEverything, plus logsFull control; can act on behalf of othersDatabase, backups, logs
Associate researcherOwn recordsOwn records; holiday requestsPersonal spreadsheet, standard CV
Line leadOwn records + whole lineOwn records; signs off the lineLine spreadsheet, annual report
Area coordinatorOwn records + 2–3 linesOwn records; assigns work to areasCross-line metrics, ANID report
DirectionWhole center: finance, projects, indicatorsRead onlyExecutive reports, billing
Technology transferPipeline, R&D contracts, TRLLeads and transfer recordsSales funnel, ANID dimension 7
DevelopmentFormal projects, finance, hoursBilling, costing, hours, purchasesBilling calendar, team hours
OperationsWhole center: core finance, peopleBudgets, holidays, contractsFinancial balances, payroll
CommunicationsOutreach, events, press, awardsPress, clippings, asset libraryMedia coverage, attendance
Guest auditorDashboards the admin authorizesStrictly read onlyOfficial audit reports

Preliminary roles matrix, after the first round of interviews. To be refined with development and direction.

06

Decisions

Where the design is heading

D1 · Give back

Every form returns something

Researchers see their available budget and deadlines, request holidays, and export a standard CV. Entering data starts to pay off for the person doing it.

D2 · Model

Add 'area' to the data model

Lines roll up into areas, so the ANID view and cross-line work are computed, not rebuilt in Excel every year.

D3 · Time

Two calendars, two filters

The financial year starts in December; scientific indicators follow other calendars. Every view says which year it uses, and you can switch.

D4 · Decisions

One billing view for the director

Projected vs actual, the running total against the target, each area's share, and the pipeline: no more reports requested by email.

These directions are being validated with AC3E's development and direction teams before detailed design.

07

Prototype

Try the director's dashboard

Interactive prototype · made-up data

Director's dashboard: are we billing what we planned?

Shown in AC3E's own design system, with a switch to the portfolio style. Change the year and the area, read the variance summary, and draft a note for the board that you approve before it's added.

Open the prototype →
✦ Built with AII specified the views from the interview findings; Claude built the prototype with made-up areas, projects and amounts. I then had it rebuilt in AC3E's own design system and chart library, so the concept reads as part of their platform.
08

The public site

Before the intranet: a new public site, built with AI

The first phase was the center's public website: new architecture for three audiences (industry, academia and funders), a trilingual site and a documented design system. I did the research, the information architecture, the mood boards and the specs; AI tools did the drafting and the building.

The AC3E home page: applied research for industry, energy and health, with a call to collaborate.
The public site. Home page, desktop.FIG. 03
The AC3E home page on a phone.
Home
The research lines page on a phone, starting with artificial intelligence and predictive analytics.
Research lines
The R&D services page on a phone, with the center's value proposition.
R&D services
The AC3E design system's token page: colours with their names and uses.
The design system. Tokens, components and usage rules, documented so the team can add pages without a designer.FIG. 04
09

Reflection

What I'm learning

✓ What workedInterviewing every group, not only the platform's current users, showed that the real product was bigger than the brief.
↺ What I'd changeThe first round mapped each group's needs; who may see and change what still has to be validated with the developers. Next time I'd bring development into the interviews from the start.