Back to cases
  • Product Discovery
  • UX Research
  • Synthetic Research

Management Platform & Edtech Discovery

How can we create a smoother experience between Performance & Career Management and Corporate Learning? A technical product discovery case using AI-generated synthetic interviews and hypotheses to validate with real people.

8

Interviews

4

Core insights

3

Prioritized HMWs

4

Case answers

01 · Cover

Management Platform & Edtech Discovery

Technical case · Product discovery. "How might we create a more fluid experience between Performance & Career Management and Corporate Learning?"

  • Understand — Briefing, hypotheses and the CSD Matrix
  • Research — 8 interviews with synthetic personas
  • Synthesize — Patterns, insights and journey
  • Propose — HMW, solution hypothesis and real validation

Jessica Fidelis · Product & UX Designer · October 2026

02 · Problem framing

What the case wants to find out and where I started

Case questions

  • Q1 · How do leaders and employees turn a need into action, and how do they perceive whether it generated growth?
  • Q2 · How do HR and L&D take part, and where does the separation between people, processes, information and products create friction?
  • Q3 · Where is the biggest pain point (Management Platform, Edtech, or the handoff), and which persona struggles most to perceive Time to Value?
  • Q4 · What would the first research study with real users be?

Initial hypotheses

  • H1 · There is a gap between learning metrics (completion) and evidence of change at work.
  • H2 · The biggest break is in the handoff between people and processes, not between products.
  • H3 · The pain may lie in the handoff between products, not within each one.
  • H4 · Not every development gap is solved with a course.

How I reached the results

  • CSD Matrix · section 3
  • Plan, AI Research Log, personas and script · section 4
  • Interviews and evidence · section 5
  • Patterns, contradictions, insights and updated matrix · sections 6 and 7
  • Journeys by role · section 8
  • How Might We and Top 3 · section 9
  • Answers to the case questions · sections 10 and 11
  • Solution hypothesis, plan and real validation · sections 12 to 14
  • Synthetic data. Every interview quote comes from AI-generated personas and is a hypothesis to validate with real people.

03 · CSD Matrix

Starting from what we know

Why the CSD matrix

  • It separates what is already known from what is assumed and from what is in doubt.
  • It prioritizes 8 questions (P1 to P8) that guide the script and the analysis.
  • It becomes a comparison point: after the interviews, each item is confirmed, weakened or dropped.
  • Built with the real Edtech and Management Platform team, it already serves as shared direction.
  • Certainties · what the briefing and documentation state
  • Assumptions · hypotheses to test
  • Doubts · what we can't safely assume
  • Prioritization · P1 to P8, what the interviews need to answer

Priorities P1 to P8

  • 01
    P1 · Evidence of development
  • 02
    P2 · Who turns a gap into action
  • 03
    P3 · How the development plan (PDI) is built
  • 04
    P4 · Where HR's responsibility ends
  • 05
    P5 · Whether the need identified in the Management Platform is reused or reconstructed by L&D
  • 06
    P6 · Where the pain is most intense and by which persona
  • 07
    P7 · Feedback between learning and performance
  • 08
    P8 · Where informal actions get recorded
Full CSD Matrix

04 · Research plan, AI Research Log and personas

AI simulation

How the research was designed

Method

  • 8 interviews of ~40 min: 4 roles × 2 companies.
  • Pilot with Paula and André; prompt frozen afterward.
  • Claude vs. ChatGPT test; Claude was chosen.
  • [E] · Stated as lived experience
  • [I] · Inferred by the persona
  • [P] · As defined in the Plan
  • [INV?] · As defined in the Plan

AI Research Log

The AI generated: personas, interviews, raw transcript and support for synthesis. I decided: the 4 × 2 design, the profile variation, the anti-bias rules, the model, and the classification of each quote.

Discarded and limits

  • Numbers and frequencies: excluded as evidence.
  • "No formal record": became a hypothesis.
  • Limits: not market evidence; 2 people per role; no quantitative data.

Companies and personas

  • Horizonte — Large-scale industry, shift work, active integration between Management Platform and Edtech
  • André · HR
  • Beatriz · L&D
  • Paula · Manager
  • Marcos · Employee
  • Nexo — Mid-sized B2B services, hybrid, no active integration
  • Fernanda · HR
  • Diego · L&D
  • Lucas · Manager
  • Aline · Employee

Single script

Product terms only come up after the persona mentions them.

  • Routine
  • Recent episode
  • Who decided the action
  • Where it was recorded
  • What changed, and the signal
Full plan and log

05 · Interviews and evidence

AI simulation

What the interviews brought

8 interviews with synthetic personas, 4 roles × 2 companies.

The biggest pain points, in order

  • 01
    No one sees the effect after the action.
  • 02
    Follow-up depends on the manager's memory.
  • 03
    The need reaches L&D without its origin.
  • 04
    The employee doesn't receive feedback.

Quotes

  • Completing or recording doesn't prove development — "…the record shows intention more than result." — André · HR
  • After the action, whoever organizes it doesn't see the effect — "…I published it, it was completed, and I don't know if it helped." — Beatriz · L&D
  • On the ground, the evidence is observed behavior — "…it's worth more than any record." — Lucas · Manager
  • The need and the action circulate through conversation — "…a loose note…" — Paula · Manager
  • There's no closure with the employee — "…I take it seriously when I see that it leads somewhere." — Marcos · Employee
  • L&D receives the request without its origin — "…already turned into a request…" — Diego · L&D
Evidence table

06 · Patterns and contradictions

AI simulation

What repeated and where the data disagree

9 patterns identified

  • 01
    #1 · Completing or recording doesn't prove development — Frequency: 6 of 8 · Evidence: 5 E · 1 I
  • 02
    #2 · On the ground, the evidence is observed behavior — Frequency: 4 of 8 · Evidence: n/a
  • 03
    ★ #3 · After the action, whoever organizes it doesn't see the effect — Frequency: 6 of 8 · Evidence: 5 E · 1 I
  • 04
    #4 · Need and action circulate through conversation — Frequency: 8 of 8 · Evidence: 5 E · 3 I
  • 05
    #5 · Access to the Management Platform concentrated at cycle moments — Frequency: 4 of 8 · Evidence: n/a
  • 06
    #6 · Development comes more from practice than from the course — Frequency: 6 of 8 · Evidence: 4 E · 2 I
  • 07
    #7 · Agreed-upon actions cool down — Frequency: 5 of 8 · Evidence: 5 I
  • 08
    ★ #8 · There's no closure with the employee — Frequency: 4 of 8 · Evidence: 1 E · 3 I
  • 09
    #9 · L&D receives the request without its origin — Frequency: 2 of 8 · Evidence: n/a
Frequency isn't strength: #9 appears in 2 of 8, but it's one of L&D's biggest pain points. Highlighted (★): the closure patterns (#3 and #8).

7 contradictions

  • 01
    C1 · Integration × real pain: 4 of 8 mention integration
  • 02
    C2 · What counts as evidence: completing, recording or observing
  • 03
    C3 · Visibility × bureaucracy
  • 04
    C4 · HR held accountable for the PDI, but outside the action
  • 05
    C5 · Horizonte × Nexo: contrast in the design
  • 06
    C6 · Autonomy × direction
  • 07
    C7 · Absence of record: real or a prompt artifact? [INV?]
Full patterns

07 · Insights and updated CSD matrix

AI simulation

What I concluded and what changed in the matrix

6 insights and the confidence level of each

  • 01
    ★ I1 · The biggest break is in verifying progress after the action — Confidence: Medium
  • 02
    I2 · The PDI is a record of the cycle, not its driver — Confidence: Low to medium
  • 03
    ★ I3 · The further from the work, the lower the visibility — Confidence: Medium
  • 04
    I4 · Value exists on the ground, but doesn't show up in the tools (business hypothesis, not from the personas) — Confidence: Low
  • 05
    I5 · Development loses to operations due to lack of protected space — Confidence: Medium
  • 06
    I6 · Employees only engage when they see where it leads — Confidence: Low to medium

What the interviews changed in the matrix

  • 01
    Weakened: "The employee has the biggest pain." The most cited pain belongs to L&D and HR.
  • 02
    Reframed: From "biggest pain in the handoff between products" to pain in the after-action, between people.
  • 03
    No evidence: Process maturity influences integration usage; overlap between the Management Platform and the training needs analysis (LNT) (the disconnect appeared, the overlap didn't).
  • 04
    Downgraded: Numbers and frequencies; absence of record.

Open questions for real people

  • 01
    Real rates of PDI creation and completion
  • 02
    Clients with active integration and PDIs with a linked action
  • 03
    Frequency of follow-up, access, feedback and updates
  • 04
    Difficulties using the platforms
  • 05
    Is the absence of a record real or an artifact?
  • 06
    Organizational maturity
  • 07
    Who closes the cycle and how HR sees the aftermath (C4)
  • 08
    New doubts: are the tools missed? What's missing for the value to show up (I4)?
Updated CSD matrix

08 · Journeys by role

AI simulation

Where the cycle breaks and when each role sees value

Journey stages: 1. Perceive the need · 2. Plan or request · 3. Act · 4. Close the cycle · 5. Perceive the value.

  • 01
    Manager
    • 1. Perceive the need (+) positive — Sees the piece that doesn't come back, at work
    • 2. Plan or request (neutral) — Agrees verbally; the PDI doesn't originate the action
    • 3. Act (−) friction — With tight deadlines, the agreement cools down
    • 4. Close the cycle (−) friction [BREAK] — No moment or place to close it
    • 5. Perceive the value (+) positive [SEES VALUE] — Sees the progress by observing; only he/she sees it
  • 02
    Employee
    • 1. Perceive the need (neutral) — Knows what's needed; doesn't know if there's a PDI
    • 2. Plan or request (−) friction — The agreement arrives through conversation
    • 3. Act (+) positive [SEES VALUE] — Learns by doing; the course complements it
    • 4. Close the cycle (−) friction [BREAK] — No one talks about what changed
    • 5. Perceive the value (neutral) — Notices it alone; no one confirms it
  • 03
    L&D
    • 1. Perceive the need (−) friction — Receives the request already turned into a "course"
    • 2. Plan or request (neutral) — Reconstructs the context by asking questions
    • 3. Act (+) positive — Publishes it and sees the completion on the Edtech
    • 4. Close the cycle (−) friction [BREAK] — No feedback: doesn't know if it helped
    • 5. Perceive the value (−) friction [RARE VALUE] — Only a stray comment from a manager
  • 04
    HR
    • 1. Perceive the need (neutral) — The need arrives through the cycle and conversation
    • 2. Plan or request (neutral) — Records it in the PDI and forwards it to L&D
    • 3. Act (neutral) — Follows completion and PDI: proxies
    • 4. Close the cycle (−) friction [BREAK] — Loses track: sees intention, not result
    • 5. Perceive the value (−) friction [VALUE DOESN'T ARRIVE] — Can't demonstrate the effect
The cycle breaks at the same point for everyone (stage 4), but each role sees value at a different moment. The further from the work, the later it appears, or never (I3). Why this format: a single 5-stage sequence per role shows where the cycle breaks (the same for everyone) and when each one perceives value (different); by role, because distance from the work changes this (I3).
View full journeys

09 · How Might We and Top 3 opportunities

AI simulation

From 7 questions to 3 opportunities

Why HMW: it turns the finding into an open question, without embedding the solution. Each HMW yields 3 ideas (WM), across 4 fronts: Product, Process, Data and Service.
  • 01
    HMW 1 · Ensure closure after the action — TOP 1
  • 02
    HMW 2 · Reduce L&D's Time to Value — TOP 3
  • 03
    HMW 3 · HR and managers demonstrating development in practice — TOP 2
  • 04
    HMW 4 · Capture what the manager observes at work — TOP 2
  • 05
    HMW 5 · Make the need reach L&D with its origin — outside the Top 3
  • 06
    HMW 6 · Show the employee where the course leads — TOP 3
  • 07
    HMW 7 · Keep the PDI alive between cycles — TOP 1

Top 3 opportunities

  • Top 1 · Cycle closure — Ensure that every development action gets a closure, sent through a channel the person already uses and consolidated into a weekly summary for the manager, and that an action without feedback shows up as "no feedback," never as "completed."
  • Top 2 · Record practice where it happens — Capture what the manager observes at work and informal actions, via phone or voice, at the end of the shift or meeting, or on a shared tablet, without requiring a terminal or a lengthy form, and define before the action what is expected to change and who will see it.
  • Top 3 · Show the effect of each course — Show, to whoever published it and whoever took the course, what changed afterward: to L&D, per course, the closure feedback cross-referenced with the Edtech's rating; to the employee, the "before and after" with the sentence they wrote themselves.
Close → Capture → Show. Without closure there's nothing to record, and without a record there's no effect to show.
Table of the 21 WM

10 · Case answers · Q1 and Q2

AI simulation

From the need to growth, and who's involved

Q1

How do leaders and employees turn a development need identified in the performance and career context into a concrete development action — and how do they later perceive whether that action truly generated growth?

The need is born at work and becomes action through conversation. Growth is only seen by whoever observes closely, and is almost never confirmed.
  • The need arises from observation and circulates through conversation. — #4 · 8 of 8
  • The action is agreed verbally, the PDI doesn't originate it, and it cools down against day-to-day operations. — #7 · I2 · I5
  • Growth is perceived through observed behavior, only by the manager. — #2
  • The employee notices it alone; no one confirms it. — #8

Q2

How do HR and Learning & Development (L&D) take part in this journey, and where does the separation between people, processes, information and products create friction?

HR records and forwards, L&D publishes and sees the completion. The friction is in the closure, not in the lack of integration.
  • People: HR is held accountable for the PDI but is left out of the action; L&D receives the request without its origin. — C4 · #9
  • Processes: there's no moment or owner to close the cycle. — #3 · #8
  • Information: the record shows intention, not result. — #1 · C2
  • Products: the Edtech reports completion, not effect; the Management Platform is accessed at cycle moments. — #5 · C1

11 · Case answer · Q3

AI simulation

Where the biggest pain is and who feels it most

Biggest pain: Management Platform, Edtech, or the handoff?

  • 01
    Management Platform — need and PDI
  • 02
    Edtech — course and completion
  • 03
    After the action — between people · biggest pain
Not within each product, nor only in the handoff between them: the pain is in the after-action, where no one sees the effect.
  • 01
    It's the most cited pain: L&D and HR don't know if the action helped, and the manager only follows up from memory. — #3 · I1
  • 02
    Integration between the products helps, but doesn't solve what happens afterward. — C1

Persona struggling most with Time to Value

L&D. Receives the request without its origin, publishes it and only sees the completion. It doesn't know if it helped, and the only feedback is a rare, stray comment from a manager. HR comes right after: it decides to invest, but only sees proxies.

When each role perceives value

  • 01
    Manager · Stage 5, close to the work; only he/she sees it.
  • 02
    Employee · Stage 3, applying it; lacking recognition.
  • 03
    L&D · Only by chance, with delay.
  • 04
    HR · It doesn't arrive; sees only proxies.
The further from the work, the later value appears, or never. (I3)

12 · Solution hypothesis

Closing the cycle is the hypothesis to test first

I believe that, by sending the closure through a channel the manager already uses (1a), consolidating pending items into a weekly summary (1b), and recording 'no feedback' instead of 'completed' (1c), more actions will get closure. We'll know it worked when HR and L&D see what changed after the action, and "Completed" stops meaning only "course done."

Case evidence

  • Top 1 · Closure
  • HMW · 1 and 7
  • WM · 1a · 1b · 1c · 7b
  • CSD Matrix · P6 · P1 · P7 · P4 · P3
  • Patterns · #3 · #8
  • Insight · I1
  • Contradiction · C2

Suggested metrics

  • 01
    Closure rate · Actions that receive feedback
  • 02
    "No feedback" rate · Diagnosis, not pressure
  • 03
    Time to closure · Median
  • 04
    Effort · Taps, seconds and abandonment
  • 05
    Visibility for HR and L&D · % of actions and courses with at least 1 piece of feedback
  • 06
    Sustained use · Managers who close 2 weeks in a row without an extra reminder
  • 07
    Business hypothesis · Recurring use and renewal (I4)
  • 08
    Targets only after the baseline.

13 · Action plan and risks

What to align, build and watch out for

Align

  • Managers and employees: channel and moment of closure
  • HR: what counts as closure and the deadline
  • L&D: useful questions per course
  • Client management: sponsorship. Without integration, start with just the Management Platform and a channel

Products

  • Management Platform: success criteria and owner of the closure within the PDI action
  • "No feedback" state and closure in 2–3 taps via link, without login
  • Weekly summary and HR view
  • Edtech: "course completed" trigger. Data: usage events

Proposal to confirm with the product team.

Processes

  • Closure within the manager's routine
  • Who reads the data and acts on it
  • Guide for the client's CS
  • Who sees what, with privacy

Not foreseen

  • Data protection (LGPD): external channels with expiring, single-use links
  • Notification fatigue
  • Simple and accessible copy
  • Response quality and baseline

Risks and caveats

  • Synthetic base · Nothing has been validated with real people
  • Bureaucracy · The closure becomes an empty ritual (C2)
  • Causation · The manager's opinion doesn't prove effect
  • Pressure · "No feedback" read as pressure
  • Integration · Dependence on clients with it active
  • Products · Current capabilities to confirm with an official source
  • Business · The effect on renewal is a hypothesis (I4)

14 · Validation plan with real users · Case answers · Q4

From what I learned to the first real research

Prepare

  • CSD Matrix with the team and technology
  • Initial hypotheses
  • Script
  • Profiles: 4 roles across varied contexts
  • Recruitment and consent via CS, with anonymization

Collect and synthesize

  • 8–10 interviews: L&D and managers (3–4 each), HR and employees (1–2)
  • Update the CSD matrix

Structure and validate

  • HMW and feasibility with leadership and technology
  • Concept test with 5–6 people

Pilot and measure

  • Define the pilot with leadership, CS and clients
  • Track the metrics

What I take from the simulation: the method, the script, the product context and the profile variation. What I adjust: open-ended questions and caution with numbers from memory, which only hold with real data.

Question

"What would be the first research study with real users that I would run tomorrow, and why?"

The first research I'd run would be an adjusted version of the one I did with AI: semi-structured interviews with open-ended questions, starting with L&D and client managers in different contexts. The simulation gave me method, script and context, but it doesn't tell me what real people experience. So I ask what the AI was instructed to avoid (frequency, volume, real metrics), without assuming the pains will be the same. Whatever comes out of it becomes a hypothesis for HMW, concept testing and the pilot.

15 · End

Thank you.

I'm available for any questions. Jessica Fidelis · Product & UX Designer

Full material

Transcripts, research plan, AI Research Log, evidence, patterns, updated CSD matrix and the HMW table.

Open in Figma