Joseph FulaProduct × Applied AI
A concept brief for Pearson · 2026
A conversation about applied AI / 01

From learner
intelligence to
meaningful
progress.

A product-led perspective on how learner signals, trusted knowledge, and AI assistance can become the right next learning action.

Joseph FulaProduct & UX leader · Applied AI builder
My AI consulting services

Independent concept prepared for a conversation with Pearson's Applied AI team.

The learning loop Concept view
Learnersignal Knowledge& context Nextstep Learningevidence ContextInterpretationFeedback
Not a new chatbot. A more useful learning moment.Explore →
The starting point / Pearson's own research

Adaptation is already making a difference.

Pearson has demonstrated the value of AI-supported practice. The interesting product question is how to make every next step grounded, useful, and measurable.

Pearson-reported research · May 2026
90%
more likely to reach initial topic proficiency with AI-powered adaptive practice, compared with non-adaptive practice.
Analysis of 62,000+ higher-ed students using Pearson Study Prep; a reported association, not an independent causal claim. Same study time.
Pearson research →
The next design question

How does intelligence become a better learning action?

01
Read the learner signalWhat was understood, where they struggled, and how confident the evidence is.
02
Guide with trustworthy knowledgeUse course-aligned content and pedagogical constraints—not a free-form answer generator.
03
Test whether learning improvedLook beyond engagement and answer quality toward mastery, transfer, and independence.

An opportunity to extend existing strengths—not a claim that Pearson lacks these capabilities. See Pearson's responsible AI principles →

A small working illustration / learner experience

One topic. Three learners. Different next steps.

Choose a synthetic learner to see how the same topic can produce a different kind of assistance.

Illustrative product simulation

Synthetic learner profiles and hand-authored instructional steps. No live LLM, adaptive model, student information, or Pearson system is connected.

Learner concept map · linear equations
Maya · reinforce a prerequisite
Signal: difficulty distributing negative coefficients
Offline scenario
Signed numbers85%
Distribution42%
Solving equations63%
Modeling problems38%
Suggested next learning action

Reinforce the distributive property

An uncertain prerequisite can make multi-step equation work harder. Show one focused example before increasing complexity.

What the learner might actually see
Try this: Expand −3(2x − 4). Where do the negative signs go?
Multiply −3 by each term inside the parentheses. Start with −3 × 2x, then consider −3 × (−4).
The values are fictional, not validated proficiency estimates. This illustrates a decision-to-UX handoff—not a production recommender.
How I'd structure the experience / conceptual architecture

The value is in the handoffs.

An applied AI system needs deliberate boundaries between learner signals, recommendations, retrieval, generation, and measurement. Explore the handoffs.

INSPECTING 01 / SIGNAL

Start with reliable learner context.

Separate observed attempts from inferred mastery. Preserve uncertainty before any recommendation.

HANDOFF OUTPUTMinimum learner-signal record with evidence source
WHEN UNCERTAINRequest more evidence; avoid premature proficiency labels
Example pilot scorecard — proposed, not measured
LEARNINGDelayed mastery & transfer
RESPONSE QUALITYSource support & hint corrections
SAFETYAnswer leakage & fallback rate
OPERATIONSp95 latency & cost per action

Proposed design pattern—not Pearson's actual architecture. Learning science validates pedagogy; AI science and engineering own production models, data governance, MLOps, observability, and deployment.

Joseph Fula / applied AI + product delivery

The connection between AI capability and product experience.

I bring 15+ years of product and UX experience together with hands-on AI workflows and rapid prototypes—complementing Pearson's AI science and engineering expertise. Explore my AI consulting work ↗

Independent consulting · AI + product

Joseph Fula

Product strategy · UX design · AI integration

What I doFour complementary capabilities
01 / DISCOVERAI workflow strategy

Map the user journey, business decisions, tool boundaries, and high-value opportunities.

02 / CONNECTLayered AI integration

Combine APIs, retrieval, purpose-built tools, agent handoffs, and human review.

03 / BUILDWorking prototypes

Turn product and AI concepts into testable, usable experiences—not just diagrams.

04 / IMPROVEUX & evaluation

Test clarity, failure paths, accessibility, and adoption; refine what matters.

Explore my AI consulting servicesjfula.com/ai-consulting/
How this could complement Pearson

Bridge the gap from model output to useful learning.

01
Frame a learner-facing use case

Understand the learner, the decision, and what an effective next step looks like.

02
Prototype the actual handoff

Make recommendations, grounded guidance, and interactions concrete for team review.

03
Evaluate with meaningful signals

Pair learner progress and usability with reliability, safety, and operational feedback.

INDEPENDENT BUILD · IMPLEMENTATION EVIDENCEUAX StudioAgent setup, visual workflow design, and export-ready build briefs. A product prototype—not proof of enterprise ML/MLOps.
See project ↗
The conversation I'd welcome

Where could a product-led AI prototype help your team turn an existing capability into a better learning moment?

Services & projects ↗Connect on LinkedIn ↗