Turning a collection of features into a product.
Shipping Product
Between 2017 and 2020, SpeakBuddy changed from a collection of speaking exercises into a structured learning product: clearer goals, scenario-based lessons, a daily curriculum and increasingly responsive conversation. The work was less one redesign than a series of shipped changes that progressively changed what the app was.
- Role
- Principal Designer
- Context
- Product Redesign, Daily Curriculum, Dialogue
- Scope
- Product Structure, Interaction Design, Delivery
- Years
- 2017 — 2020
A redesign can change more than how a product looks.
The early problem was structural. SpeakBuddy had lessons, characters and speaking exercises, but the learner still had to work out what to do and how those pieces connected.
Over several releases, that changed: lessons gained clearer goals, conversations became more contextual, daily practice became a sequence, and speaking started moving from choosing answers toward forming your own.
One curriculum — A fixed amount of learning, sequenced each day
Zero blank screens — Every lesson opens with a goal and a scenario
Product Redesign (2017 — 2018)
From separate exercises to a learning experience with a point of view.
The redesign was not primarily visual. Content and navigation were consolidated while the lesson itself became more intentional. By the 2018 v5 release, scenarios could involve multiple characters, each lesson had a learning goal and key expression, and the old correct-or-incorrect answer selection was being removed.
- Learning surfaces consolidated into a clearer product structure
- Goals, key expressions and scenario context built into lessons
- Moved away from selecting answers toward producing them
Daily Curriculum (2018 — 2020)
Turn the app from a library into something you could finish today.
The next step was deciding what the learner should actually do when they opened the app. Daily Curriculum organised practice around level and purpose, grouped lessons into a sequence and gave each day a finite amount of work. Instead of choosing from everything SpeakBuddy could teach, the learner could simply continue.
- A fixed amount of learning presented each day
- Lessons organised into units with review and assessment
- Level and learning goals used to shape what appeared next
Free Chat (2019 — 2020)
Before LLMs, open conversation was a dialogue-flow problem.
Free Chat pushed the product beyond scripted exchanges. Learners formed their own responses, the system interpreted what they said and the character reacted while the conversation continued toward a designed goal. The technology was constrained, but the design problem was already the one we would return to years later with generative AI: how do you keep an open conversation useful, understandable and moving forward?
- Learners constructed responses instead of choosing a prepared answer
- Character responses changed according to what the learner said
- Dialogue remained directed toward a conversational objective
Great app that does exactly what I need, when I need it. — App Store review
Give the learner freedom without giving them the product design problem.
Across the redesign, the same tension kept appearing: reduce unnecessary decisions while increasing what the learner could actually say and do.
- Does the learner know what to do next?
- Can the lesson make its purpose clear before asking for effort?
- Can we remove navigation choices without removing useful control?
- How open can the conversation become before the experience loses direction?
The learning loop became the product.
By 2020, SpeakBuddy had a recognisable structure: understand the goal, practise inside a scenario, speak, receive a response, progress through a finite curriculum and return the next day.
The technology would change substantially after this — AI feedback, personalisation and eventually LLM conversation — but the core product questions were already established here.
- Consolidate
- Set the goal
- Sequence the practice
- Open the dialogue
- Ship
- Learn from use