✛ ✛
[ AI · COACH ]TALK TO KPLUS

TALK TO
KPLUS

[ SMARTER ]

Your AI running coach that learns your pace, your goals, and your recovery — then tells you exactly what to run next.

[ ANYWHERE ]

Just talk. Ask anything about training, races or form, anytime, anywhere — KPLUS answers in your ear like a real coach beside you.

ROLE

Concept · System & UX · Art Direction

SCOPE

Voice-first AI running coach

TYPE

Self-initiated concept

YEAR

2026

A coach you talk to is only worth building if it actually knows you.

01The problem

Most AI coaches today are a chat window in front of a general-purpose model. Ask one what to run tomorrow and it will answer warmly and confidently — without having seen any of your actual runs.

The interface feels new, but the advice is closer to the average of the internet than to anything about you. For someone training toward a specific time, that general guidance only goes so far.

GENERIC MODEL
“What should I run tomorrow?”
“An easy recovery run sounds good — maybe 30–40 minutes at a comfortable pace.”
GROUNDED IN YOUR DATA
“What should I run tomorrow?”
“Keep it easy — around 6 km at 5:40. Your last three sessions were all hard and inside 48 hours.”
pace 5:40weekly load · highrecovery · low

Illustrative — the same question, with and without your data behind it.

02The thesis

So the part worth designing isn't really the chat — it's the system behind it. A coach is only as good as the data it can draw on. Many tools stop at the conversation layer; the more meaningful work is the layer underneath — your own training history, gathered, organised and made queryable, so each answer is rooted in what you actually did.

A

Own the data

Your runs kept on a stack you control, rather than inside a service you can't see into.

B

Retrieve, not guess

Bring in the exact part of your history each question needs, instead of answering from generalities.

C

Answer what matters

What to run, whether you've recovered, what pace to aim for — drawn from your own record.

03How it works

In practice that's a small pipeline rather than a single prompt. Sessions arrive from the watch and the KPLUS app, are tidied into one place you own, and the model only responds once it has pulled the part of your history the question calls for.

INGESTwatch · KPLUS app
→
STOREone place you own
→
RETRIEVEthe runs this question needs
→
GROUNDanswer uses your numbers
→
SPEAKvoice, in your ear
Signal it readsWhat it tells the coach
Pace & splitsEffort, target paces, and where you faded
Distance & volumeWeekly load and how fast it's ramping
Heart rateIntensity zones and drift within a run
HRV & resting HRReadiness and accumulated fatigue
Sleep & recoveryWhether to push or hold back today
Elevation & weatherThe context that explains a “slow” run

The retrieval step is where the value sits — it's the difference between a general “take it easy” and a specific “easy 6 km at 5:40, because of how your week has gone.”

04What the data unlocks

And the data is worth far more than a single answer. Once your history lives in one place, the same foundation supports a whole range of things a general model simply can't do:

Knows your baseline

Your normal paces, distances and effort — so “easy” means easy for you.

Sees the trend

Spots a plateau, a steady build, or the early signs of overtraining across weeks.

Predicts the day

Turns months of training into a realistic race-pace estimate.

Remembers everything

Every run stays in context; nothing resets between conversations.

Reads the context

Sleep, recovery, weather and terrain shape the advice — not pace alone.

Compounds over time

Every logged run makes the next answer a little sharper.

05Why voice

The moments a coach is useful — mid-run, just after, or lacing up early — tend to be the moments your hands are busy. Voice takes away the logging and the typing; you simply ask, the way you might ask someone running beside you.

It also suits KPLUS, a brand built around the helmet and the ride. The coach can live in your ear, rather than behind one more screen.

06The direction

The visual language is intentionally the counterpoint to the KPLUS brand site. That one is cinematic, dark and slow; this one is electric-blue, monospaced and brutalist, with a pixel-dither that dissolves the interface into pure signal.

One loud, one calm — together they show the brand can move between two registers and still feel like itself.

07Why it matters

This is also where KPLUS is well placed. It already owns the hardware and the sessions — the very data most coaching apps have to gather from scratch. A loop where the data and the coach belong together is hard to replicate: the more you run with it, the better it understands you, and the more it's worth coming back to.

That's the real idea here — not “adding AI”, but turning data the brand already holds into a coach only KPLUS could build.

CapabilityTalk to KPLUSTypical AI coachTraining-log app
Grounded in your data✓✗you read it
Remembers your history✓✗✓
Spots trends & fatigue✓✗✗
Voice-first, hands-free✓~✗
You own the data✓✗~

Illustrative comparison — where a data-owned, voice-first coach would sit against the common alternatives.

Concept — a self-initiated systems-and-design exploration, not a shipped product. There are no live users or data; the comparisons and charts throughout are illustrative, shown to make the thinking visible.