02 — Wireframing

From scattered pain to a single shape.

Seven recurring pain points pulled from interviews. One central failure mode they all feed into. Eight lo-fi sketches that tested whether an ambient layer could carry the load.

01Problem framing

Each red node is a verbatim complaint from at least three different interviews. They all converge on one structural failure — async tools fragment ownership, and the standup is a weak patch on top.

02IA & critical flows
Flow 01

Onboarding

A new EM connects Slack + Linear + GitHub in under 90 seconds.

design constraint

OAuth flows are non-blocking — the user can keep exploring while sources index in the background.

Flow 02

Daily check-in

A PM opens Helix, scans the Pulse Map, acks two Risk Lens items, leaves with a draft Async Digest.

design constraint

No state persists between visits. Each session starts fresh — the system's job is to re-orient, not remember.

Flow 03

Incident triage

A senior IC gets a P1 ping, opens Risk Lens, traces the upstream signal, and dispatches a runbook.

design constraint

90-second human-in-loop window before any escalation reaches a channel. Easy to extend, impossible to skip.

03Lo-fi sketches
Sketch 01
Hypothesis

Can a single-pane dashboard replace standup?

→ Failed — became another forgotten tab.

Sketch 02
Hypothesis

Does an inbox-style Risk Lens lower triage cost?

→ Partial — works for ICs, fails for EMs.

Sketch 03
Hypothesis

Can a Pulse Map node graph make momentum legible?

→ Yes — at the cost of mobile parity.

Sketch 04
Hypothesis

Should the Digest auto-send or require ack?

→ Ack — sends without ack felt invasive.

Sketch 05
Hypothesis

Does color-coded risk bias judgment?

→ Yes — moved to single-accent (violet).

Sketch 06
Hypothesis

Will users tolerate read-only? (no ticket actions)

→ Yes — they explicitly preferred it.

Sketch 07
Hypothesis

Can sources be muted per-channel?

→ Yes — became the dampening control.

in-app
/
slack
Sketch 08
Hypothesis

Should the Digest live in-app or in Slack?

→ Both — user toggles per workspace.

04Decision log
  • Surface modelA:Single dashboardB:Ambient layer (no home screen)BHigher engineering cost; harder to demo in 30 seconds.
  • Color semanticsA:Traffic-light (RYG)B:Single accent (violet) + tonal shiftBLost an industry-standard cue; relies on user training.
  • Risk priority surfaceA:Inbox metaphorB:Ranked list with dismissalBLower familiarity; required onboarding copy.
  • Digest deliveryA:Auto-send dailyB:Ack-then-sendBTwo extra clicks per send; some teams asked for auto-mode.
  • Mobile strategyA:Native parityB:Pulse-only mobile companionBNo incident triage on mobile in v1; flagged for v1.1.
  • Ticket actionsA:Two-way write back to LinearB:Read-only, deep-link outBUsers can't close from inside Helix; slows triage by 6s avg.

Every decision here cost something. Writing the cost down up front is the only way I caught myself defending bad choices later.