A no-list, in the open

What Wingman won’t do.

“Innovation is saying no to a thousand things,” Steve said. Most product pages tell you everything they will do. Here is the other half — the features we considered and deliberately did not build, with the one-line reason. We update this in the same commit we say no.

Things we won’t replace

Real CRMs, real RevOps stacks, real billing tools — they do their jobs better than we ever will, and we are the intelligence layer above them.

  • A "real" CRM (Salesforce / HubSpot / Pipedrive / Dynamics / Attio / Close / Outreach / Zoho)

    We integrate with all eight. Our built-in CRM gap-fill is an on-ramp, not a destination — it graduates cleanly when you outgrow it.

  • CPQ

    Salesforce CPQ exists. Our quotes are for teams without one — multi-entity revenue accounting + complex pricing rules are not in Wingman.

  • Commission accounting (Spiff / CaptivateIQ)

    We PROJECT commission per deal (chip + ladder + QBR) for the rep. The system of record stays in your plan tool.

  • A forecasting product (Clari / Gong Forecast)

    We ship a Clari-shaped adapter and keep the in-meeting + per-deal layer as moat. Don’t out-Clari Clari.

  • A marketing-automation suite

    Hubspot Marketing, Marketo, Pardot. We do not send 100k-record nurture blasts.

  • Full ABM scoring (Demandbase / 6sense)

    6sense ships intent. We integrate; we don’t rebuild.

  • Territory + quota allocation engine

    Xactly / Anaplan are this. We assume territories are decided elsewhere.

  • A sales-enablement learning platform (MindTickle / Highspot)

    We ship battle cards + coaching scorecards inside meetings. We don’t ship a course library.

  • A standalone notetaker bot

    Recall.ai is one Recall.ai. We use them; we don’t reimplement them.

  • A help-desk product (Zendesk / Intercom)

    Our cases surface is gap-fill for teams without one. Don’t double up.

Things we won’t build because they hurt sellers

A few features that "sales tools" routinely ship and that we considered + rejected on adoption + trust grounds.

  • Manager-up surveillance dashboards as the default

    Bernie lens. Rep-data goes to the rep first; manager rollups require opt-in consent. Reps adopt what helps them, reject what manages them.

  • Auto-send AI-written emails on the rep’s behalf

    Every cadence draft sits in `draft → under_review → approved → sent`. No code path auto-sends. The rep is the gate.

  • Hidden buyer-side recording

    Recording consent is a hard gate; the chime plays on start + stop. No "demo-mode" override.

  • Activity-as-attainment dashboards

    We measure outcomes, not call-volume. Activity is a leading indicator; we surface it but never as the primary metric.

  • AI cues that interrupt the rep mid-sentence

    TTL + max-per-minute + dismissal-policy. A bad cue is worse than no cue.

  • "Coach my rep" alerts to the manager during a live call

    Coaching is post-call. Manager-during-a-call is a panic-button, not a feature.

  • Cross-tenant data sharing without opt-in

    Competitor-intel + winning-lines pools are read = write — both ends opt in, neither sees identifying data.

Things we won’t build because they hurt the architecture

A few would-be features whose blast radius isn’t worth the win.

  • A general "lowcode workflow builder"

    Salesforce Flow exists. We ship account-actions as the AI-native alternative — a curated list of one-click moves the operator can grow.

  • A custom-UI builder

    Per-user panel layouts + per-tenant role defaults + the canvas render-spec give us the configurability we need without a drag-and-drop builder.

  • Redis as a source of truth for anything

    Redis is ephemeral only — queues, pub/sub, rate limits, session ticks. Postgres is the source of truth. Forbidden, in CLAUDE.md.

  • A standalone mobile app for managers

    The rep is the user; managers reach the same surfaces over the web. Mobile is for the rep on a call, not for dashboard-scrolling.

  • Push notifications by default

    Quiet hours + severity-gated. We default OFF for non-critical events because rep attention is the scarcest resource we touch.

  • AI that "auto-advances" deal stages

    Stage-policy engine enforces required fields per destination stage. The rep advances; we surface the swing (commission ladder), we don’t move the deal for them.

  • A "magic" auto-discovery of every meeting attendee

    Voice → contact attribution suggests; the rep confirms. Never auto-applies — biometric matching against contact lists without confirmation is a trust violation in regulated industries.

Things we don’t do — yet

Honest about the gap. These are not on this page as "never" — they are on the build plan, with named phases.

  • A native Zoom + Teams meeting bot (without Recall.ai)

    Adapters scaffolded + calendar-aware dispatch + host-aware selection ship. Awaiting Zoom Marketplace + AAD vendor approval — gated on vendor, not on us.

  • Meta Ray-Ban Display HUD cue surface

    Phase 5. The cue stream is already a single event channel; the renderer is the missing piece.

  • Counterfactual cue evaluation

    For every cue we fire, log the alternative + the rep’s response. After 10k cues, a real causal estimator. Currently folklore.

  • The audio layer of the closed-won celebration

    The visual takeover ships; the sound is a synth placeholder until a commissioned chime lands. Synth ≠ shipped.