Traeger Pilot.
Our recommendation and view on the SOW.

For Jesper, Mikael, and the Ignite team ahead of the Hjalmar working call. A short read with a clear recommendation, a task-level 90-day plan, and a resourcing view that puts Ignite's AI-engineering edge on the page.

TL;DR — the recommendation

1 · Where we are

The pitch meeting ran with Hjalmar (VP Brand), Justin (VP Tech), and Jared (VP Product) in the room. The follow-up from Traeger:

• They liked everything we showed. Explicit interest in how AI can improve user experiences.
• Six examples called out by name (see Section 2).
• Hjalmar's stated priorities, in his order: successful cooking sessions, churn retention, getting people to grill more, making Traeger a true companion, selling more pellets.
• Hjalmar is separately moving a retail-partnership / co-sell conversation (Ace, Costco) with their Head of Sales + IT.

The narrowing is ours, not theirs. Jesper's draft reply to Hjalmar proposes the two-module pilot. The Hjalmar working call is where it gets ratified or reshaped. Hjalmar is championing internally — the brand-and-engagement framing in his feedback is him doing translation work for the rest of the room.

2 · The six examples → pilot map

Each of the six examples Traeger highlighted has a home. Walking in with this map shows we read the email.

3 · Module pick

Both modules are strong pilot candidates. Present them as such. Soft-default to Pellet Depletion if Traeger asks us to choose.

Either path lands a real Phase 1 by Day 90. The non-selected module becomes Phase 2 in the Week 12 readout roadmap.

4 · Module A — Pellet Depletion Prediction

What it is: an AI prediction layer over WiFIRE telemetry that forecasts pellet depletion from cook history, recipe profile, temperature curve, and grill model. Surfaces the reorder prompt at the user's highest-motivation moment — when they're planning a cook — rather than mid-cook when it's too late to convert.

Build scope · 90 days

• Cohort-isolated depletion model on one pellet SKU, one retail partner, one user segment.
• Reorder prompt surface inside the existing Traeger App (extension, not replacement).
• Retailer-channel handoff — delivery this week or pickup nearby in 2 hours.
• Internal dashboard: model accuracy, reorder conversion, retailer-channel split.

Why this lands per evaluator

5 · Module B — Real-Time Cooking Intervention

What it is: a context-aware guidance layer that reads the cook in flight — temperature curve, duration, the user's own past cooks — and surfaces specific, timely nudges. Stall-phase coaching, finish-temperature suggestions, lid-discipline reminders. Grounded in the user's own history, not generic advice. Includes context-aware recipe adaptation as an extension prompt type.

Build scope · 90 days

• Cook-state classifier (stall, finish-approach, abnormal-deviation) on WiFIRE telemetry stream.
• Personalization layer keyed to the user's previous cooks of the same protein/recipe.
• Two-to-three prompt types shipped to one cohort and one recipe family (e.g., brisket).
• Post-cook rating capture + cook-completion measurement.

Why it's the stronger Phase 2 than Phase 1

• Engagement lift is harder to defend as a single number to a CFO who's never seen our work.
• "Ego-safe" coaching is real research-grade UX — it deserves the time we'd squeeze in 12 weeks.
• Inference cost variance is real if any personalization runs through an LLM; we'd want to model the cohort first.
• A wrong nudge mid-cook is worse than no nudge. Validate the data plane via A, then layer B with confidence.

6 · 90-day plan — SOW shape

Three phases, twelve weeks, weekly working demos. The plan below is what we'd commit to in SOW v1 — task-level, not narrative.

Discovery · Weeks 1–3

Pilot Build · Weeks 4–11

Readout · Week 12

• Attach debrief with primary, secondary, tertiary metric results.
• Retailer channel readiness assessment (A) or cook-completion uplift summary (B).
• Churn-cohort signal as cross-cutting deliverable.
• Costed 12-month roadmap: Phase 2 module, the four other highlighted examples, adjacent categories.
• Reference architecture handoff document for Traeger's team.

7 · Team, resourcing & the AI-engineering edge

The pitch told Traeger we build AI-augmented and ship fast. The resourcing here is how we make that visible. Lean team, weighted to AI/ML engineering, working in tight weekly cycles with a working demo every Friday.

8 · Success metrics

Module A — Pellet Depletion

• Primary: reorder conversion rate, cohort vs. control.
• Secondary: model prediction accuracy — days-of-supply forecast vs. actual depletion.
• Tertiary: retailer channel split — pickup vs. ship.
• Cross-cutting: churn-cohort engagement signal across the pilot window.
• Defensible lift target: +15–25% reorder rate vs. control. Floor co-set in Discovery.

Module B — Real-Time Intervention

• Primary: cook completion rate (finish vs. abandon/pull mid-cook) — cohort vs. control.
• Secondary: post-cook rating uplift.
• Tertiary: cook frequency change for cohort over pilot window.
• Cross-cutting: churn-cohort engagement signal — drop-off detection improvement.
• Defensible lift target: measurable uplift on at least two of three. Sharpened in Discovery.

9 · Investment band

Single-module pilot. Parallel-track variant (Module A lead + Module B prototyped in the same window) pushes net to $275K–$315K — offered only if Traeger asks.

10 · Exchange of value

11 · Parallel retail-partnership track

Hjalmar is moving a retail-partnership / co-sell conversation with Traeger's Head of Sales and IT — targets named so far: Ace, Costco. Different stakeholders, different procurement path, different timeline.

It intersects with Module A (which needs one named retailer for the pickup-in-2-hours value prop) but should not be absorbed into pilot scope. Coordinate, don't converge.

• Acknowledge the parallel track explicitly on the Hjalmar call.
• Ask Hjalmar to nominate the retail partner for Module A now or defer to be confirmed when his Sales/IT track matures. Either is workable.
• If deferred: scope Module A Discovery weeks 1–3 retailer-agnostic for the architecture, bind to a named retailer by week 4.
• If nominated: bring the retailer's tech team into Discovery week 2.

Jesper's separate follow-up with Hjalmar next week on the partnership track is its own conversation — commercial/co-sell framing, not product pilot.

12 · Hjalmar working-call game plan

Energy is different from a cold meeting. Traeger has said yes to direction. We're sharpening — they're ratifying. Working-session mode, not pitch mode.

Walk in with

• A customer-facing one-pager — two modules side by side, our default visible, six-example map, 90-day shape, no pricing.
• The plan from Section 6 (Discovery / Build / Readout), restated unchanged.
• Success-metric framework — proposed, not closed.
• A short list of retailer candidates (Home Depot, Lowe's, Ace, Costco, Traeger.com DTC) — for them to react to, not for us to commit.

Drive these decisions

• Module pick. Both equally strong. Soft-recommend Pellet Depletion if the team has locked that pre-call. If Hjalmar pushes for both, hold the line: one module pilot, the other locked as Phase 2 in the Week 12 readout.
• Executive sponsor. Hjalmar is the natural pick. Justin is strong backup (owns the data plane). Avoid no-sponsor-named.
• Data access path. Direct API · read-only mirror · sandboxed export. Determines Week 1 dependencies.
• Retailer — or deferral. Either Hjalmar nominates now or defers to his parallel track. Both workable.
• Churn measurement layer. Confirm we fold it into Discovery as cross-cutting — gives them a Week 12 churn signal without changing pilot scope.

Leave open

• Price.
• Final success-metric thresholds — co-set in Discovery Week 1.
• Project Gravity / CFO framing — wrong audience for this call.

Don't bring up

• "Loud 1-star reviews" framing — reads as a critique of Traeger's existing app team.
• "The Pitmaster App" as a named product. Frame as an extension of the Traeger App that their team ships.
• Smart-hub counter-pitches against Weber. Pocketed.
• Premature commitment claims ("we'll dominate," "huge"). Norrlander allergy to brashness is real.

13 · Risks & trade-offs

• Cohort isolation in WiFIRE. Real data-engineering cost lives here. Discovery Week 1 must close this.
• Retailer pickup integration (Module A). Depends on Traeger's existing retailer API maturity. If absent for the chosen retailer, that's a 6–12 week parallel build we can't absorb. Mikael's read needed on the call.
• Inference cost variance (Module B). Justin will ask for numbers we don't have until we model the cohort. Scope cost ceiling in Discovery before committing to Build.
• Wrong nudge mid-cook (Module B). Tolerance for false positives is far lower than for late reorder alerts. The shadow-mode phase exists for this.
• App-team sign-off. Whichever module we pick, Traeger's app team owns the surface. Engage them in Discovery Week 1, not Week 4.
• Parallel partnership track drift. If we bind Module A to a partnership outcome that hasn't landed, pilot stalls. Mitigation: retailer-agnostic architecture through week 3, bind by week 4.
• "Apps. Agents. AI. On Webflow." positioning needs to read as complement to the AWS-native stack, not competition. Lead every architecture conversation with "alongside, not on top of."
• Project Gravity isn't in the room. The pilot must produce a CFO-ready output so the next room up has something to react to.

14 · Next steps for the Ignite team

1. This week. Internal alignment on (a) module default — keep Pellet Depletion or flip to Real-Time Intervention given Hjalmar's stated priority, (b) the pricing band, (c) who joins the Hjalmar call besides Jesper and Mikael, (d) position on the parallel partnership track — retailer named in the call or deferred.
2. Before the call. Customer-facing one-pager built from this memo: two modules side by side, 90-day plan summary, six-example map, "Already in motion" sidebar acknowledging the parallel track. No pricing. Ready to drop into chat.
3. On the call. Drive the five decisions in Section 12. Notes by Ignite. Jesper same-day recap email confirming locked outcomes.
4. Next week — separately. Jesper's follow-up with Hjalmar on the parallel partnership track. Commercial/co-sell framing, not product pilot. Worth its own internal prep call.
5. Within 48 hours of the call. SOW v1 draft — scope, deliverables, success metrics (co-set), team, timeline, price. Ignite review before it goes to Traeger.
6. Week of SOW send. Pre-brief Hjalmar one more time before it lands. He's defending this internally — we arm him.