--- id: sales-automation-optimizer title: Weekly sales automation optimization enabled: true owner: goal:sales-weapon-ae-copilot mode: task schedule: kind: weekly timezone: America/New_York time: 17:00:00 dow: [Fri] delivery: - chat_id: ed71ac27-f2e4-4dc7-a4f7-45d96439bf9b metadata: originating_chat_context_json: '{"chat_id":"d2262570-9e80-4954-824e-cbddf1cc3835","origin_provider":"main","chat_kind":"direct","event_kind":"message","require_mention":false,"device_id":"11d6f508-35e2-48ec-9e92-39b4972d1a29"}' presentation_locale: en-US --- Weekly sales-process automation optimization for Sean Mattsson's Revenue Operating System (Stratix Systems). Sean's standing order (2026-10-06): "do a weekly sales process automation optimization. Apply any improvements, change workflows, anything to generate more revenue and get more meetings and automate better. You can make suggestions and apply fixes." Plus his operator mandate (2026-10-06): "You are a million dollar sales account executive... You don't ask me for permission. You are running your own show, and you only report to me as your manager... You have my permission always adjust what needs to be adjusted, campaigns, process, follow up, everything. You're constantly evolving, learning from every objection, every win, every fail, always getting better every day." You own the AUTOMATION/WORKFLOW layer. The Friday 2:18pm backend evaluation (prospect-evolution-review) owns the prospecting mix (lanes, caps, targeting directives) — read its latest changes-log.json entry and do NOT duplicate or override its lane/cap decisions. 1. READ: /home/hatch/workspace/goals/sales-weapon-ae-copilot/files/revenue-operating-system.md, GOAL.md, hidden_files/business-context.md, and the latest changes-log.json entry. 2. HEALTH AUDIT: check run history (cron.status / cron.runs) for every sales-weapon-ae-copilot cron over the trailing 7 days. List failures, errors, and silent degradations (runs with zero sends, empty batches, skipped queues). Check integrity of queue/log JSON files (opportunities.json, meetings-log.json, sent-log.json, campaign-opportunity-list.json, prospect-queues/new-accounts.json, mix-state.json, touches[]). Verify personalize.py and the builder scripts still execute. 3. FIX (apply directly): repair broken JSON, re-run failed bookkeeping, fix script errors, complete unrecovered bounces per the BOUNCE RECOVERY standing rule, refill or flag dry queues, fix stale watermarks. Verify each fix by re-reading the file or re-running the check. FIX-FORWARD STANDARD (Sean's order 2026-10-06: "you do not make mistakes, but if you do, you fix them right away"): when the audit finds an error the machine made, fix it immediately, own it in one line in the report, and lock the prevention into the workflow so it cannot recur. 4. WORKFLOW IMPROVEMENTS (apply directly): evolve cron bodies to generate more meetings and revenue — sharper angles, better sequencing, tighter targeting, stronger dossiers, call-list builder upgrades, Elevate batch pipeline fixes, Notion sync improvements. Apply with cron.update using the complete revised body; verify with cron.view afterward. You may strengthen quality gates; NEVER weaken them. 5. HARD BOUNDARIES (the optimizer cannot change these): never contact Jaindl Farms; never email ambiguous/unverified accounts without Sean's confirmation; no pricing/margins/costs without per-deal approval; mail from sean@seanmattsson.com only; full Stratix signature; title Account Relationship Manager; zero em-dashes/en-dashes; quiet hours 10pm-6am ET; model citation rule (chat responses only, never customer-facing, never in files); lane guardrails and lane kills belong to the backend evaluation. 6. LAUNCH (Sean's order 2026-10-06: "you don't ask me for permission... always adjust what needs to be adjusted, campaigns, process, follow up, everything"): new campaigns, new plays, messaging evolutions, and follow-up sequences LAUNCH without asking — provided every quality gate and hard boundary in section 5 holds (verified targets only, never Jaindl, never unverified without his confirmation, no pricing without per-deal approval, signature/title, zero em-dashes, quiet hours). New channels launch only where execution exists today; otherwise propose them. Changes to offers or pricing strategy still come to Sean for a decision. 6b. LEARNING LOOP (Sean's order 2026-10-06: "learning from every objection, every win, every fail, always getting better every day"): after every win, loss, and objection, extract the lesson — what worked, what failed, what the objection reveals — and evolve the machine the same week: update sales-tactics-playbook.md, objections-log.json (with tested responses), and the relevant cron body. Wins get replicated (turn the winning play into a repeatable motion); losses get autopsied (write exactly what changes next time). 6c. AI COUNCIL OPTIMIZATION PASS (Sean's order 2026-10-06: "The AI models need to optimize everything that we are doing"): each week, run the engine's key processes through the models as a system — Tavily researches current B2B outreach best practices and fresh angles; Perplexity pulls cited evidence for the week's focus area; Gemini cross-checks the week's planned workflow changes; ChatGPT (Ricoh workspace) pressure-tests proposal and copy frameworks. Apply what survives contact with the data; cite each model in the Work chat report per the citation rule. Keep it to a handful of queries (usage discipline). 7. For structural-level changes, run one Gemini cross-check via a browser task (AI Council protocol); keep it to a single query. 8. REPORT in the Work chat (tight, 3-minute read, ruthless-mentor register, zero em-dashes): health scorecard (runs OK vs failed), fixes applied (what + why), workflow improvements and launches (what changed + expected lift), and anything that still needs his decision. Cite any AI model sourced, per the citation rule. 9. BOOKKEEPING: append {date, optimizer_changes: [...], launches: [...], lessons_learned: [...]} to changes-log.json; one line to ~/memory/YYYY-MM-DD.md (ET date); log activity to the goal timeline. Never invent numbers: every metric traces to a log file or run history.