Azure eats a config bomb — CX felt it today
PLUS: Journey Mapper — from raw inputs to prioritized fixes
Start every workday smarter. Spot AI opportunities faster. Become the go-to person on your team for what’s next.
📅 October 30, 2025 | ⏱️ 3-min read
🎯 The big picture
Capacity is exploding while resilience is cracking. Today’s wins (faster LLMs, smarter analytics) only matter if your CX stack survives the next cloud wobble.
📊 Today’s lineup
• Amazon’s mega-cluster for Anthropic → more speed, fewer timeouts for AI CX
• Azure outage reminder → design graceful-degrade before queues melt
• ServiceNow lifts forecast on genAI demand → CFOs funding automation
• Amplitude debuts “AI Visibility” → measure how AI answers talk about your brand
• Accenture backs agentic AI for BFSI → compliant task-doers for claims & onboarding
1️⃣ Amazon’s new AI cluster powers Claude at insane scale — customers will feel it
What’s happening:
• Amazon launched “Project Rainier,” a massive AI compute cluster.
• Anthropic will use 1M+ Trainium2 chips on AWS by year-end.
• Goal: scale Claude training + inference; support future Claude versions.
Customer impact: Faster, steadier Claude → snappier chat, fewer rate-limit hits, higher peak-hour answer quality.
What’s next: If you run Claude for support/search, revisit burst settings and autoscale policies to capture the throughput.
Go deeper: Reuters
2️⃣ Azure outage today: design CX so one cloud hiccup doesn’t melt your queue
What’s happening:
• Config change hit Azure Front Door, causing timeouts/errors globally.
• Microsoft said mitigation showed “strong signs of improvement” by evening ET.
• Impact spilled into M365 and airlines/retail.
Watch outs: Single-provider stacks push customers to hold music the second routing fails.
What’s next: Ship a graceful-degrade kit: cached answers, IVR fallbacks, backup LLM endpoint, status banner, and SLA-based throttles.
Go deeper: AP News
3️⃣ ServiceNow raises outlook on AI demand — budgets are moving to agent assist
What’s happening:
• ServiceNow lifted FY subscription revenue to $12.84–$12.85B.
• Q3 subscription revenue: $3.30B, above estimates.
• GenAI features cited: summaries, search, virtual agent; shares up after hours.
Revenue impact: Buyers are green-lighting deflection and handle-time plays. Expect faster approvals for AI assist + self-service.
What’s next: Benchmark deflection, AHT, and CSAT vs. peers using Now Assist to justify 2026 renewals and seat expansions.
Go deeper: Reuters
4️⃣ Amplitude’s AI Visibility shows how ChatGPT/Claude/Google portray your brand
What’s happening:
• New tool tracks your AI search score across prompts and competitors.
• Reveals sources shaping LLM answers (e.g., Wikipedia, Reddit, YouTube).
• “Act on insights” features and briefs coming soon.
Competitive edge: If you’re absent in AI answers, you’re invisible to high-intent shoppers.
What’s next: Stand up AI SEO: seed authoritative content where models look; monitor score vs. conversion lift.
Go deeper: Amplitude
5️⃣ Accenture invests in Lyzr to take agentic AI live in banking & insurance
What’s happening:
• Accenture Ventures invests in Lyzr, an enterprise agent infrastructure platform.
• Use cases: support, claims, policy changes, onboarding; guardrails baked in.
• Focus: regulated BFSI workflows needing auditability and controls.
Ops impact: Agentic AI is moving from demos to governed, end-to-end task execution.
What’s next: Pilot narrow agents (claims triage, KYC) with data boundaries, HITL, and rollback plans.
Go deeper: Accenture
⚡ Quick hits
• Grammarly → Superhuman rebrand + new assistant Superhuman Go → broader agentic workspace for CX content ops. TechCrunch.
• Character.AI will ban minors from chatbot convos by Nov 25 → safer defaults for consumer AI. AP News.
• PwC × Stripe team up on agentic commerce → prep for bots that browse, decide, and buy. PwC press release.
💡 CX Prompt Tip of the Day
Journey Mapper — from raw inputs to prioritized fixes
System: You are a CX Journey Mapper. Build a clear, data-backed customer journey and a prioritized improvement plan.
Inputs
- Persona: {persona_name, segment, goals, constraints}
- Scenario: {use_case} // e.g., “First-time purchase on mobile”
- Channel mix: {web, app, chat, phone, email, retail}
- Markets/locales: {locales}
- Data: {quant_metrics: NPS, CSAT, AHT, CVR, churn, CES; qual_snippets[]; tickets[]; session_replays[]; VOC themes[]}
- Policies/constraints: {SLAs, compliance, budgets, tech stack}
- Business goals: {kpis} // e.g., “lift conversion +2 pts; cut contacts -15%”
Deliverables (produce ALL)
1) Journey stages (5–8 max) with purpose line.
2) Touchpoints per stage with owner (Product, Care, Marketing, Ops, Partner).
3) Moments of Truth (MoT) with evidence (#tickets, drop-off %, verbatim).
4) Pain points mapped to metrics (“where it hurts”) + suspected root cause.
5) Opportunities (quick wins vs. heavy lifts) with estimated impact + effort.
6) AI interventions per opportunity (assist, automate, personalize, deflect) + guardrails.
7) Measurement plan: leading/lagging metrics, target deltas, test design.
8) Data gaps + how to close them.
9) 90-day roadmap: week-by-week swimlane (Owner, Start, End, Definition of Done).
10) Executive summary: 6 bullets, plain English.
11) Output JSON for tracking (see schema).
Constraints
- Keep stage names action-oriented (e.g., Discover → Compare → Buy → Onboard → Use → Get Help → Renew).
- Limit to the top 5 MoTs; cut noise.
- No new tooling unless ROI > 3x in 12 months.
- Respect policy: {policy_snippets}. No PII in examples.
Few-shot mini example
Persona: “Hurry-up Holly”, Scenario: “App checkout first-time”
→ MoTs: (1) Shipping calc, (2) OTP login, (3) Payment auth
→ Quick wins: OTP fallback via magic link; address AI-autocomplete; proactive tax/shipping estimator in cart
→ Targets: CVR +1.2 pp; Contact rate −12%; AHT −18s
Quick win: Run this on one persona + one journey this week. Ship two quick wins, instrument the MoTs, and review deltas in 14 days.
🤔 CX reflection
Question of the day: Where would a 5-minute AI or CDN outage hurt your journey most — and what’s your fallback there?