Business Case Executive Summary ES

Telefónica Tech's AgentOps

AI-Driven L1 Support Automation — from a proven foundation to enterprise scale. Specialized agents for cost, observability, security, remediation, and ITSM — serverless architecture, multi-channel, and knowledge that compounds.

5x
Volume Scaling
60%
Capacity Freed
24/7
Autonomous Ops
5+
Specialized Agents

The innovation tax

Skilled engineers spend the majority of their time on repetitive, low-complexity tickets that follow predictable patterns — classifying, verifying, investigating, drafting. This "innovation tax" prevents scaling without proportional headcount growth.

Reactive Model
Teams respond to issues as they arrive, with no capacity for proactive service delivery or strategic customer engagement.
Linear Scaling
Growing the customer base requires proportionally growing headcount — a model that is neither sustainable nor economically viable.
Knowledge Silos
Complex issue resolutions stay locked with individuals rather than becoming reusable, searchable team assets.
Response Limits
Human-dependent support cannot deliver 24/7 near-instantaneous responses at scale without massive operational cost.
Talent Misallocation
Engineers capable of architectural guidance spend their time on routine L1 ticket resolution rather than high-value work.
Scalability Ceiling
Current infrastructure cannot handle 5x ticket volume growth without major investment in headcount and tooling.

From MVPs to Platform

Two independent MVPs — Cost Assistant and Observability Agent — validated with 3 select clients in Q1 2026, achieving 80% faster information retrieval. Now evolving into a unified multi-agent platform.

40–60%
More Volume Handled
~80%
Vision Delivered
Key Design Principles
Human-in-the-loop — all AI responses as reviewable drafts
Domain-specific specialized agents, not one-size-fits-all
Automatic knowledge capture from every resolved ticket
AgentCore Memory (STM + LTM) — agents adapt to each client with near-human contextual understanding, no manual parametrization
Extensible — new agents added without redesign
Validated Scenarios
Cost Analysis "How much did we spend this quarter on EC2?" — instant breakdown by service, account, and region with optimization recommendations.
Active Alarms "What are the active alarms in production?" — real-time alarm status with guided troubleshooting: alarms → metrics → logs → traces.
Log Analysis "Show me error logs from the last hour" — CloudWatch log analysis with pattern detection and root cause identification.

Five specialized AI agents

A supervisor agent triages each query, gathers context, and routes to the right specialist. Each agent has access to native AWS APIs and MCP tools, with cross-account access to multiple AWS accounts.

Supervisor
Supervisor Agent
Analyzes intent, validates completeness, requests clarification if needed, and routes to the right specialist agent.
Observability
Observability Agent
CloudWatch metrics, logs, and alarms. X-Ray distributed traces. Guided troubleshooting flow: alarms → metrics → logs → traces.
Cost
Cost Agent
AWS cost queries, forecasting, anomaly detection, and rightsizing recommendations. Access to Cost Explorer, Compute Optimizer, Budgets, and Pricing APIs.
Security
Security & Compliance Agent
Security auditing, compliance checks, configuration analysis, and vulnerability detection across your AWS accounts. Coming soon.
Remediation
Remediation Agent
Autonomous L1 incident resolution with human-in-the-loop approval. Executes corrective actions on client infrastructure — always with operator confirmation.
ITSM
ITSM Agent
Ticket lifecycle management — create, comment, update, and close incidents. Bidirectional integration with your ITSM platform.

Multi-channel serverless architecture

Access the platform from any channel — web chat, phone call, or ITSM ticket. The same supervisor handles every request, with AgentCore Runtime, Memory, and Gateway providing the serverless foundation.

Channel 1 — Web Chat
React SPA CloudFront AgentCore Runtime Supervisor Specialist Agent AI Response
Channel 2 — Voice (AI Voice Agent)
Phone Call AI Voice Agent Nova Sonic S2S AI Agent MCP → AgentCore Gateway Supervisor
Channel 3 — ITSM (ServiceNow)
ServiceNow API Gateway Lambda Bridge AgentCore Runtime Supervisor Response → Ticket
Amazon Bedrock
Claude Sonnet · Guardrails
AgentCore
Runtime · Memory · Gateway
DynamoDB + KMS
Encrypted · PITR · TTL
CloudWatch
Logs, Metrics, Alarms

From foundation to production

From hands-on experimentation to client validation to enterprise production — a journey driven by curiosity, confirmed by real results.

December 2025
First MVPs
Cost Assistant and Observability Agent built independently after AWS training events. Validated with AgentCore Runtime, Strands SDK, and Bedrock from day one.
Q1 2026
Validation & Strategic Vision
Presented to leadership. Deployed to 3 select clients — 80% faster information retrieval. AWS training on multi-agent orchestration confirms the path. Decision to unify under a single supervisor with ITSM integration.
April – May 2026
Architecture & Foundation
Rebuilding from the ground up — unifying both MVPs under a single supervisor, designing voice flows, and laying the groundwork for plug-and-play agent onboarding.
May – September 2026
Delivery & Build
End-to-end orchestration flow with multi-account AWS support. Channels added incrementally — web chat (live), internal ITSM (connected), AI Voice Agent (live). Next: ServiceNow ITSM bridge.
October 2026
Testing & Go-Live
End-to-end validation across all channels, AI accuracy testing, security review, and production deployment — live for our managed clients as a fully operational service.
Future Opportunities
AWS Support Enterprise Experience Releasing 2026 — agentic diagnostic flows that could reduce engineering effort and shift AI inference costs to AWS.
AWS DevOps Agent Customer-specific context through deployment pipeline integration. Proactive automation triggered by infrastructure events.
Total Duration
6
Months Build
10
Months Total
Oct 2026
Go-Live

Investment breakdown

$53,465/year in AWS infrastructure — fully offset by sandbox credits. Serverless economics mean you pay only for what you use.

Annual cost by service
Infrastructure specs
RegionEurope (Frankfurt) — eu-central-1
Bedrock InferenceCross-Region, On Demand
Tokens1,000 in / 1,000 out per req
S3 Storage1 TB Standard + 500 GB IA
DynamoDB500 GB, On-demand + Reserved
Lambda + FargateServerless compute
Observability800 GB/mo logs + spans
Data Transfer1 TB in + 1 TB out / month

Operational transformation

From reactive to proactive. From knowledge silos to compounding team assets. From linear scaling to exponential capacity.

Metric Current State AgentOps Phase 1 AgentOps Full
L1 Response Time Minutes to hours Seconds (AI analysis) Seconds, 24/7
Support Availability Business hours Business hours (triggered) 24/7/365 autonomous
Volume Capacity Baseline +40–60% 5x baseline
Engineer Time on L1 ~100% Reduced (draft review) ~40% (60% freed)
Knowledge Capture Individual / ad-hoc Bedrock Knowledge Bases Continuous learning KB
Operating Model Reactive Semi-automated Fully proactive
For Support Teams
Analysis in seconds. Focus on complex issues. Knowledge builds automatically. AWS details extracted instantly.
For Customers
Faster resolution. Expert-level analysis. Self-service knowledge library. Consistent quality across all interactions.
For the Business
Scale 5x without proportional cost. AI-powered differentiation. Knowledge compounds over time. Serverless economics.

Automate the routine.
Capture the knowledge.
Unleash the talent.

The foundation is proven. The Q1 2026 enhancements are complete. The production build is underway. The business case is clear.

Dec 2025
Foundation Validated ✓
Q1 2026
Enhancements Completed ✓
Apr 2026
Production Build Starts
Oct 2026
Go-Live

The people making it happen

A joint initiative driven by Altostratus engineering, sponsored by Telefónica Tech leadership, and supported by AWS Partner Technical Account Management.

Altostratus · Drivers
Tech Lead
Maykel Cano Serrano
Maykel Cano Serrano
AWS Cloud Infrastructure Manager
Altostratus · Telefónica Tech
LinkedIn
Marcos Moreno Torres
Marcos Moreno Torres
Senior Cloud Architect
Altostratus · Telefónica Tech
LinkedIn
Ester Nieto
Ester Nieto
Product Manager
Altostratus · Telefónica Tech
LinkedIn
Telefónica Tech & Altostratus · Sponsors
Jordi Costa Mur
Jordi Costa Mur
Head of Hybrid Multicloud & Hyperscalers
Telefónica Tech
LinkedIn
Álvaro Paniagua Martín
Álvaro Paniagua Martín
Chief Operating Officer (COO)
Altostratus - Part of Telefónica Tech
LinkedIn
Amazon Web Services
Oriol Cañellas Tortell
Oriol Cañellas Tortell
AWS Partner Technical Account Manager
Amazon Web Services
LinkedIn

Dive deeper

Pick the format that matches your moment. A quick executive snapshot to share with leadership, or the full business case with architecture, costs, risks, and the complete project plan.