AI Act and the future of communications: challenges for VoIP and contact centers
The AI Act is changing the game for telecom operators. How to prepare VoIP and contact center data and architecture for new…
VoIP, contact centers, BSS/OSS, MNP routing. For tier-1 operators and enterprises with their own telecom infrastructure.
The telecom practice covers design and implementation of systems for communications operators and enterprises with their own telecom infrastructure. Includes VoIP platforms (SIP routing, MNP, billing), contact centers (IVR, ACD, omnichannel), BSS/OSS, integration with government registries.
In 2026 the focus is AI Act preparation in telecom (especially for contact centers with AI agents), legacy OSS modernization, transition to cloud-native architecture.
AI Act classifies systems making customer service decisions as high-risk. This means documentation, audit trail, human oversight, regular testing. Preparation takes 6–9 months — start now.
If launching a new tariff takes 6 weeks due to legacy OSS, you lag behind competitors who can launch in 5 days. OSS modernization is a strategic priority, not tech-debt in the backlog.
An operator used to voice channel needs 3–6 months of retraining to work effectively with chat/email/social. The technical part is just 30% of the real project.
It is more honest to say "you do not need this yet" than to sell an engagement that will not deliver ROI.
Audit of SIP infrastructure, BSS/OSS, contact center, integrations. Gap analysis against state-of-the-art and regulatory requirements.
Target architecture design, legacy migration plan, AI strategy for contact center with compliance-by-design.
Pilot launch on one segment (e.g., new MNP router or omnichannel in one department) with full monitoring.
Phased expansion to full traffic/operator volume with parallel run of legacy and new system. Continuous KPI monitoring.
SLA support, AI model optimization, regular compliance reviews, capacity planning.
Project lead: DooxSwitch (VoIP platforms, SIP routing, MNP) and Softengi (AI agents for contact centers).
Brought in when needed: Softline (integration with government systems and registries), SL Global Service (cloud-native transformation).
They deploy an AI agent in the contact center to reduce AHT. Six months later — regulator inspection finds problems with audit trail and human oversight. Fine + forced decommissioning.
What we do instead: compliance-by-design from the start: audit trail for every AI decision, human-in-the-loop for escalation, regular bias testing.
Attempt to replace the entire OSS in one project. Plan duration 12 months, actual 24. Old OSS breaks before new one stabilizes. Customer loss due to billing errors.
What we do instead: parallel run for at least 6 months, phased migration by customer segments.
They think omnichannel just means giving voice operators access to chat and email. Three months later — chat response time 8 hours, NPS drops.
What we do instead: for omnichannel — separate training program 3–6 months, operational model redesign, new KPIs.
No precise savings percentages — actual numbers depend on the client's starting point. Instead — concrete architectural decisions and organizational changes.
The DooxSwitch team builds MNP routing for two tier-1 operators. Migration from legacy — 8 months with parallel run. Failed call rate reduced by 40%.
AI agent handles typical queries (balance, transaction status, card block). Escalation to human — for complex cases. Compliance-by-design for AI Act: full audit trail, human review for escalation cases.
Previously launching a new tariff took 6 weeks due to legacy BSS. Cloud-native BSS with API-first architecture. Now new tariff — 5 days from decision to launch.
Nine recent expert articles — from thematic overviews to specific architectural decisions.
DooxSwitch Platform · FreeSWITCH · Asterisk · Kamailio · OpenSIPS · Cisco UCM
Genesys · Five9 · NICE inContact · Amazon Connect · Cisco Webex Contact Center
OpenAI · Anthropic Claude · Microsoft Copilot · custom LLM agents · sentiment analysis
Amdocs · Comarch · Oracle BRM · Ericsson Charging · MATRIXX Digital Commerce
Prometheus · Grafana · Datadog · NetFlow analyzers · Wireshark
TM Forum · ISO/IEC 27001 · AI Act · ETSI · GSMA · 3GPP
EU AI Act took effect in August 2024. Mandatory requirements for high-risk systems — from August 2026. AI agents in customer service contact centers are classified as high-risk because they make decisions affecting customers.
For new deployments — yes, no alternatives. For existing legacy — depends on business case. Cloud-native — lower TCO, faster new service launch. Legacy on-prem — better latency, full control, no cloud-provider lock-in.
For tier-1 operator — 6–10 months with parallel run. Without parallel run — should not do at all, because failed routing costs customers faster than budget.
In 2026 AI closes first-line (60–70% of typical queries). Complex cases — human. Savings — mostly AHT reduction and more parallelism, not headcount reduction. Headcount reduction as a goal usually fails on NPS.
For high-risk systems: (1) risk management system; (2) data governance and training data quality; (3) technical documentation; (4) record-keeping (audit trail); (5) human oversight; (6) accuracy + robustness + cybersecurity. Preparation — 6–9 months.
Real projects rarely fit in one competency. See which other areas we work in.
30-minute discovery call with a telecom architect. We will discuss current state and options, no commitments.