1 hr ago
AI’s Next Frontier Is Making Conversation Infrastructure Reliable
AI systems can already think, write and solve many problems.
But talking with people on the phone is harder than simply finding an answer.
A phone conversation needs quick responses, accurate listening and smooth turn-taking.
Small delays or misunderstandings can make people think the system is broken.
Customers also want to feel understood and may still want a human available.
The article says many businesses use voice AI but are not very satisfied with it.
It argues that special infrastructure is needed to connect AI agents reliably to telephone networks.
This infrastructure should work with different AI models and keep companies in control of their customer data.
If it works well, customers will experience a normal conversation without noticing the technology behind it.
AI agents may be highly capable but still struggle with pauses, interruptions, accents and human handoffs during customer calls.
Deloitte’s 2026 research found nearly three-quarters of surveyed organizations expect to use AI agents moderately by 2027.
Qualtrics found that feeling understood was most closely linked to resolving customer issues, but AI agents performed worst on that measure.
Gartner reported that 87% of customers still consider human access essential when companies use generative AI for service.
Deepgram found that 80% of companies use voice agents, while only 21% are very satisfied with them.
- Who
- AI companies, businesses using voice agents, and their customers are central to the issue; the article was written by Bjas Murthy, CEO of Prudent Technologies.
- What
- The article argues that AI’s next major challenge is reliably delivering intelligent, natural conversations over telephone networks.
- Where
- The challenge occurs in phone-based customer service, across the infrastructure connecting AI agents with telephone networks.
- When
- The article cites Deloitte’s expectations for 2027 and research from 2025 and 2026; it does not provide a publication date.
- Why
- Voice conversations expose latency, speech-recognition, turn-taking and data-management problems that can prevent capable AI from serving customers effectively.
Key facts
- Expected AI-agent adoption
- Nearly three-quarters of organizations surveyed by Deloitte expect to use AI agents at least moderately by 2027.
- Customer preference
- Gartner found that 87% of customers consider access to a human essential when companies use generative AI for service.
- Voice-agent usage
- Deepgram’s 2025 State of Voice AI study found that 80% of companies already use voice agents.
- Voice-agent satisfaction
- Only 21% of companies in Deepgram’s study were very satisfied with their voice agents.
- Customer understanding
- Qualtrics studied more than 7,000 consumers and found that feeling understood was most closely linked to resolving an issue.
- Proposed solution
- The article advocates an independent layer connecting AI agents with telephone networks and supporting cloud, private-infrastructure or on-premises deployment.









