Skip to content

Voice Is the Second Wave of AI (After Chat)

· Jan Cervenka, Co-Founder & CEO · 4 min read
voice-ai ai-receptionist call-automation conversational-ai customer-support

The first wave of modern AI was chat.

A text box is the easiest interface in the world:

  • it’s cheap to ship
  • it’s easy to retry when the model is wrong
  • and it fits neatly into websites, apps, and internal tools

That’s why “chatbot + knowledge base” exploded first.

But chat isn’t the final form of AI in the real world.

Voice is the second wave. Not because voice is trendy, but because voice is where real-time business happens: customer support, reservations, claims, logistics, scheduling, triage, and the countless “I need help right now” moments where typing is the last thing people want to do.

Why chat came first

Chat won wave one for a boring reason: it’s forgiving.

If a chat assistant is slightly off:

  • the user can rephrase
  • the user can skim and ignore the wrong part
  • the user can keep browsing while the assistant thinks

Chat is also naturally asynchronous. It doesn’t need perfect timing, interruption handling, or turn-taking.

Voice does.

Why voice is wave two (and why it didn’t happen earlier)

Voice has always been “automated” — we’ve all suffered through phone trees and IVR menus. But that wasn’t the same as real conversation.

Voice is harder than chat because it forces the system to handle:

  • latency (silence feels broken)
  • interruptions (people talk over each other)
  • ambiguity (people are vague on the phone)
  • noise and accents
  • high-stakes outcomes (bookings, changes, refunds, claims)

Old-school IVR systems solved this by avoiding conversation entirely: “Press 1 for hours. Press 2 for billing.”

Customers hate it because it routes, it doesn’t resolve.

So voice had to wait until the underlying tech could do something different: fast, natural, back-and-forth conversation.

Voice isn’t just “chat, but spoken”

This is the part many people miss.

Voice changes behavior.

When people talk, they:

  • explain context faster
  • interrupt themselves
  • change their mind mid-sentence
  • ask follow-ups immediately
  • express emotion (stress, urgency, confusion)

That makes voice a better interface for real problems, not just FAQs.

It also makes voice harder to get right, because you’re dealing with messy humans in real time.

Why voice matters to businesses (beyond “answering calls”)

Voice wins whenever the user needs:

  • speed (they want help now)
  • hands-free interaction (driving, working, holding a kid, in a shop)
  • emotional reassurance (complaints, confusion, urgency)
  • a two-way resolution (not just information)

That applies to a huge set of industries:

  • customer support and contact centers
  • banking and insurance
  • healthcare admin (not medical advice, but scheduling, reminders, logistics)
  • travel and hospitality
  • deliveries and field services
  • retail returns and order issues

And importantly: even “digital-native” consumers still use phone support for help and support.

McKinsey reported that in a survey of 3,500 consumers, live phone conversations were among the most preferred methods of contacting companies for help and support, including among Gen Z.
Source: McKinsey: Where is customer care in 2024?

So voice isn’t a legacy channel. It’s a core channel.

The market signal: businesses are investing heavily in voice + call automation

This isn’t just a vibe shift. There’s real budget moving.

Grand View Research estimated the global call center AI market at about USD 1.99B in 2024 and projects it to reach USD 7.08B by 2030.
Source: Grand View Research: Call Center AI Market Report

That’s a lot of investment chasing one outcome: handle more voice interactions with better quality and lower cost.

The big constraint for voice: trust

Voice isn’t only a model problem. It’s a trust problem.

A big part of the modern phone experience is spam and scams. People ignore unknown numbers.

TransUnion reported that nearly 8 in 10 consumers consider phone important for communicating with businesses, but many consumers also block calls from numbers they don’t know.
Source: TransUnion research on phone channel importance and call blocking

This is one reason voice AI has to be built like a product, not a demo:

  • clear caller identity (so people answer)
  • consistent experience (so people trust it)
  • safe handling when uncertain (so it doesn’t confidently do the wrong thing)

What will separate “toy voice bots” from real voice AI

Wave one (chat) had a million “good enough” assistants.

Wave two (voice) will be less forgiving.

The winners will be the systems that combine:

  • natural conversation (so it doesn’t feel like an IVR menu)
  • structure and guardrails (so it doesn’t hallucinate or take wrong actions)
  • reliable data sources (so availability, policies, and outcomes are correct)
  • safe fallbacks (so edge cases don’t become disasters)

In other words: voice agents need to be engineered like systems, not prompted like chatbots.

Where Switchly.ai fits into this second wave

Switchly.ai exists because voice is still the most under-served part of “AI for business.”

Chat assistants are everywhere now.

But voice is where businesses actually lose time, lose customers, and get interrupted mid-work. And voice is also where customers still expect a fast resolution without pressing buttons into a phone tree.

Our bet is simple:

  • chat was wave one
  • voice is wave two
  • and the businesses that adopt reliable voice-first AI early will have an unfair advantage

What comes next

The long-term endpoint isn’t “voice replaces chat.”

It’s that AI becomes available in whatever interface is most natural for the moment:

  • chat when you’re browsing
  • voice when you’re moving, working, or urgent
  • multimodal when context matters

Chat proved the model can be useful.

Voice is proving the model can be operational.

And that’s why voice is the second wave of AI.