← Back to Insights
Header image for article: Your IVR Is About to Get Calls From Robots

Your IVR Is About to Get Calls From Robots

ZippiAi Team••5 min read

This week, two consumer AI agents learned to pick up the phone and call businesses. Most contact centres still answer with “Press 1 for billing.”

On 16 September, two of the most-downloaded AI assistants in the United States shipped the same feature within hours of each other. Instinct, the San Francisco startup reportedly in talks to raise at a $10 billion valuation, launched Instinct Concierge: your assistant will now phone a restaurant that doesn’t take online bookings, get you onto a dentist’s cancellation list, or argue with your cable company about a bill. Meta’s Muse, which launched on 8 September, expanded its beta for outbound calls to US businesses the same day.

That is a small product update with a large implication for anyone who runs a phone line. For thirty years, the caller on the other end of an IVR has been a human being with a keypad. From this month, a growing share of inbound calls will come from software that was built to talk, not to press buttons.

How big is this, really?

Bigger than a beta flag suggests. According to Sensor Tower data cited by TechCrunch, Muse passed 730,000 US downloads in its first days, more than the Meta AI app managed in its first five (707,000), and briefly reached number two in the US app charts. Instinct raised $350 million at a $2.5 billion valuation in August and is now reportedly negotiating $1 billion at four times that price. Both apps are free at the entry tier; Muse charges $20 and $100 a month for heavier use.

The direction of travel is also clear from the feature list. Instinct already lets a user’s assistant contact other people’s assistants within a trusted network. Muse connects to email, calendars and payments, and checks out through Link by Stripe. These are not chatbots that answer questions. They are agents that complete tasks, and completing a task with a business very often means calling it.

Customers already prefer their own AI to yours

The demand side of this shift is not speculative. Gartner surveyed 3,566 B2B and B2C customers in February and March 2026 and found that customers were roughly three times more likely to use a third-party generative AI tool than a company-provided chatbot when resolving a service issue. Use of third-party GenAI during service interactions nearly doubled in a year; use of company chatbots has been statistically flat since 2022.

The same survey explains why. Among customers who use GenAI, 58% have used it to complete a task on their behalf (74% in B2B). Only 7% used a company chatbot in their most recent service interaction, even though 49% said they would have been willing to. And customers give company-owned automation a short leash: just 27% would try a chatbot again after one negative experience.

Put those numbers next to this week’s product launches and the picture is simple. Customers are delegating service tasks to an agent they trust more than yours, and that agent is now capable of dialling your number.

IVR vs. conversational voice AI: what happens when the caller is a machine

The difference between a touch-tone IVR and a conversational voice front door has usually been argued in terms of customer effort. An AI caller changes the terms of the comparison.

Scenario

Touch-tone IVR (DTMF menus)

Conversational voice AI

Caller states intent in a sentence

Must map free speech to a numbered menu; often fails and loops

Extracts intent directly and routes or resolves

Caller has already gathered context (account, order, issue)

Asks for it again, keypad entry, no memory

Accepts it in conversation; can verify and carry it to a human

Caller is an AI agent acting for a customer

Agent must guess menu structure and time its tones; long hold music burns the agent’s budget and the customer’s patience

Two systems negotiate the task in natural language; hand-off to a human only when needed

Outcome measured

Menu completion, containment

Task completion, first-contact resolution

Data captured

Key presses

Full transcript, intent, sentiment, outcome

The last row matters as much as the first. A keypad tells you which number was pressed. A conversation tells you why the customer’s agent called, what it tried to do, and whether it succeeded. That is the data a support organisation needs to decide what to automate next, and an IVR simply does not produce it.

The humans still matter, which is the point

None of this argues for removing people. Gartner found that 87% of customers say access to a human agent is essential when companies use GenAI in customer service. Call Centre Helper’s coverage of Simplyhealth’s deployment with PolyAI shows what a well-designed voice AI front door looks like in practice: three intent-led voice assistants resolving roughly 30% of total call volume without a human, while CRM data routes vulnerable customers straight to a person.

The IVR was designed to protect agent time by making callers do work. A conversational layer protects agent time by doing the work, and it can do that for a human caller or a software one. As the inbound voice share of contact-centre traffic dipped below 50% for the first time in 2025 (Call Centre Helper’s annual report), the calls that remain are the ones that could not be solved in an app. Increasingly, an app will be making them.

What to check this quarter

Three practical questions for any operation still running a menu tree at the front of its phone line. First, can your entry point understand a caller who states a full request in one sentence, including a caller who is reading from a task list? Second, does context collected at the front door survive the transfer to a human, or does the customer (or their agent) start again? Third, what does your reporting show about why people called, rather than which digits they pressed?

If the answers are no, no, and “digits”, the calls arriving from Instinct and Muse will not look like an opportunity. They will look like a spike in abandonment.

At ZippiAi we build Aura, a voice AI companion and analytics layer for enterprise support. The comparison table above reflects how we think about the front door of a contact centre: it should hold a conversation with whoever, or whatever, is calling, and it should tell you what happened.


Sources