How Brands Can Build Better Conversational AI That Won’t Go Unheard

If technology matures at a pace closer to dog years than human, then we’ve already witnessed a century of conversational intelligence since Siri first appeared on our iPhones 15 years ago. As ChatGPT and a constellation of other large language models (LLMs) continue to evolve at an exponential speed and scale, legacy interactive voice response (IVR) systems are being replaced as consumers increasingly expect intuitive artificial intelligence agents across chat and voice.

From retailers, to financial institutions, and fast-food chains, hundreds of thousands of businesses serving millions of users are already integrating AI agents into their consumer experience apparatus. In fact, one-third of organizations are in the final implementation phase of conversational AI in customer service, while 28 percent have already fully deployed.

While consumer comfort and familiarity with conversational AI continue to grow, so do expectations. The Wild West days of two, three years ago — rife with hallucinations, prompt injections, and reputational landmines from businesses throwing generic LLMs at any and all traffic — are finally giving way to some law and order. Today, performance is paramount, and deployments must be far more deliberate.

With the conversational AI market projected to surpass $40 billion by 2030, a more than 250 percent increase from just last year, how can brands ensure their conversational AI strategies and investments prioritize consumers and not simply the hype cycle?

Master the Primitives

The barometer I often give our customers to gauge if their conversational AI is any good is painfully simple: is it good at conversations? Most of us wouldn’t tolerate a human who constantly interrupts, takes long, awkward pauses, or gives unhelpful information to basic questions. Consumers are no longer giving AI agents that grace either.

Speech recognition, noise cancellation, low latency, turn detection, natural tone, interruption handling, and voice activity detection are all primitive capabilities that should be the baseline for any conversational agent. It won’t matter if you have the Rolls-Royce of AI under the hood if the conversational layer is a Pontiac Fiero. If the interface can’t speak or listen at a natural cadence, consumers will drop into the uncanny valley and reject the interaction.

Take a fast-food run I recently made with my family. After I found the nearest location on Google Maps, I called the restaurant and was greeted by an AI agent that took my order. Within a minute or two, I placed my order (which was accurately repeated back for confirmation), received SMS verification with my receipt, and completed secure payment with two-factor authentication — all with the windows down and three hungry kids in the back seat.

The most phenomenal part of this experience came about a week later when I placed a subsequent order at the same restaurant. The AI agent immediately recognized me, recalled my previous order, and asked if I wanted the same meal. I simply had to say what I wanted substituted and confirm my on-file payment method. I was off the phone in less than 60 seconds. It may be a minor use case, but the personalization and precision of this experience was anything but (and honestly, if these LLMs can ace the SATs, they should be able to take a pizza order).

Baked into all of this (no pun intended) must be robust trust and security. What if, instead of a fast-food order, I was checking recent transaction history with my bank? If a bad actor spoofs my number and calls the agent back, without proper verification, they could simply ask for account details and access sensitive information from the previous call. Businesses must ensure their conversational AI agents treat every new interaction, whether it’s a request for payment history or a pepperoni pizza, as an opportunity to reinforce identity.

Targeted and Transparent

There’s a stubborn mindset among some business leaders that conversational AI is purely a deflection tool, used not to resolve issues or improve consumer experience, but to reroute inbound traffic away from overburdened contact centers. If traditional IVRs are just replicated with LLMs, the outcome for consumers is often the same: whiplash as they’re bounced between dead-end menu options.

Instead of overhauling the entire IVR at once:

• Identify the primary reasons consumers reach out.
• Select a few that are simple and safe to automate.
• Measure the results (e.g., containment rate, average handling time, agent efficiency, complaint resolution rate).

For retailers with global consumers, this intentional approach should also extend to languages. If your IVR is in English, for instance, take those initial automated options and ensure they’re available in a handful other languages that cover 90 percent to 95% of consumers, then pivot and scale as appropriate.

However, consumers aren’t calling businesses to chat about their weekend plans; they want speedy and satisfactory resolutions. Agents must not only be able to carry a conversation but have the access and agility to connect with broader applications to resolve them: booking a refund, changing an order, scheduling an appointment, completing a purchase. Unsurprisingly, half of all consumers who were dissatisfied with their AI interaction said it was because the agent didn’t resolve their issue, and two-thirds said they would prefer to use AI agents if they fully solved their issue faster than a human.

If a targeted approach is critical, then so too is a transparent one. While businesses continue to take varying stances on how candid they are about agent interactions, it's likely we’ll see increased regulatory pressure going into 2026 to drive greater consistency and accountability.

This could take the shape of:

• Branded numbers and icons on phones during outbound calls — AI agent: Appointment reminder from Joe’s Auto Body.
• Automatic disclosures upon pickup — ”This is Sherri, an AI agent from Joe’s Auto Body reminding you of your appointment tomorrow at 2 p.m. Do you have any questions I can help answer?”

I experienced this firsthand with a company that helps expedite passport applications. After I registered (including opting into its customer support) and uploaded the required documents and photos, I received an automatic outbound call along with a SMS message from its AI agent to answer questions about my application status.

If businesses can complement transparency with deep personalization — ”This is Matt, an AI agent to help you navigate your passport application. I see you’ve been on step three for a few days; do you have any questions I can support you with?” — it not only makes agents seem intelligent but immediately establishes trust and confidence from the first touchpoint.

The New Brand Ambassadors

In the early-to-mid ‘90s, websites were the bleeding edge of consumer engagement, highly customizable platforms designed to inform, inspire and interact with audiences. Thirty years later, conversational AI is quickly picking up the digital mantle as the new frontline interface.

At a time when 90 percent of businesses believe their customers are satisfied with their conversational AI experiences, but only 59 percent of consumers agree, those that can balance core performance capabilities with security, transparency, and strategic clarity will deliver more frictionless interactions — ones that won’t fall on deaf ears.

Andy O'Dower is the vice president of product management for Voice and Video at Twilio, a customer engagement platform that drives real-time, personalized experiences.