Agentic AI Just Cleared Its ROI Bar. For Dealers, the Test Is the Phone.

On June 18, a number moved through customer-experience circles that would have read like a vendor's wish list a year ago: 96% of organizations running agentic AI in production said it met or exceeded their return-on-investment targets in 2026. The figure comes from a benchmarking study by CCW Digital and SoundHound, and it is worth reading carefully — both for what it says and for what it does not.

For a dealership, the temptation is to file this under "more AI hype." That would be a mistake, but so would treating an enterprise contact-center survey as a promise about your store. The useful move is the analyst's move: separate the signal from the marketing, then ask where — if anywhere — it touches the parts of a dealership that actually leak money.

Myth: Customers will always route around a "robot" on the phone.

What the data shows: 50% of organizations with agentic AI now report customers are more inclined to engage with AI chat and phone systems — a reversal of years of documented self-service avoidance. (CCW Digital and SoundHound, June 2026)

What the study actually found — and who was asked

The headline numbers are striking. Of the deployments surveyed, 54% met ROI expectations and 42% exceeded them. Eighty-two percent of teams said the rollout was as easy as expected or easier, and 90% expect at least a quarter of their customer interactions to be fully resolved by AI within five years.

Three caveats keep this honest. First, the study was commissioned by a voice-AI vendor, so treat it as directional rather than independent. Second, the respondents are enterprise CX and operations leaders at companies with at least 500 employees and 500 million dollars in revenue — not franchised rooftops. Third, and most important for anyone picturing an empty BDC, 94% of these organizations still keep humans monitoring or supporting at least some AI-handled resolutions. This is not a story about removing people. It is a story about a technology that has moved from "deflect the customer" to "resolve the request," with staff supervising the edges.

The detail that matters most for dealers is quieter: voice is no longer the laggard channel. Agentic AI now runs on 74% of surveyed organizations' live chat, 67% of their email, and 53% of their phone lines. Phone was the channel everyone assumed would be last to work. It is now past the halfway mark.

Why the timing lands for dealers

Context decides whether a tool is interesting or essential. Here is the context dealers are operating in right now.

The Q2 2026 Cox Automotive Dealer Sentiment Index shows current-market sentiment improving to 43, its second straight quarterly gain — but the three-month outlook fell to 47 from 56, dropping back below the break-even line of 50. The cost index climbed to 74, its highest in more than a year. More than half of dealers (55%) named the economy as the single biggest factor holding back their business, and the profit index, at 36, remains well under water.

Read those numbers together and a pattern emerges: costs are up, the outlook is softening, and margins are thin. In that environment, the expensive lever — adding headcount to chase more volume — is the least attractive one. The cheaper lever is recovering revenue you have already earned the right to capture, and then lost. Dealers clearly sense the shift: in a Digital Dealer survey reported in December 2025, 76% of U.S. dealers said they planned to increase AI spending in the coming year, with AI voice agents among the most-targeted tools. The open question is where to point that spending. The most defensible answer is the phone.

The phone is where the leak is measurable

Most dealership revenue leakage is unglamorous. It is not a failed marketing campaign; it is calls that ring out, especially in service.

The platform vendor Numa, drawing on data from roughly 600 dealerships, estimates that a typical service department misses an average of 158 calls a month, and that stores in the 75th percentile miss around 216. At an average repair order of about 450 dollars, the math gets uncomfortable quickly.

Source: Numa analysis of ~600 dealerships, 2025. The math assumes every missed call represents a recoverable repair order, so read it as the size of the opportunity, not a guaranteed loss.

Even discounted heavily, the figure is large — and it compounds, because roughly 75% of callers who do not reach a live person on the first try never call back. They call the shop down the road.

The sales side has its own version of the same problem, and the research here is old enough to be boring, which is what makes it trustworthy. Harvard Business Review's classic study "The Short Life of Online Sales Leads" found that firms contacting a web lead within an hour were nearly seven times more likely to have a qualifying conversation than those who waited just one hour longer — and more than 60 times more likely than those who waited a full day. After hours and over lunch are precisely when a showroom's human coverage is thinnest and a shopper's patience is shortest.

What "agentic" changes versus the old phone tree

Skepticism is fair here, because dealers have been sold an "AI phone" before and gotten a glorified menu tree. The distinction in the 2026 research is capability, not branding. Earlier systems generated answers and then left the customer to finish the job. Agentic systems execute multi-step tasks: look up an open repair order, check whether a part arrived, book the appointment, and route to a human when the request falls outside their lane. In the survey, 28% of enterprise deployments now resolve complex, non-standard issues end to end; the rest blend AI with human hand-offs.

For a dealership, the realistic 2026 use case is not "replace the BDC." It is narrower and more defensible: cover the calls that currently go to voicemail. After hours. During the 8 a.m. service rush. Over lunch. While advisors are walking a customer to the lane. Those are the gaps where a missed call becomes a competitor's repair order, and they are exactly the gaps an always-on voice agent is built to fill. We have written more on where that hand-off line should sit in our look at the dealership phone-coverage gap.

How to evaluate it like an analyst, not a buyer

If the June data tempts you to act, act like you are underwriting an investment, not buying a gadget.

• Size the leak before you shop. Pull 30 days of your phone system's missed and abandoned-call report, broken out by department and hour. You cannot price a fix you have not measured.
• Tie the pilot to revenue, not "calls answered." The metrics that matter are appointments set, repair orders booked, and lead-to-first-response time — not raw answer rates.
• Keep humans in the loop and define the hand-off. Ninety-four percent of the enterprises in the study do. Write down exactly which calls the agent completes and which it transfers.
• Watch CSAT, not just deflection. The point of agentic systems is resolution, not avoidance. If satisfaction slips, the tool is being used as a wall instead of a bridge.
• Pilot one queue first. Service overflow is usually the cleanest place to start, because the value is easy to attribute and the scripts repeat.

The honest takeaway

The June study is a real marker. For enterprises that deploy carefully, agentic AI has crossed from pilot theater into measurable ROI, and voice has quietly become a mainstream channel rather than an experiment. That does not mean a dealership should rush, and it does not mean the technology runs itself — the same data shows humans still supervising nearly everywhere.

What it means is more specific and more useful: the question for a dealer in 2026 is not whether AI can answer a phone. It is which of your phones is losing money, and at what hour. Find that, and you have found the only business case that matters.

If you want to hear how an AI voice agent actually handles dealership calls — and where it should stop and hand off to your team — that is the conversation we have every day at Carbuki.

Sources

• CCW Digital and SoundHound AI, "Research Finds 96% of Organizations Report that Agentic AI Deployments Met or Exceeded ROI Expectations in 2026," June 18, 2026 — report announcement and full study
• Cox Automotive, "Q2 2026 Cox Automotive Dealer Sentiment Index," May 26, 2026 — coxautoinc.com
• DealershipGuy, "76% of U.S. dealers plan to increase AI spending in the coming year — survey," December 2025 — news.dealershipguy.com
• Numa, "The Real Cost of a Missed Call at a Car Dealership," 2025 — numa.com
• Harvard Business Review, "The Short Life of Online Sales Leads," 2011 — hbr.org