July's Recall Wave Is a Reachability Problem, Not a Repair Problem.

The first two weeks of July 2026 brought a heavy run of vehicle recalls, around 25 separate actions covering an estimated 1.4 million units in the United States, with the largest campaigns coming from major manufacturers. The coverage, as always, focused on the defects: a starter that can overheat, a rear axle carrier that can fracture, lights that can fail. Those are real engineering problems, and the manufacturers are fixing them.

But the defect is rarely the part that goes wrong at scale.

The gap the headlines skip

Look at what NHTSA publishes about what happens after a recall is announced. Nearly 80% of recalls from 2020 to 2024 had a final remedy available within 60 days. Yet average completion rates land around 61.5% for 2021 recalls, 71.6% for 2022, and 69.2% for 2023.

Hold those two facts next to each other. The fix exists and is ready for four out of five recalls within two months. Roughly a third of affected vehicles still never get repaired. That 15 to 20 point gap between a remedy is available and the owner actually came in is not a manufacturing problem or a parts problem. It is a reachability problem, and reachability is an engineering discipline that almost nobody measures, because it is invisible until you go looking.

The hard part of outreach is never the conversation

The hard part of any high-volume outreach system, recall, service reminder, any campaign that has to move real people to act, is almost never the conversation. It is getting a real, able-to-act person on the line at all. The transcript is the easy, glamorous part. The unglamorous part is everything that has to be true before the transcript matters: the phone number is current and not the number of a car sold two owners ago; the call lands at a time a human will pick up rather than at 8 p.m. when it feels like a scam; the language matches the person; and when the system is not sure who it reached, it degrades gracefully instead of confidently marking a wrong number contacted.

Grade the thing that matters

Any team deploying voice AI at volume drifts toward grading the agent on the parts that are pleasant to grade: does it sound natural, does it follow the script, is the summary tidy. Those matter, but they are downstream of effective reach. It is easy to review a batch of outreach a system proudly reports as completed, see a great dashboard, and then discover on a listen that a meaningful slice reached voicemail, a dead number, or a language the script did not handle, each one logged as a success because the system counted a dial, not a human who could act.

Questions to pressure-test any outreach system

If you want to pressure-test an outreach system, your own or a vendor's, reachability is where to spend the questions, not the demo:

• Show me last week's failed contacts, not the highlight reel: the wrong numbers, the no-answers, the language mismatches, the calls the agent was unsure about.
• What does the system do when it is not confident it reached the right person? Does it retry at a better hour, hand to a human, and suppress the record from completed, or bank the failure as a win?
• How does it decide when to call? Time-of-day moves contact rate more than voice quality ever will.
• How does it handle the second and third owner, the aging car whose contact record is three moves out of date?
• How would you know the system is getting better over time rather than quietly getting worse?

If a vendor can only answer those with a live demo, you have learned the most important thing about them.

This is the argument for automation, done right

None of this is an argument against automating outreach. The opposite. Reachability is a volume-and-timing problem humans cannot brute-force; a campaign across a million vehicles cannot be worked by a team calling nine to five. The machine's advantage is that it can call the right person at the right hour in the right language and retry intelligently, but only if it is built and measured around reachability rather than transcript polish. Build it around the transcript and you get a beautiful system that reports success while the gap sits exactly where it was.

The recall headlines will keep coming. Treat each one as a reminder that the defect is the manufacturer's problem to solve and the reach is everyone else's. A prediction, with a date: by the end of 2026, in any category that runs outreach at scale, the teams that report effective reach as a first-class metric will be visibly outperforming the teams still reporting raw completion, in outcomes, not dashboards.