We Made 1,097 AI Cold Calls in the UAE. Here's What Actually Happens.

What exactly did we measure, and how do we know it's real?

Between 2 June and 12 July 2026, MAJ Leads ran an AI voice agent on one live outbound calling campaign in the UAE. Every dial, every connected second, and every outcome logged itself automatically in our own production database. Nothing here is a survey, an estimate, or an industry-wide average. It's 1,097 real dials from one real campaign, reported honestly, including the parts that don't flatter the pitch.

The core numbers cover 1,097 dials placed over six active weeks, plus one additional dial logged on 27 July that we've kept in the total rather than quietly dropping. All times in this article are Gulf Standard Time (GST). No client is named or identifiable anywhere in this piece: every figure below is a fleet-level aggregate, not one business's private performance data.

Two definitions carry the whole analysis, so we're stating them once and using them consistently. A call counts as "connected" once it runs 10 seconds or longer. It becomes a "conversation" once it passes 30 seconds. Anything shorter than 10 seconds is a no-pickup, an immediate hang-up, or a wrong number, not a connection. Sentiment and qualification scoring ran on a smaller, transcript-scored subset of 634 calls. We repeat that n=634 every time we cite it, because a sample that size deserves the caveat, not a headline percentage presented as though it covered all 1,097 dials.

How many of the 1,097 calls actually connected?

525 of the 1,097 dials connected for at least 10 seconds, a 47.9% connect rate. Roughly one call in five, 21.1% (232 calls), ran long enough to count as a real conversation of 30 seconds or more. The rest, 51.9% of dials (569 calls), never crossed the 10-second threshold at all.

Attempt number matters more than almost anything else we measured. First-attempt dials connected 53.7% of the time (431 of 802). By the second attempt at the same number, the connect rate dropped to 31.9% (94 of 295), a fall of nearly 22 percentage points. If a number doesn't pick up on the first attempt, the odds of a second attempt working are meaningfully worse, not roughly the same.

What happens once a call actually connects?

Connecting is not the same as succeeding. Here's how the 525 connected calls actually broke down:

Zoom out to all 1,097 dials and the booking rate is 3.4% (37 calls): roughly one in every 30 dials ends in a booked meeting. Zoom in to just the 232 calls that became real 30-second-plus conversations, and the booking rate climbs to 15.9% (37 of 232). Once a prospect actually talks to the agent for half a minute, better than 1 in 7 of those conversations converts.

The median connected call lasted 28 seconds. The average, 36.2 seconds, sits higher because of a tail of longer calls: 76 calls (6.9% of all dials) ran past the 60-second mark. Across the whole campaign, the agent spent 367.5 minutes on the phone, a little over six hours, to produce 37 booked meetings.

Does calling in the morning actually work better?

Conventional cold-calling advice says call early, before the day gets away from people. Our data says the opposite, at least for this campaign. Calls placed in the 09:00–11:00 morning block connected 44.5% of the time (157 of 353 dials). Calls placed in the 12:00–16:00 afternoon block connected 50.6% of the time (340 of 672 dials), a 6.1-point gap in the afternoon's favour.

13:00 was the single best hour to dial, connecting 53.3% of the time. 09:00, the hour most cold-calling scripts treat as prime time, was in fact the worst, at 38.8%. The late-afternoon hour of 17:00 was nearly as weak, at 40.6%. Dials tail off at both edges of the legal calling window, and the middle of the day carries the volume.

Wednesday was the strongest day to dial, at 54.1%. Monday and Thursday, the two days furthest from the middle of the working week, both underperformed it. No dials were placed on Friday, Saturday, or Sunday, which matters for the compliance section below.

When do calls actually turn into booked meetings?

Connect rate and booking rate are not the same curve. Of the 37 total bookings, most landed in the late afternoon, not around 13:00, the hour with the best connect rate. We're presenting these as raw counts, not percentages: with only 37 bookings spread across nine hours, a per-hour percentage would imply more precision than the sample supports.

16:00 produced more bookings (8) than any other hour, despite a connect rate (49.1%) that was middling, not the best. Whatever is driving that, who picks up at that hour, or how a conversation that starts late in the working day tends to unfold, it converts once it connects.

Do people actually get hostile with AI cold callers?

This is the finding that surprises people most. Of the 634 calls we scored for sentiment, 86.9% (551 calls) were neutral. Only 5.4% (34 calls) were positive, and 7.6% (48 calls) were negative. Genuinely hostile reactions (shouting, swearing, demanding to be removed in anger) showed up on exactly one call out of 634, 0.16%.

The dominant reaction to an AI cold call, in other words, is not anger. It's mild, neutral politeness, the same tone most people use to get any unwanted call off the phone quickly, human or otherwise. Sentiment also turned out to be the strongest qualification signal in the data: of the 34 positive-sentiment calls, 61.8% (21 calls) went on to qualify, against 3.6% (23 of 634) qualified across the full scored subset. A prospect who sounds pleased in the first 30 seconds is worth far more follow-up attention than the raw qualification rate alone suggests.

Were these 1,097 calls even legal to make?

UAE Cabinet Resolution No. 56 of 2024 sets a hard calling window for marketing phone calls: 9:00 am to 6:00 pm, Article 5(3). Every one of the 1,094 timestamped dials in this campaign fell inside that window. Zero calls before 09:00 GST. Zero at or after 18:00 GST. Zero on Friday, Saturday, or Sunday. That's 100.0% compliance, and it wasn't produced by someone checking a report afterward.

What does this mean if you're considering outbound AI calling?

A few honest takeaways, roughly in order of how much they should change your plan. Afternoon beat morning for this campaign by a real margin, not a rounding error: if your team is still front-loading dial lists into the first hour of the day out of habit, that habit is worth testing against your own data rather than assumed. The fear that AI cold calling provokes widespread anger is not supported by what actually happened on these lines: 0.16% hostile, out of 634 scored calls, should not be the reason a business avoids trying outbound AI. And connect rate and booking rate are genuinely different curves, so optimising purely for "more connects" will not automatically produce more bookings.

None of this guarantees these exact percentages will repeat on a different list, industry, or campaign. n=1,097 is one campaign's worth of evidence, not a universal law. What it does show is that the numbers are knowable. A properly instrumented AI receptionist and outbound calling deployment logs every dial, second, and outcome automatically, which is the only way to get an honest answer to "does this actually work" instead of a vendor's estimate. Our services overview covers what a deployment like this involves to set up and run; if inbound booking matters more to you than outbound dialing, our AI appointment booking page covers that side specifically.