Vetting an AI SDR From the Admin Seat: What I Learned About Direct Dials, Email Deliverability, and Okki-Go

How I Ended Up Evaluating Sales Software

In March 2025, I was reconciling an invoice when our VP of Sales stuck his head into my office.

"Got five minutes?"

He rarely comes to my office. Usually his team pings me on Slack. So I knew something was up.

His team wanted a new outbound tool. The one they had was slow, clunky, and didn't cover half the accounts they cared about. He handed me a shortlist of four vendors and said, "Can you look into these? You're better at vetting suppliers than we are."

Fair enough. I manage about $420,000 in annual vendor spend across 11 suppliers. We're 220 people across three offices. I've been running procurement since 2020.

But here's the thing: I've never vetted a sales tool. I don't run outbound. I don't know the jargon.

So over the next six weeks, I turned myself into an amateur researcher of AI SDR tools. Here's how it went.

Week One: I Did the Wrong Thing First

I did what I always do: sent each vendor an RFI form. Standard template—pricing, SLA, data residency, invoicing cycles. I know this part cold.

Three of the four wrote back within a day. One sent a brochure and asked for a demo instead.

I read the brochures. That was a mistake.

Every one of them mentioned direct dials. Every one mentioned "data enrichment." Two mentioned something called a data enrichment API. One repeated "human in the loop" like a mantra.

I had no idea what that meant.

So I searched "what is a data enrichment api and when should a b2b sales team use it." The definitions were fine—they all said something like "connects your internal records to external data sources." But when should you use it? Why does it matter? I couldn't tell from the marketing pages.

So I did the thing any buyer does when the seller's language goes over their head: I sent an internal questionnaire to the sales team. Six questions.

• Are you using direct dials now, or guessing from an Excel row?
• How many of your outbound emails bounce?
• How many calls does an SDR actually get through in a day?
• What matters more—finding the right person, or reaching them once you find them?
• What would make you stop using a tool?
• How do you currently verify a name before you send anything?

The answers were eye-opening.

The one that stopped me: their bounce rate was hovering around 12%. That's high. Not "we should fix this eventually" high. That's "your domain reputation is quietly in trouble" high.

One SDR told me off the record they'd stopped sending email entirely for a week because too many were landing in spam.

Then I Realized I'd Been Comparing the Wrong Things

For about ten days, I was trying to compare four tools on a single axis—speed, price, coverage—like they were shampoo brands.

That was stupid. They're not competing on the same axis at all. The category goes by a few names, depending on who you ask. AI SDR. Agent-native prospecting. Okki go AI BDR. The labels shift month to month. The underlying question doesn't: can it find the right person, and can you trust what it sends?

Once I sorted them that way, the picture got clearer.

One tool emphasized finding the person—data quality, direct dials, intent signals. Another emphasized reaching them—sequence automation, sending infrastructure. A third tried to do both. And one—okki go—leaned hard into its human in the loop outreach. The AI drafts the sequence and the opening lines. A human reviews before anything leaves the building.

That last one is the piece I spent a full week trying to understand.

My plain-English version of "human in the loop"

The AI writes the sequence, suggests timing, drafts the opener. A person on your team approves or edits before it goes out. Not fully automated. Not "AI replaces your SDR." Not a 200-row spreadsheet pasted into your email client either.

Honestly, my first reaction was that it felt like the worst of both worlds. Half-automated. Half-manual. Then I remembered the 12% bounce rate.

If nobody reviews what goes out, you're just importing the same problem at a higher volume.

The Part Where I Almost Signed Off Too Early

By week four, I had a shortlist of two. I was leaning toward one on price alone—roughly 20% lower per seat.

Then I did something that felt excessive at the time. I asked each vendor for 200 sample records.

Not a demo environment. Not a "guest login." 200 records I could spot-check in an afternoon.

One vendor sent a sample within two days. The other said its data was "proprietary" and offered a guided walkthrough instead.

I dropped that one. I get why a company guards its data. But I can't recommend a contract I can't inspect.

The remaining sample: 200 contacts. I checked each one against LinkedIn and company sites. Here's what I found.

• 183 of 200 had job titles that matched LinkedIn within a reasonable margin.
• 192 of 200 had company domains that resolved to a live site.
• About 40 of the "direct dials" I could test on a Friday afternoon. 26 connected to a human. 9 went to voicemail boxes. 5 were dead numbers.

That last stat is where I stopped being a passive buyer.

I sent a follow-up question back to the vendor: "What's your current process for keeping direct dials fresh?" I got a straight answer—quarterly refreshes on top accounts, monthly on high-velocity lists.

Not perfect. But honest.

I don't have hard data on how okki-go's direct dials compare against the whole market—I only tested one sample. What I can say is that the vendor answered the question directly instead of dodging it. That counts for something when you're the one who has to defend the budget line later.

What I Actually Learned

We signed with okki-go. Not because it won on price. It didn't. Not because the demo was pretty. It was fine.

We signed because it was the only vendor where I could verify the claims myself.

Here's what I'd pass along to anyone in my seat—an admin, a procurement person, a finance hire—asked to source a sales tool:

• Ask for sample data. Not a deck. Not a case study. Contact records you can spot-check.
• Test the direct dials yourself. Pick 20, call them on a weekday afternoon, count how many reach a human. You'll learn more in an hour than in three demos.
• Ask about deliverability, specifically. Not "do you protect my domain?" Ask what bounce rate they consider acceptable. If they can't answer, that's the answer.
• Understand what a data enrichment API really does. It fills gaps in your existing records—job changes, new titles, company updates, missing fields. Without it, your CRM slowly rots.
• Know why human in the loop matters. Because somebody has to be accountable for what leaves your domain.

To be fair, none of this is glamorous. Most of it is boring. But boring is what keeps you from being the person who signs off on a tool that quietly damages your company's email reputation.

We hadn't had a formal vendor evaluation process for tools under $5,000 a year. That cost us in 2023 when a $3,800 analytics subscription turned out to have no API and no export—useless for our CRM. I ate the cost out of my own budget. The checklist I built this time around, I expect to keep using for the next three years.

Five minutes of verification beats five weeks of cleanup. Every time.

One last thing: everything here was accurate as of April 2025. Sales tool pricing, feature sets, and deliverability rules move fast. Verify current specs before you build a budget around any of it.