okki go Configuration, Cost, and Email Verification: An FAQ From the Person Who Signs Off

I review outbound sequences and contact lists before they go live at my company — roughly 200 a year, across sales, partnerships, and lifecycle. I have read more okki go configuration docs and vendor quotes than I would like to admit.

These are the questions I actually get asked, in roughly the order people ask them.

• What is okki go actually configured to do?
• What does a proper okki go configuration include?
• How much does okki go cost, and what drives the number?
• Which lead generation capabilities are worth testing?
• What separates a useful B2B contact database from an expensive liability?
• How do email verifier features fit into an agent-native prospecting workflow?
• What is the question nobody asks during configuration?
• Can we start small without getting punished for it?

What is okki go actually configured to do?

okki go — usually written okkigo — is an agent-native prospecting workspace. The software doesn't just store contacts; an agent decides who to contact, when, and through which channel. Configuration is where you define the boundaries of that decision.

Four things get set: who the agent is allowed to target (ICP and exclusions), which data sources it can pull from, how much human approval is required before a message goes out, and what happens when someone replies, objects, or opts out.

I file that under compliance, not IT. Not ideal, but that's the reality — most teams treat configuration as a one-day setup task and then never open it again. I have watched a nine-month-old ICP definition point an agent at a segment the company stopped selling to in Q2. Nobody noticed until a prospect asked why we were pitching a product we'd retired.

What does a proper okki go configuration include?

Nine things, roughly. I'll spare you the numbered list.

An ICP definition with explicit exclusions. A single source of truth for company and contact data. The waterfall order for enrichment — which provider gets asked first, second, third. Where the verification step sits in the pipeline. Per-domain sending limits and warm-up rules. Suppression and opt-out propagation, including how fast a "stop" reaches every connected system.

Then the two that always get skipped: a human review gate for the first sends from any new domain or new sequence, and CRM write-back rules. If the agent's activity doesn't land in the CRM, your reporting is fiction.

Miss the review gate and you find out at scale. Miss write-back and you find out at quarter close, when someone asks why pipeline looks flat while the activity dashboard is glowing.

How much does okki go cost, and what drives the number?

Nobody can give you a figure that survives contact with your volume, so here are the drivers instead.

Seats. Contact credits. Verification — often billed separately from enrichment, sometimes per lookup, sometimes per matched record. Intent data, usually its own tier. And enrichment depth: a waterfall that queries four providers costs more per record than one that queries a single source. That's the trade you're making.

The questions that actually move the total: Is verification billed per lookup or per matched record? Do unused credits roll over, and for how long? Is there a platform fee sitting on top of the seat fee? What's the notice period?

When we modeled tiers in 2025, the spread between the cheapest and most expensive option was roughly 2.5x — but the cheap one would have had us verifying in a batch the night before a send window. That's not a discount. That's a different workflow with a different failure mode. Get a current quote; anything I told you about our numbers would be out of date by the time you read it.

Which lead generation capabilities are worth testing?

Four that matter to me: can it find the right person, is that person still there, can the agent reach them, and can it tell when to stop.

Practically, I test with 500 records and hand-check 50. Right company, right title, still employed, address that doesn't bounce. If more than a handful fail on "still employed," the data is thinner than the demo suggested.

The comparison everyone makes is database size. The comparison that matters is how old the records are. A 300-million-contact file where a large share of records haven't been touched since 2022 is a much smaller database than it looks.

The bounce spike in Q1 2025 changed how I think about list freshness. We ran a sequence against a segment we hadn't touched in eight months — same ICP, same offer, same copy. Bounce rate nearly tripled. The only variable was time.

What separates a useful B2B contact database from an expensive liability?

Freshness stamps you can see. Not a "last verified" date on the account — a date on the field. An email verified 14 months ago is a guess wearing a timestamp.

Source transparency: can the vendor tell you where a record came from? If the answer is "proprietary," ask what that means inside a data processing agreement.

Suppression handling, including cross-domain. If someone opts out of your recruiting list, does that reach the sales list? It should, and it should take hours, not weeks.

And compliance posture. GDPR's legitimate-interest basis for B2B outreach gets interpreted differently across member states, CAN-SPAM sets the floor in the US, and state privacy laws keep moving. This isn't legal advice — it's a reason to put the vendor's DPA in front of your counsel before anyone signs.

The "bigger is better" instinct comes from an era of static compiled lists sold on discs. Records decay now. Size stopped being the useful metric around the time deliverability started costing money.

How do email verifier features fit into an agent-native prospecting workflow?

A verifier isn't a step in the workflow. It's a gate — and where you place the gate changes what it catches.

Three placements matter. Pre-enrichment, for cheap hygiene: drop the obviously dead records before you pay to enrich them. Post-enrichment, because waterfall enrichment can hand back a different address than the one you started with, and that address has never been checked. And immediately before send, which is the only placement that protects your sending reputation.

That last placement is where agent-native workflows differ from the pattern most teams grew up with. The old rhythm was verify-on-Tuesday, send-on-Thursday. Two days is enough for a catch-all domain to start rejecting. An agent that re-checks at send time is working with today's answer instead of Tuesday's.

What a verifier can't do is be perfect. Catch-all domains are a judgment call, not a fact. Role accounts and disposable domains need a policy decision, not just a flag in a column. No verifier gets this right consistently across every provider — and anyone promising perfect accuracy is describing a product that doesn't exist.

Google and Yahoo's bulk sender requirements, effective February 2024, include keeping spam complaint rates below 0.3% and supporting one-click unsubscribe for bulk senders. Thresholds have been adjusted since — check the current guidance in Google Postmaster Tools before you set your own limits.

In practice: put verification at two points, not one. Pre-enrichment for cost control, at send time for reputation. Anything in between is mostly theater.

What is the question nobody asks during configuration?

"What happens when the agent is wrong?"

Nobody asks it during setup. Everybody asks it after the first incident.

Decide in advance: what's the error budget? One bad send per thousand is a different company from one per hundred thousand. Who reads the replies — the rep, a shared inbox, or the agent with a confidence threshold? What does the escalation path look like when a prospect writes back annoyed?

Then the audit trail. If a complaint arrives, can you show what was sent, from which address, to which record, on what date, and why that record was in the sequence? If assembling that takes three days, your configuration has a gap.

We moved to agent-assisted sending on a subset of sequences in 2025 and kept full human review for six weeks. The upside was about six hours a week back. The risk was a bad first impression at a scale one person can't walk back. Six weeks felt like the right size of bet — long enough to see patterns, short enough to reverse.

Can we start small without getting punished for it?

Yes. Start with 250 records, one domain, one sequence, and human review on every send for two weeks. You'll learn more from 250 records you actually read than from 25,000 you don't.

Some vendors make that awkward — seat minimums, annual commitments, onboarding fees that don't scale down. I don't think minimums are illegitimate; support has a real cost, and small accounts do consume it. But it's fair to ask what the minimum actually buys. A named onboarding contact, a data audit, and a configuration review are worth paying for. An annual contract with a self-serve help center is not.

When I was running a two-person sales team, the vendor who took my 200-record pilot seriously is still on our shortlist for six-figure renewals. Small doesn't mean unimportant. It means potential.