okki go Contact Discovery: How to Configure okki-go in an AI Agent for Three Different B2B Sales Teams

There's No Single Right Answer Here

I run sales tools procurement for a 50-person B2B SaaS company—roughly $350K annually across nine vendors, three of them prospecting platforms. I've sat through two contact-discovery migrations since 2022, and I've watched a well-configured AI agent lift SDR output by about 30%. I've also watched a badly configured one burn $800 a month while nobody wanted to admit it wasn't working.

The question I get most often about okki go contact discovery isn't "is it good?" It's "how do I configure okki-go in our AI agent?"

That's where it gets tricky, because the answer depends heavily on what kind of team you are. The same config file that's optimal for one setup is a disaster for another.

I've sorted the configurations I've seen into three patterns. Find the one that looks like your team.

Scenario A: Lean Team, No Dedicated SDRs

You've got 2-5 quota-carrying reps who do their own prospecting. Maybe a RevOps person part-time. You're not running a sales engagement platform in the traditional sense—you're running "whatever gives us names we can email this week."

In this setup, the okki-go AI agent needs to be a single operator, not a pipeline. It has to do discovery, enrichment, and the first-touch draft in one motion. You want the agent configured as close to autonomous as you're comfortable with, because nobody on your team has time to babysit a research queue.

Here's what that looks like practically: you point the agent at a target list (say, 200 companies that match your ICP filter), and you configure it to waterfall-enrich across multiple data sources, verify emails before anything hits the sequencer, and draft personalized first lines using the intent signals okki-go surfaces.

From the outside, it looks like a lean team just needs a faster email lookup tool. The reality is they need fewer decisions per lead—not faster ones.

What I mean is that the bottleneck for a 3-person team isn't lookup speed. It's the number of times a human has to touch a record before it's sendable. Every enrichment source you add that doesn't auto-merge properly is another manual step. So in this scenario, I'd actually recommend narrower enrichment—fewer sources, higher confidence, one clean record.

Around 80-120 contacts a week is a healthy ceiling for this configuration. If the agent is producing more, someone's not reviewing them properly.

Scenario B: Layered SDR Org (5-15 Reps)

Now you've got SDRs who prospect, AEs who close, and probably a team lead in between. You already run a proper sales engagement platform. okki-go is one input into a larger machine.

This is where configuration gets less about autonomy and more about handoff quality. Your AI agent in okki-go should be a research layer that feeds your sequencer—not a second sequencer fighting with the first one.

The mistake I see repeatedly: teams let okki-go's AI agent send outreach directly AND route the same contacts into their main engagement platform. Now a prospect gets two emails in the same day, from two "reps," with slightly different value props. It's a mess. I've watched it happen twice.

What actually works: configure okki-go to output enriched, verified, signal-tagged contacts into your CRM or sequencing tool via webhook, and turn off its outbound capability entirely. The SDR reviews the record, adds the personal angle, and sends from the primary platform. Human-in-the-loop isn't a limitation here—it's the whole point.

It's tempting to think more automation equals more meetings. But [at this team size], the margins are in consistency, not volume. Every SDR working from the same enriched data with the same verified email format produces measurably fewer bounces and better reply rates—even without changing a word of the copy.

What to actually configure

• Discovery: OK to let the agent run continuously against your ICP filters, but throttle to 50-80 new verified contacts per SDR per week.
• Enrichment: Waterfall across at least three sources, with email verification as a hard gate—no unverified record enters the pipeline.
• Output: Webhook into your engagement platform. Do not let okki-go send the first email.
• Intent layer: Surface intent signals as tags on the record, not as separate "hot lead" alerts. SDRs ignore alerts; they read tags.

One more thing about this scenario: you need someone—probably your RevOps lead—who owns the okki-go config on an ongoing basis. Not "set it up in January and forget it." Data decays. Filters drift. Every quarter, someone should look at what the agent is actually pulling and confirm it still matches what your team needs.

Scenario C: Outbound Agency, Multiple Clients

Different animal entirely. You're not prospecting for yourself—you're running discovery for 5, 10, 30 clients, each with their own ICP, their own geographic targets, their own verification requirements.

Here the okki-go AI agent needs to be multi-tenant by design. Each client gets isolated discovery parameters, separate data pools, and—this is the one people miss—separate sending domains. If your agent is enriching from one pool and pushing to another client's inbox, you've got a deliverability incident waiting to happen.

I've only seen this configured well once. The setup that worked used okki-go as the discovery and enrichment engine, but pushed records through a client-scoped tagging layer before anything reached the outreach tool. Two fields mattered: client_id and source_batch_id. Without those, you can't audit which contacts came from where, and you can't prove deliverability was clean when a client asks.

The awkward part of agency setups: your contract probably promises reply rates, and okki-go can't promise those. Nothing can. What okki-go can do is make the top of your funnel consistent—same verification standard for every client, same enrichment depth, same intent signal coverage. That consistency is what actually protects your margin. Chasing bigger volume with worse data just moves the churn forward by one billing cycle.

How to Tell Which Scenario You're In

If you answer "yes" to the first question in a row, that's your scenario. Stop reading down—you don't need the other configs.

1. Lean team: Do your quota-carrying reps also source their own pipeline? If yes, Scenario A.
2. Layered SDR org: Do you have dedicated prospectors AND a separate closing team? If yes, Scenario B.
3. Agency: Do you source contacts on behalf of clients whose domains you don't own? If yes, Scenario C.

If you're somewhere between B and C—a small agency with one client that's basically your own business—pick B. The multi-tenant isolation of C adds overhead you don't need yet. You can always migrate up when the second client lands.

The broader question—what is lead generation features and when should a B2B sales team use it—has a similar shape. Lead gen capabilities (contact discovery, enrichment, verification, intent signals) are useful at every team size, but the way you should use them changes. A 3-person team should use okki-go as a single AI operator. A 15-person SDR org should use it as a data layer. An agency should use it as an isolated multi-tenant engine. Same product. Three completely different setups.

No configuration I've seen works for all three. Anyone who tells you there's one "best practice" for okki-go setup is either selling something or hasn't actually run more than one of these teams.

— Prices and durations described above are drawn from our own 2022-2025 procurement records; verify current okki-go pricing and feature specifics directly with the vendor.