OKKI Go Natural Language Prospecting: An 8-Step Checklist for Agent-Native Workflows

Who This Checklist Is For (And When to Skip It)

This is for RevOps leads, sales ops managers, and SDR team leads who already have a CRM and at least one outbound channel running—usually a 5 to 50-person team. If you're a solo founder sending 20 emails a day, this list is overkill. Bookmark it for when you hire your first SDR.

Total steps: 8. Realistic setup time: 2-4 hours if your CRM is already clean, closer to a full day if it isn't. The order matters—steps 2 and 3 are the ones most teams skip, and they're the ones that decide whether the rest of the workflow works or just burns your domain.

Step 1: Lock Your ICP—Including the Exclusions

Write your ICP as a filter, not a persona. Something a query can actually act on.

Two lists. The "include" list: industry, headcount range, tech stack signals, funding stage, geography. The "exclude" list: anyone you've lost to in the last 18 months, current customers, anyone in an active deal, and—this is the one most people forget—companies with a headcount growth rate above, say, 60% year over year. Fast-growing companies churn your sequences because their org chart changes every quarter.

I review about 200 outbound sequences a year before they go live. The rejected ones almost always skip the exclusion list. Data hygiene downstream can't fix a list that shouldn't have been built.

Step 2: Waterfall Enrichment Before You Touch a Sequence Tool

Here's something most vendors won't tell you: any single enrichment provider covers 60-70% of a given B2B contact list on a good day. Not 95%. Not "near complete." The number varies by region, seniority, and how recently the person changed jobs.

Waterfall enrichment—running a record through multiple providers in sequence and stopping at the first match—gets you closer to 85-90% coverage. That's the honest range. Anyone promising more is either counting bounced rows or guessing.

In OKKI Go, the enrichment step slots in before your sequence builder, not inside it. The order is:

• Upload raw list →
• Run waterfall enrichment (email, direct dial, LinkedIn URL, company data) →
• Tag enriched-but-still-missing-email contacts separately →
• Only then move the enriched segment into sequences

Keep the un-enriched segment in a "LinkedIn-only" bucket. You'll route them in step 6.

Step 3: Verify Email Before Anything Else Touches the List

This is where email verification service features stop being a checkbox and start being load-bearing.

A verification service should give you four distinct outputs, not a binary valid/invalid:

1. Deliverable — safe to send
2. Risky / accept-all — the domain accepts everything; send at low volume, monitor
3. Unknown — usually a timeout or a server that won't respond; retry once, then discard
4. Invalid — hard bounce waiting to happen

Only the first category goes into your main sequence. Risky goes into a separate low-volume test. Unknown and Invalid get dropped—don't "try them anyway."

Everything I'd read said verification was a one-time pre-import step. In practice, I found that re-verifying the same list after 30 days catches a meaningful chunk of contacts whose status has changed—job moves, domain changes, catch-all server updates. Our Q1 2024 audit showed roughly 8-12% of previously deliverable rows shifted category within a quarter. Not a huge number, but enough to matter if you're sending at volume.

Verification inside an agent-native workflow is different from verification as a pre-flight check. The agent re-verifies on a rolling schedule, flags drift, and routes the changed rows for re-segmentation. That's the feature you actually want.

Step 4: Layer Intent Data on Top of the Verified List

Not every deliverable contact deserves the same sequence. Intent signals tell you who's in-market this week.

Priority tiers, roughly in order of reliability:

• Direct intent (job posting for a role you sell to, visiting your pricing page)
• Adjacent intent (competitor research, review-site comparisons)
• Broad intent (industry news, funding announcement, leadership change)

Route tier 1 into a short, high-touch sequence. Tier 2 goes into your standard cadence. Tier 3 is nurture.

The mistake I see most often: intent signals used to justify more sends rather than better targeting. Intent doesn't make a weak offer work. It just makes a strong offer land at the right moment.

Step 5: Write OKKI Go Sequences in Natural Language

This is where OKKI Go's natural language prospecting actually earns its place. Instead of building a sequence from a dropdown tree, you describe the sequence in plain English.

Example of what you type:

"Write a 4-step sequence for RevOps managers at Series B SaaS companies who visited our pricing page in the last 14 days. Tone: peer-to-peer, no fluff. First email references the pricing page visit without being creepy. Each follow-up adds a new piece of value—case study, benchmark, or a specific question. No exclamation points."

The agent produces a draft. You edit. You don't run it raw.

After 5 years of running outbound for teams of every size, I've come to believe that the quality of your prompt matters more than the quality of your sequence template. A generic template with a sharp prompt beats a clever template with a vague one, every time.

Two rules for prompts that actually work:

• Name the objection you expect ("they'll say they already have a tool—address it directly in step 3")
• Cap the length ("every email under 90 words"). Without a cap, you get paragraphs.

Step 6: Cover the LinkedIn Side with the Un-Enriched Bucket

Remember those contacts from step 2 who couldn't be enriched with an email? They're not dead leads—they're LinkedIn prospects.

LinkedIn prospecting complements the email side, and it doesn't require an email address at all:

1. Connection request with no pitch
2. Wait 3-5 days
3. Message referencing something specific (a post, a role change, a shared group)
4. If they respond, move them back into the email track using an email scraped from the profile (with verification, obviously)

Keep LinkedIn and email streams separate in your CRM—same person, two touchpoints, one timeline. OKKI Go's CRM enrichment syncs both so you don't double-touch someone on the same day through two channels.

Step 7: Insert a Human-in-the-Loop Review Before Any Send

This is non-negotiable for the first 30 days of any new sequence.

The agent drafts. A human reads the first 20 sends. Every single one. If the personalization is off, if the tone drifts, if the offer doesn't land—you catch it here, not in a reply-all disaster.

After 30 days of clean runs, you can loosen this. Maybe spot-check 10% of sends. Maybe only review the highest-value accounts. But the gate exists. Agent-native doesn't mean unsupervised—it means the agent does the volume work and the human owns the judgment calls.

It took me about three years—and I'd guess 400+ sequence reviews—to understand that the review step isn't about catching errors. It's about the person doing the review understanding why the agent wrote what it wrote. That's how you learn to write better prompts.

Step 8: Sync Back to CRM and Audit Weekly

Every reply, every meeting booked, every unsub, every bounce—it all needs to land in your CRM with the original source attribute (which list, which intent tier, which sequence).

Weekly audit, 20 minutes, four questions:

• What's our bounce rate this week? Anything above 2% needs a re-verification pass.
• Which intent tier produced the most meetings? Adjust routing.
• Which step in the sequence has the steepest drop-off? That's your rewrite target.
• Any contact touched more than 6 times in 30 days across both channels? Pull them out.

Things That Go Wrong (And What to Do)

You skipped the ICP exclusion list. Sequences start hitting competitors, current customers, and accounts in active deals. Fix the top of the funnel first, then re-verify and re-send.

You ran enrichment and verification in the wrong order. Verification before enrichment means you're verifying rows you're about to replace. Always enrich first, verify second.

Your intent tier 1 list is too big. If "high intent" is 40% of your list, your criteria aren't selective enough. Tighten it until tier 1 is under 10%.

You let the agent send without review. You'll find out what went wrong in a Slack message from a prospect who saw your email. Restore the review gate for two weeks, then reassess.

You treated verification as one-and-done. A verified list goes stale. Build a rolling re-verification schedule—30 days is a reasonable default—and let the workflow flag drift.

That's the whole loop. Eight steps, and the two that matter most are the ones most teams rush: enrichment before verification, and human review before scale. Get those right and the rest of the workflow—natural language sequences, LinkedIn side, intent routing—does what it's supposed to do.

Pricing, coverage, and product capabilities change; verify current details at okkigo.com before finalizing your setup.