TL;DR: AI lead-routing builds split into three tiers. Tier 1 (basic routing, no LLM, no enrichment) costs $1–3k to build and $0–50/month to run. Tier 2 (enriched routing with LLM scoring, Clearbit or Apollo, ICP fit reasoning) costs $3–7k build and $100–400/month. Tier 3 (full ops automation: drafted reply, pre-call brief, calendar match, follow-up kickoff) costs $7–15k build and $300–1,000/month. The three things that move price: enrichment depth, LLM intensity, and how much human review stays in the loop. Most studios stop at Tier 4 of the workflow. The ROI lives at Tier 6.
A lead-routing automation is one of the highest-ROI AI builds for a service business. It's also one of the most consistently mis-quoted, because the cost depends on three things people forget to scope: enrichment, LLM intensity, and how much human review you keep in the loop.
Here are real numbers from automations we've shipped in the last year.
What "lead routing" actually means
A typical lead-routing automation does some subset of these:
- Capture. Form fill, email, or LinkedIn message comes in.
- Enrich. Look up the company, person, role, and recent activity.
- Score. Compare against ICP rules. Rate fit and intent.
- Route. Send to the right rep, channel, or queue.
- Draft. Pre-write a personalized response or first reply.
- Brief. Generate a pre-call summary for the rep.
The cheapest automations stop at step 4. The most expensive go through 6. The right scope depends on what your team actually needs.
The three real cost tiers
Tier 1: Basic routing — capture, light scoring, route to a Slack channel or HubSpot owner. No LLM, no enrichment.
- Build: $1k–$3k
- Run: $0–$50/mo
- What it does: Pulls form fills, applies a rule-based score, routes.
Tier 2: Enriched routing with LLM scoring — capture, enrichment via Clearbit/Apollo, LLM-based fit assessment, route.
- Build: $3k–$7k
- Run: $100–$400/mo (mostly enrichment API + LLM)
- What it does: Tier 1 plus deep ICP fit scoring, with reasoning the rep can read.
Tier 3: Full ops automation — Tier 2 plus drafted reply, pre-call brief, calendar match, follow-up sequence kickoff.
- Build: $6k–$15k
- Run: $200–$800/mo
- What it does: A nearly hands-off lead pipeline from form fill to first call.
These aren't priced on a single ladder. Most clients land at Tier 2.
Build cost breakdown
For a Tier 2 automation, here's where the build hours go.
| Phase | Hours | Cost |
|---|---|---|
| Discovery, ICP scoring rules, integrations map | 6–8 | $600–$800 |
| Form / email capture and webhook routing | 4–6 | $400–$600 |
| Enrichment integration (Clearbit, Apollo, or free fallback) | 4–8 | $400–$800 |
| LLM scoring with structured output | 8–12 | $800–$1,200 |
| HubSpot/Slack/CRM write-out | 4–8 | $400–$800 |
| Retries, error handling, observability | 4–6 | $400–$600 |
| Internal dashboard or admin view | 4–8 | $400–$800 |
| Testing, edge cases, evaluation | 4–6 | $400–$600 |
| Documentation, handoff, Loom | 2–4 | $200–$400 |
Total: 40–66 hours, $4k–$6.5k. Inside Tier 2's quoted range.
A studio quoting Tier 2 and spending 100+ hours is losing money. A studio quoting Tier 2 and spending 20 hours is shipping something that will break in month two.
Run cost breakdown
Monthly running cost for Tier 2, at typical small-business volume (300–800 leads/mo):
| Service | Cost |
|---|---|
| Hosting (Vercel + a small worker, or n8n Cloud Starter) | $20–$50 |
| Enrichment (Clearbit, Apollo, or free Hunter alternatives) | $50–$300 |
| LLM API (OpenAI or Anthropic) | $20–$80 |
| Vector DB (only if using RAG for ICP definitions) | $0–$50 |
| Monitoring | $0–$30 |
| Total | $90–$510/mo |
The biggest variable is enrichment. Clearbit at scale is $1,400+/mo for full features. Apollo runs $99–$199/mo for typical small-business volume. Free alternatives exist (Hunter free tier, manual fallback) but the data quality drops.
Where the LLM bill actually lands
The LLM cost is smaller than people think for routing-class workloads.
A typical Tier 2 automation makes 1–3 LLM calls per lead. At GPT-4o or Claude Sonnet 4 prices in 2026, that's roughly $0.01–$0.03 per lead. At 500 leads/mo, that's $5–$15. At 5,000 leads/mo, $50–$150.
Compare to enrichment: Clearbit at $1.50/lookup × 5,000 = $7,500/mo. Enrichment is the line item to negotiate.
When this pays back
The math we see most often.
A B2B service business gets 400 inbound leads/month. A rep spends 8 minutes on each one (read, qualify, draft a first reply). That's 53 hours/month, or $5,300/month at a $100/hr blended rate.
A Tier 2 automation cuts this to 1.5 minutes per lead (read the brief, send the pre-drafted reply, book or pass). That's 10 hours/month, or $1,000.
Savings: $4,300/month. Build pays back in 1–2 months. Run cost is well under the savings.
This is the unit economics for typical service businesses. It gets better at higher lead volume and worse at lower.
When this doesn't pay back
Three cases.
1. Lead volume under 100/month. The savings don't justify the build. Stay manual or use Zapier.
2. Leads are already pre-qualified. If your form is short and 80% of leads are already qualified, the routing automation has nothing to add. Spend the money elsewhere.
3. Routing rules change weekly. A custom build is fixed code. If your team is going to want to tweak ICP rules every Friday, build it on Make or n8n with a UI editor instead.
What we ship by default
A Tier 2 build for a typical client looks like this:
- Stack: A Next.js API route or n8n Cloud workflow. Webhook from form, calls enrichment, calls LLM, writes to CRM, posts to Slack.
- Enrichment: Apollo or Clearbit if budget allows. Hunter + LinkedIn enrichment via a third party if not.
- LLM: Claude Sonnet 4 or GPT-4o for scoring. Cheaper models for drafting.
- CRM: HubSpot or Pipedrive write-out via their API. Source attribution baked in.
- Slack: A formatted message with score, reasoning, and a "draft reply" attached.
- Monitoring: Sentry + a simple dashboard showing run rate, success rate, scoring distribution.
We document the workflow, hand over the keys, and check in at 30 days post-launch.
Three questions to decide this week
- What's your monthly lead volume, and how much rep time per lead? Multiply. If the answer is under 20 hours/month, this isn't your bottleneck.
- Do you have an explicit ICP definition? If no, the automation can't score against it. Define ICP first.
- Will routing rules change weekly? If yes, build on Make or n8n. If they're stable, custom is the right call.
If you want a scoped quote for a lead-routing automation tuned to your specific ICP and volume, send us the form fields you collect, your CRM, and a rough lead-volume number. We'll come back with a tiered proposal in two days.