Smith.ai for HubSpot: Lead Scoring and Routing Guide

If you’re capturing leads by phone or chat and pushing them into HubSpot, your next constraint isn’t volume - it’s quality and speed. Smith.ai’s integration can help you screen, score, and route leads the moment the conversation ends. The difference between a 3-minute and a 3-hour response is the difference between won and lost. This guide shows you how to design scoring, routing, and SLAs around the Smith.ai + HubSpot stack so sales gets signal, not noise.

Most teams obsess over top-of-funnel acquisition, then pour everything into one bucket called “New Lead.” That’s not a system; it’s a waiting room. If your inbound mix includes calls and live chat - where intent and context are highest - you can’t afford to treat every contact equally.

Lead scoring separates fit from fluff. Routing gets the right person to the right rep within minutes. Together, they protect your sales calendar from junk and give qualified buyers an experience that feels intentional, not reactive. If you’re using Smith.ai to capture chats and calls, you already have richer context than a form submit. Put it to work.

Done right, scoring and routing reduce manual triage, improve speed-to-first-touch, and lift conversion without adding headcount. Done poorly, they create work for ops and resentment for sales. This guide is the former.

Smith.ai focuses on front-line conversation capture - 24/7 chat and phone answering - with human receptionists assisted by AI. When integrated with HubSpot, those conversations can create or update contacts, attach transcripts/notes, and flag key answers from intake questions. That data can trigger workflows that score the lead and route it to a rep, a queue, or a follow-up sequence.

Practically, this looks like: new contact created from a Smith.ai interaction, mapped fields update (e.g., phone, email, company, notes), and one or more properties indicating call/chat outcomes. From there, HubSpot workflows handle the rest - scoring, assignments, notifications, and tasks. The app’s exact fields and options may vary, so confirm the latest property mapping in your install.

The headline: you get structured data and conversation context at the moment of capture. If you’re still manually forwarding emails from a shared inbox to sales, this is a step-change in responsiveness and consistency.

Scoring and routing only work when your data model is boring and predictable. Before installing anything, stabilize the basics in HubSpot. You want clear lifecycle stages, source taxonomy, and ownership logic so there’s no ambiguity when a new Smith.ai lead arrives.

Minimum viable data model components:

• Lifecycle: Subscriber → Lead → MQL → SQL → Opportunity → Customer (use your org’s version, but keep it linear).
• Source fields: Original Source, Latest Source, and a consistent subsource taxonomy - e.g., “Smithai - Chat,” “Smithai - Phone,” “Smithai - After-hours.”
• Fit vs. Intent properties: Create two numeric properties - Lead Fit Score and Lead Intent Score. Keep them separate so ops can tune independently.
• Routing properties: Lead Route Tier (A/B/C), Territory/Segment (e.g., EMEA, SMB/ENT), and a boolean like “Ready for Sales” to trigger owner assignment.
• Operational notes: Qualification Notes (text), Booking URL, Last Conversation Summary, and a checkbox for “Do Not Call/Email” if compliance flags arise.

If your team hasn’t formalized pipeline structure or property governance, start there. It’s hard to bolt advanced routing onto a shaky CRM foundation. If needed, bring in help to standardize your CRM and HubSpot pipeline before you automate complexity.

Lead scoring fails when it becomes an academic exercise. Your model should be simple enough to explain on a whiteboard and precise enough to separate tire-kickers from buyers. Two parallel scores - Fit and Intent - keep logic clean and tuning faster.

Fit measures who they are: company size, industry, role, territory, tech stack, and disqualifiers. Intent measures what they do: requesting a quote, mentioning budget/timeline, visiting pricing pages, or calling after-hours for urgent needs. Add both scores to get a composite, but keep the components visible.

Use bands, not brittle thresholds. For example: A (80+), B (60–79), C (40–59), D (<40). Route A and B to humans fast; send C and D to nurture or lower-priority queues. Update bands over time as conversion data accumulates.

Fit scoring: ICP signals worth tracking

Start with your actual customers. What do your won deals have in common? Mirror those traits in your Fit score. Common positive signals include headcount (or revenue) in your target range, ICP industries, buyer roles with authority, and in-territory location. Negative signals might be consumer emails for a B2B product, out-of-territory regions, or mismatched industries.

Assign points that reflect real revenue likelihood, not wishful thinking. If you close zero deals below 10 employees, that’s a heavy negative. If regulated industries require expensive implementation, maybe they’re neutral until intent is high. Keep the score mapping in a shared doc and set a quarterly review cadence.

Smith.ai conversations often reveal job titles, company details, and use cases upfront - gold for Fit scoring. Make sure those fields map cleanly into HubSpot properties you can reference in workflows.

Behavioral scoring: use intent they actually expressed

Behavioral signals beat vague “engagement.” Treat a phone call asking for pricing as a stronger signal than a webinar view. From a Smith.ai chat or call, capture specific intents such as “requested quote,” “asked for demo,” “urgent issue,” or “timeline < 30 days.” Allocate points accordingly.

Complement conversation signals with web analytics: viewed pricing page (+10), viewed case studies (+6), returned within 24 hours (+4), clicked calendar link (+12), submitted payment inquiry (+15). Avoid double-counting (e.g., chat + call on the same day) unless the buyer changed behavior meaningfully.

Weight recency. A high-intent chat from this morning should outweigh a case study view from last month. Use engagement time windows (7 days, 14 days) to decay older points automatically.

Decay and negative scoring to keep the funnel honest

Every scoring model needs gravity; otherwise, old activity props up non-buyers. Implement a decay function: subtract 10% per week of inactivity from Intent, and 5% per month from Fit if disqualifying info appears.

Negative scoring should be explicit: student emails, competitors, spam indicators, or markets you don’t serve. It’s better to move a contact out of sales’ way than to pretend nurture will transform them.

Document what earns negatives and why. This isn’t a blacklist; it’s resource allocation. As long as the rules are transparent, sales will trust the output.

Great routing feels boring. The right owner appears, tasks are created, and the buyer gets a call back quickly. Aim for rules that are stable, testable, and minimal. Start with a route tier derived from score bands, then apply territory or specialization logic.

Example approach: A/B leads with territory EMEA go to Team EMEA round-robin; AMER go to Team AMER. C leads go to an SDR queue with a lower SLA. D leads skip owner assignment and go to nurture. VIP flags (enterprise size + urgent intent) route to a senior queue with on-call coverage.

Make after-hours behavior explicit. If Smith.ai captures a high-intent call at 9 pm local time, do you wake someone up, or book the soonest morning slot automatically? Decide the policy once, enforce it in workflows, and communicate it to the team.

This sequence gets you from installation to reliable scoring/routing. Adapt names to your taxonomy, and confirm current field mappings in your Smith.ai integration.

1) Install and authenticate: Add the Smith.ai app from HubSpot’s marketplace. Authenticate your account so new conversations can create/update HubSpot records.

2) Property mapping: Map conversation attributes to HubSpot contact properties - source subtype, call/chat disposition, qualification notes, and any custom intake questions. Keep names human-readable and consistent.

3) Create score properties: Add “Lead Fit Score” (number) and “Lead Intent Score” (number). Add “Lead Route Tier” (dropdown: A/B/C/D). Add “Ready for Sales” (checkbox) and “SLA Start Time” (datetime).

4) Scoring workflows: Build two workflows - one calculates Fit (based on profile data and disqualifiers), the other calculates Intent (based on call/chat outcomes and web behavior). On update, both write to the composite route tier. Keep workflows modular so you can tune pieces without redeploying everything.

5) Routing workflows: Trigger on “Ready for Sales is true” AND “Lead Route Tier is A or B.” Branch by territory/segment. Use team rotation or round-robin for fairness. For C leads, assign to SDR queue; for D leads, enter nurture only. Create tasks automatically with due dates aligned to SLA.

6) Notifications and calendars: Fire Slack/email alerts for A/B leads, include the conversation summary and next-step CTA (e.g., link to the rep’s calendar). For after-hours policies, either assign to on-call or create an automated calendar booking link response.

7) SLA tracking: On lead creation or tier upgrade, stamp SLA Start Time. Use a workflow to measure time-to-first-touch and mark “SLA Met” or “SLA Missed.” Surface this on dashboards by team and rep.

8) QA and shadow mode: Run routing in shadow for a week - write the hypothetical owner to a “Proposed Owner” property without reassigning. Compare to sales’ expectations, fix edge cases, then go live.

9) Post-launch monitoring: Review 20–30 routed leads with the team weekly for the first month. Tune negative scoring and territory exceptions first; avoid adding niche rules unless you see recurring patterns.

If you’d rather have an operator design the workflows end-to-end, this lives squarely in marketing automation workflows - connectors, properties, routing, SLAs, and reporting.

Response speed is a competitive moat. Write SLAs that anyone can understand and that a workflow can enforce. Separate SLAs by route tier so you don’t hold C leads to A-level standards.

Reference benchmarks by channel. For a high-intent phone call, target under 5 minutes during business hours. For live chat escalations, under 10 minutes. After-hours, set a clear opening window (e.g., by 9:30 am local). Document exceptions and make “missed SLA” an objective metric on dashboards.

Finally, align capacity to your SLA. If your inbound mix surges on Mondays, don’t set an SLA you can’t keep. Better to route overflow to a backup team or auto-book a call than to overpromise and let hot leads cool.

Don’t drown in vanity metrics. Track the five that connect scoring/routing to revenue. Build one dashboard per team and review weekly.

1) Speed to first touch: Median and 90th percentile, by tier and by channel (chat vs. phone). If the tail is long, you’re missing buyers even if the median looks good.

2) MQL→SQL conversion: By route tier and by Smith.ai subsource. If Tier A isn’t converting 2–3× Tier C, your scoring is off or the handoff is broken.

3) Meetings booked from Smith.ai: Count and rate - meetings that came directly from chat/call interactions. Include no-show rate; it reveals qualification gaps.

4) Pipeline created: Opportunities and value initiated within 14 days of first touch, by tier. This is the sanity check for whether your model maps to money.

5) SLA adherence: % met by team/rep, plus outcomes on missed-SLA leads. If missed SLAs correlate with lost deals, it’s a staffing or routing problem, not a sales “motivation” problem.

Over-scoring engagement: Not all clicks matter. Weight conversation intent higher than content consumption. One phone call asking for pricing beats three blog views every time.

Routing spaghetti: Too many exceptions create unpredictability. Start with territory and tier. Only add special-case rules after you’ve seen the same edge case at least 10 times.

Owner churn: Reassigning owners mid-cycle kills accountability. Freeze ownership after the first live connect unless territory is wrong. Use a “Proposed Owner” field for suggested changes without ripping accounts away.

Missing UTMs and source taxonomy: If everything shows up as “Direct” or “Chat,” you can’t optimize. Enforce a consistent subsource like “Smithai - Chat” vs. “Smithai - Phone” and include UTMs on links sent during conversations.

Time zone blind spots: If you serve multiple regions, after-hours in one territory is prime time in another. Route by local time or set an on-call rotation.

If your RevOps backlog keeps growing - properties to clean, workflows to rebuild, reports to fix - you don’t need another presentation. You need a senior builder who sits in your CRM, ships the scoring model, tunes routing, and proves it with conversion lifts. That’s the operator model: one accountable owner for strategy and execution across HubSpot, content, paid, and automation. Learn more about our approach on the about page.

Proof beats promises. On Cleverence (B2B SaaS), we grew from 0 to 80K organic visits/month and 0 to 2,000 inbound leads/month while driving a 16% lead-to-purchase rate. Those outcomes happen when scoring, routing, and content play well together - not because a slide said they would.

You don’t have to use a receptionist/chat platform to route leads - but it helps when conversations create better intent data. Broadly, you have three paths. Native-only in HubSpot: use built-in lead scoring and assignment rules. Pros: fewer moving parts, one data model. Cons: less conversational context, fewer after-hours capture options.

Receptionist/chat platforms integrated to HubSpot: capture richer signals in real time (what the buyer asked, urgency, budget hints). Pros: higher intent clarity and immediate workflows. Cons: requires clean mapping and discipline in ops.

DIY patchwork: forms, manual emails, spreadsheets. Pros: no new vendor. Cons: slow, error-prone, invisible SLAs. If you’re spending real money on acquisition or if sales capacity is tight, DIY quickly becomes the most expensive option.

Week 1–2: Data model cleanup, property creation, install/configure Smith.ai integration, map fields, draft scoring rules. Shadow-mode routing in place by the end of Week 2.

Week 3–4: Live routing for Tier A/B, sales feedback loop, SLA tracking, dashboard v1. First conversion lifts appear purely from faster response and clearer ownership.

Week 5–8: Tune negative scoring, refine territory exceptions, add meeting automation, and scale to more sources. By now, your MQL→SQL rates should stabilize and tell you where to double down.

Quarterly review, always. Look at the last 90 days of A/B leads: which traits predicted revenue? Adjust Fit scores accordingly. Which conversation intents turned into pipeline? Adjust Intent weights and chat/call scripts.

Audit routing metrics monthly. Are SLAs consistently met? Are some reps overloaded while others idle? Tweak round-robin pools or add a backup queue for spikes. Keep changes small and testable - avoid overhauls unless the business changed materially.

Finally, document everything. A simple runbook with score mappings, routing diagrams, and SLA rules prevents “tribal knowledge” drift and speeds onboarding.

If you’re already taking calls and chats via Smith.ai, you own the fastest intent signal in your funnel. The win comes from turning that signal into routing and SLAs your team can execute daily. Stand up a clean data model, split Fit and Intent scoring, keep routing boring and fair, and measure speed-to-first-touch like a core KPI. If you want a second set of senior hands to build it end-to-end, book a free Growth Audit and we’ll show you the fastest path to predictable response and conversion.

Article by Anton Chisliuk · AI Marketing Ops - leads, sales, content, and systems, one senior operator.

- Do I need Smith.ai for lead scoring in HubSpot?

No. HubSpot can score and route natively. Smith.ai adds higher-quality, real-time intent from calls/chats, which usually improves accuracy and response speed. If you already get phone/chat inquiries, the integration is worth evaluating.

- What properties should I create to support scoring and routing?

Create Lead Fit Score (number), Lead Intent Score (number), Lead Route Tier (dropdown), Ready for Sales (checkbox), SLA Start Time (datetime), and Qualification Notes (text). Keep names consistent and documented.

- How fast should we respond to Smith.ai leads?

For high-intent chats/calls, target under 5–10 minutes during business hours. After-hours, set a clear morning follow-up window (e.g., by 9:30 am local) or use an on-call rotation for Tier A leads.

- Can I route leads by territory and still use round-robin?

Yes. Use a two-step workflow: first branch by territory/segment, then apply round-robin within the team. This balances fairness with specialization and keeps rules transparent.

- How do I prevent junk leads from reaching sales?

Use negative scoring (student emails, competitor domains, mismatched industries) and a minimum route tier threshold. Route low-score contacts to nurture and only promote them if Fit or Intent improves.

- What reports show whether scoring works?

Track MQL→SQL conversion by route tier, speed-to-first-touch by tier and channel, meetings booked from chat/call, pipeline created within 14 days, and SLA adherence. If A/B tiers don’t materially outperform C/D, revisit your weights.

- How do we handle after-hours high-intent calls?

Decide a policy upfront: on-call routing for Tier A, or auto-book first morning slots. Encode it in workflows and communicate it to reps. Consistency beats ad hoc heroics.

- Will complex routing slow down HubSpot?

It can. Keep workflows modular, avoid nested branches beyond two levels, and prefer clear tiers over many one-off exceptions. Test in shadow mode and monitor execution times after launch.