The default outbound stack for a RevOps team running AI-augmented prospecting in 2026: list-build, enrichment, ICP scoring, sequence dispatch, and CRM sync, with AI inside every layer instead of bolted on at the end. The premise is not that any one tool wins in isolation. It is that a 5-to-15-rep team can send fewer, better-researched messages and still fill pipeline, because the research each message needs now runs in software rather than in an SDR’s browser tabs.
Two things moved this year and both change how you build it. Outreach rebuilt around agents and moved to outreach.ai, retiring its Engage seat plans for AI-credit-metered Amplify tiers. And every layer here now publishes an MCP server, which changes what the glue between them has to do.
How the pieces fit
- Clay is the data and AI workbench. ICP list-building from any seed (titles, technographics, intent signals), waterfall enrichment across 150+ data partners, and AI columns that summarize an account or draft an opener per row. Clay is where the list takes shape. Billing runs on two meters you size independently — Actions for platform work, Data Credits for third-party data — so the headline price is the sum of two sliders. Launch starts at $167/mo billed annually ($54 for 180,000 actions/year plus $113 for 30,000 data credits/year); Growth starts at $446/mo for 480,000 actions and 72,000 credits. The lead-enrichment Skill and the ICP list-builder are the two plays that live here.
- Apollo supplies cheap base data, mobile numbers, and the SMB sequencer. Professional is $79/seat/mo billed annually and grants 48,000 credits per seat per year upfront; Basic is $49 and Organization $119 with a three-seat minimum. For a sub-$10M ARR team Apollo is both the data floor and the send engine — you don’t add Outreach on top. Website-visitor identification and the parallel dialer are separate add-ons at $119/team/mo each.
- Outreach is the send engine for the mid-market and up. Branching sequence logic, A/B on send times, deep Salesforce coupling, and the audit trail a 25-plus-rep org needs. Its Spring 2026 release added Omni, a conversational layer over pipeline and accounts, and Agent Studio for building custom agents from workflow templates. Pricing is now four Amplify tiers named by AI-credit allowance — Essentials 10,000, Core 25,000, Plus 50,000, Pro 100,000 — combining a seat price with a consumption meter. There is no published figure; the page commits only to “per user pricing, no platform fees.”
- Claude writes the personalized first-touch. Each prospect’s opener is generated from the Clay-enriched row — the account’s hiring signal, funding event, or role context — so the send references something real instead of a merge tag. Claude runs as a Clay column or through n8n. Team seats are $20/seat/mo billed annually ($25 monthly) for teams of 2 to 150.
- n8n is the integration layer. A new lead in HubSpot fires a webhook → Clay enriches → the scored row writes back → the sequencer enrolls it → a hot reply routes to Slack. Every cross-tool hop runs through n8n rather than a hard-wired connector. Pricing is quoted in euros: Starter €20/mo billed annually for 2,500 executions, Pro €50 for 10,000, Business €667 for 40,000 with self-hosted deployment included. Community Edition is still free to self-host.
- HubSpot is the CRM source of truth. Every enriched contact, sequence touch, and reply lands back on the HubSpot record for pipeline reporting and forecast. Sales Hub Professional is $90/seat/mo billed annually ($100 monthly), carries 3,000 HubSpot Credits, and adds a one-time $1,500 onboarding fee. Operations Hub is now sold as Data Hub, with Professional at $720/mo annually.
The handoffs that make it a stack
The value is the chain, not the six logos. A form fill or a list import hits an n8n webhook, which fires a Clay run that waterfalls the contact, scores it against your ICP rubric, and has Claude draft the opener. Then n8n writes the finished row into HubSpot and enrolls only the rows above the score threshold into Apollo or Outreach. When a prospect replies, n8n routes the positive ones to Slack for a human and syncs the disposition back to HubSpot. One event, four tools, no human until the reply. Each arrow is a trigger that fires the next action, which is why the stack compounds instead of just co-existing.
Every layer now answers an agent
Clay runs a hosted MCP server at https://api.clay.com/v3/mcp, Apollo ships Apollo MCP, HubSpot and n8n both have servers, and Outreach went furthest — a server, a client, and a revenue-orchestration app inside ChatGPT, announced in June 2026. That makes a second path into the same data: ask a chat client which accounts replied this week, or have it enrich a shortlist, without opening six tabs.
It does not replace the pipeline. n8n still owns anything that runs on a schedule, retries on failure, or has to leave an audit trail, and an agent that silently skips a step is worse than a workflow that fails loudly. Use MCP for the ad-hoc questions and the exploratory pulls; keep the enrolment path in n8n.
Why this combination
Pre-AI outbound was Apollo + Outreach + Salesforce: a sequence-and-spray model that used volume to cover for weak personalization. The AI layer inverts it. Clay and Claude move the per-account research that used to eat an SDR’s morning into a table that runs at list scale, so the team sends fewer messages with account-specific context in each. Reply rate is the metric to watch — treat any multiplier a vendor quotes as unproven until your own sends move it. The combination keeps six tools because each does one job well; folding it into a single all-in-one suite (Apollo’s full stack, or Outreach’s) gives up the workbench depth that only Clay provides.
What it costs
The stack is bimodal, and the sequencer you pick sets the tier.
SMB / Apollo tier (sub-$10M ARR, 5-15 reps): budget roughly $33K-$52K/year all-in. For an 8-rep team: HubSpot Sales Hub Professional at $90/seat/mo is $8,640/yr, Data Hub Professional another $8,640, Apollo Professional $7,584, Clay Growth $5,352, Claude Team $1,920, and n8n Pro about €600. That lands near $33K, plus HubSpot’s one-time $1,500 onboarding fee. Fifteen reps pushes it past $50K, mostly through CRM and Apollo seats.
Mid-market / Outreach tier (25+ reps): budget $100K-$200K+/year. Outreach no longer publishes a seat price, so this range is triangulated, not quoted: Vendr data puts a 50-user Amplify Core deployment near $72,000/year, and 2026 pricing trackers land Core around $100-120/seat/mo with Plus at $120-160. At 25 reps that is roughly $36K-$48K before AI credits are consumed. Add CRM seats for the same headcount, a Clay Enterprise contract, and implementation work that the same trackers put at $5,000-$25,000, and the total clears six figures. Note that the old “$2,000-5,000/year platform fee” line no longer applies — Outreach’s own page now rules platform fees out.
The hidden costs sit in four meters and one headcount line. Clay’s two meters expand separately, so buying more actions does not buy more data credits. Apollo grants its credits per seat per year rather than monthly, which front-loads them. Outreach’s AI-credit consumption is the layer nobody forecasts correctly in year one. HubSpot’s credit allowances behave the same way, and n8n’s euro denomination adds an FX line for a US buyer. Model the two Clay meters against real list sizes before you commit — build a 20-row test table, read the per-run action and credit cost in the log, then extrapolate. That is the line that surprises teams.
When this stack is the right pick — and when it isn’t
Pick it when outbound is a primary channel, you have someone who will own the n8n and Clay build, and your ICP is specific enough that per-account personalization moves reply rates. That is the center of gravity: a $5-50M ARR B2B SaaS team that has outgrown spray-and-pray but is not ready to hand outbound to a fully-autonomous agent.
Skip it in three cases. If you are inbound-led, most of this sits idle — drop Apollo and Outreach, and route with a meeting-router like Chili Piper or a chat layer like Intercom instead. If nobody will own the integration and enrichment build, the “AI in every layer” premise collapses into six disconnected logins, and an all-in-one serves you better. And if you are sub-$10M ARR but eyeing Outreach for its brand, don’t — its per-seat-plus-credits model is priced for a team that already has an ops function; stay on Apollo’s sequencer until headcount actually crosses 25 reps.
Common variations
- Account-based motion. Add 6sense or Demandbase for intent and account scoring, and feed those signals into the Clay list-build; the ABM stack is the right reference. Swap when marketing and sales run named-account plays rather than volume outbound.
- Enterprise / strategic outbound. Replace HubSpot with Salesforce and add Gong or Chorus for conversation intelligence on booked meetings. Swap when deal complexity and coaching scale justify the price — not before.
- Fully-autonomous lower tier. Hand the lowest-value segment to an AI SDR and reserve human reps for the upper-mid-market; the AI SDR stack is the dedicated build. Note that Regie AI now sells RegieOne as a replacement for the sequencer layer rather than a bolt-on beside it, so this swap removes Apollo or Outreach instead of sitting next to them. Swap when volume outgrows what Clay-into-a-sequencer handles by hand.
- AI-native CRM base. If you are starting CRM fresh and want an API-first substrate instead of HubSpot, see the AI-native CRM stack — same Clay + Claude + n8n spine, different system of record.
What this stack does NOT replace
- An actual ICP rubric — see ICP and signal-based selling; the tools execute the rubric, they don’t write it.
- A meeting platform (Chili Piper, Calendly) to convert positive replies into booked time.
- A conversation-intelligence tool (Gong, Chorus) for the post-booking sales motion.
- A demand-generation engine — paid, content, partnerships — that feeds the Clay list-build inputs in the first place.
- The judgment call on whether to go autonomous — read what an AI SDR actually does before you replace reps with agents.