PartnersWatch AI

AI that sets up, runs and explains your partner program

Answer a few questions and get a working program draft. Then let AI score applicants, draft the follow-ups, run workflow steps and model commission changes, while you approve anything that matters. On your own model key.

Design my program Talk to sales

OpenAIAnthropicGeminiMCP

partnerswatch.com/admin/setup/design

Program Designer · review draftDraft

From your answers: B2B SaaS · resellers + referral · monthly billing

Partner groups

Referral partners, Resellers

Tier ladder

Silver → Gold → Platinum by won volume

Commission rules

20% recurring 12 mo · $500 spiff · 90-day clawback

Mission path

Onboard → certify → first deal

Application questions

Region, customer count, services offered

EditApply 12 changes

What the AI does

Nine jobs, one model key

Seven are in the product today. Two are being built and are marked as such.

Setup

AI Program Designer

From your signup answers to a draft program: groups, tiers, commission rules, a mission path and application questions.

See it

Recruit

Applicant fit scoring

Every application scored 0–100 against your ideal partner, with strengths, risks and a suggested group.

See it

Everywhere

Agentic assistant

One chat over a manager and specialist roles that use your program's tools and pause for approval.

See it

Optimize

AI workflow steps

An ai_step node in any playbook that returns structured output the next step can use.

See it

Commission & pay

Commission simulator

Run a proposed rule over past deals and see the cost and who earns more or less.

See it

Developers

Integration MCP server

Coding agents connect your product to the program: guide, connections, self-tests, readiness.

See it

Coming soon

Recruit agent

Keeps a ranked prospect list from your ideal partner profile and drafts the invites.

Coming soon

Fraud & anomaly radar

Risk scores on conversions and partners, with commission held for review.

Optimize

Natural-language analytics

Ask a question in plain words and get a table, a chart and a one-line answer back, answered from an allow-listed set of metrics, never raw SQL.

01 · AI Program Designer

From signup to a running program

Tell us what you sell, who your partners are and how you bill. The Program Designer drafts the whole setup — partner groups, a tier ladder, commission rules with previewed payouts, an onboarding mission path and application questions — and shows it as a list of changes you apply, edit or drop.

✓Works for SaaS channels, agencies, referral programs and franchise networks
✓Commission rules checked against the same preview engine as live payouts
✓Nothing is created until you apply the draft

partnerswatch.com/admin/setup/design

Program Designer · review draftDraft

From your answers: B2B SaaS · resellers + referral · monthly billing

Partner groups

Referral partners, Resellers

Tier ladder

Silver → Gold → Platinum by won volume

Commission rules

20% recurring 12 mo · $500 spiff · 90-day clawback

Mission path

Onboard → certify → first deal

Application questions

Region, customer count, services offered

EditApply 12 changes

02 · Applicant fit scoring

Score every applicant the same way

Each application to your program is scored against your ideal partner profile. The score comes with the strengths and risks behind it, a suggested group and tier, and the questions worth asking before you decide.

✓0–100 fit score with the reasons attached
✓Suggested group and tier for the approval
✓Questions drafted from the gaps in the application

partnerswatch.com/admin/exchange/applicants/brightline

Brightline AgencyAI fit score

Strong fit

Audience match92

Region coverage80

Sales capacity74

Strengths

Serves your target segment
Existing services practice

Risks

Small team for two regions

Suggested group Agency partners · Silver. Ask: “Who would own onboarding for new customers?”

03 · Agentic assistant

Ask for the outcome, approve the action

The assistant is one chat over a small team: a manager breaks your goal into steps and hands each to a specialist role that uses the program's own tools. Anything that reaches outside the platform or cannot be undone pauses for your approval. Partners get the same assistant in their portal, limited to their own data.

✓Reads and acts through the same permissions you have
✓Every tool call shown as it runs
✓Outward and irreversible steps wait for approval

partnerswatch.com/admin/assistant

Assistant

Which Gold partners haven't registered a deal in 60 days? Draft them a nudge.

list_partners tier = Gold

list_deals since 60d

create_task 3 drafts

Three Gold partners are quiet: Harbor & Co, Quill Labs and Cedar Group. I drafted a nudge for each with their best-performing link.

Send 3 emails? Needs your approvalApprove

04 · AI workflow steps

AI as a step, not a side panel

Drop an ai_step into any playbook. Give it a prompt and the shape of the answer you want; the step returns structured output that the next node can branch on, write to a record or put in an email. Simulate the playbook on sample data before it runs.

✓Structured output validated against your schema
✓Classify, summarise, extract or draft inside a workflow
✓Runs on the workspace's own model key

playbook node · ai_step

{
  "type": "ai_step",
  "prompt": "Classify this request and draft a reply",
  "output": { "category": "string", "urgent": "boolean", "reply": "string" }
}

result → next node

{ "category": "co-op funds", "urgent": false, "reply": "Thanks — we…" }

05 · Commission simulator

Change a rate without guessing

Before you change a commission rule, run it over your past deals. See the extra payout cost, how many partners earn more or less, and a one-sentence summary of the change that you can send to partners as it is.

✓Uses the same engine as live commission
✓Who wins, who loses and what it costs
✓A plain-English summary for partners

partnerswatch.com/admin/commissions/simulate

Simulate: 20% → 25% for Goldlast 12 months

Payout cost

+$18.4k

Partners earning more

Earning less

Margin stays above your 60% target. Partners will read: “Gold partners now earn 25% of every subscription for 12 months.”

06 · Integration MCP server

Let a coding agent do the integration

Point Claude Code, Cursor or any MCP client at the integration server with an operator key. The agent reads the integration guide, creates test connections, runs self-tests and reports readiness for live, all scoped to your workspace and environment. A second MCP server exposes the program itself to AI tools.

✓Guide, connections, webhooks, provisioning and readiness tools
✓Scoped by key, permission and live or test environment
✓Guided prompts for integrate, verify, launch and troubleshoot

Developer docs

~/.cursor/mcp.json · or Claude Code, ChatGPT, any MCP client

{ "mcpServers": {
  "partneros-integration": {
    "url": "https://partnerswatch.com/api/mcp/integration",
    "headers": { "Authorization": "Bearer <operator key>" }
  }
} }

coding agent · session

Wire up provisioning

get_integration_guidepayload, auth, retries

create_connectiontest environment

run_integration_selftest4 / 4 checks pass

integration_readinessready for live

Being built

Next on the AI roadmap

Recruit agent

Keeps a ranked prospect list from your ideal partner profile and drafts the invites.

Coming soon

Fraud & anomaly radar

Risk scores on conversions and partners, with commission held for review.

Coming soon

Your model, your data

Bring your own key. Keep your own tenant.

Add an OpenAI, Anthropic or Gemini key and every AI feature in your workspace runs on it. Keys are encrypted and scoped to the workspace that added them. The usage log keeps the feature, model, token counts and latency for each call, so you can see what AI costs you, and never stores prompt or output text.

✓OpenAI, Anthropic and Gemini adapters
✓Per-workspace encrypted keys; no cross-tenant reuse
✓Usage by feature and model, without prompt text

partnerswatch.com/admin/settings/ai

AI providersYour keys

ProviderKeyStatus

Anthropicsk-ant-…4f2aDefault

OpenAIsk-…91cdActive

Google GeminiplatformFallback

AI usage · 30 days

assistant

62%

playbook.ai_step

24%

fit scoring

14%

Keys encrypted per workspace. Usage logs keep tokens, never prompts.

How we build AI features

Four rules every AI feature follows

Drafts, not decisions

AI output lands as a draft you review. Program changes are a diff you apply; outward actions wait for approval.

Your model, your bill

Bring an OpenAI, Anthropic or Gemini key. Keys are encrypted per workspace and only used for that workspace.

Scoped to the tenant

Tools run with the signed-in user's permissions inside their own workspace. Partners only ever reach their own data.

Shows its work

Every tool call is visible, and each model call is logged by feature, model, tokens and latency, never the prompt text.

Integrations

Models and tools AI works with

All integrations

ANAnthropicAIBring your own Anthropic key; AI features run on your account.OPOpenAIAIBring your own OpenAI key; AI features run on your account.GGGoogle GeminiAIBring your own Gemini key, or use the platform default.MSMCP serversAILet Claude, ChatGPT, Cursor or any MCP client work with your program and wire up your product.SWSigned webhooksiPaaS & customHMAC-signed webhooks in and out, with delivery logs, replay and secret rotation.RAREST API (generic)iPaaS & customAny REST or Bearer API, with field-mapped pushes and shared-secret inbound.

Ready to run your partner program in one place?

Book a walkthrough and see recruitment, PRM, automation, payouts, CX, and dashboards working together — on one platform.

Start free Talk to sales Sign in