Early access — building with design partners

Personalised ML for every subscriber, run by an agent.

SubCore Agent trains models on your warehouse data and picks the right treatment for each subscriber — replacing static rules and one-size-fits-all campaigns.

subcore-agent · rollout brief preview
approved
brief.mdReady for review
DecisionPersonalized free LLM API usage cap
CohortTrial users >80% cap in first 14 days
Projected uplift+4.2pp trial-to-paid
GuardrailMAU change ≤ -1.5%
See how the agent got here
treatment assignment108,690 users
Max cap31,488
highest free allowance
Lenient cap37,785
moderate ceiling
Strict cap28,548
tighter ceiling
holdout: 10% control · randomized cap
Vision

Personalised ML for every subscriber decision, run end to end by one agent.

Most subscription teams still drive growth, monetisation, retention, and conversion with static rules and one-size-fits-all campaigns — because building a per-user ML loop on top of the warehouse is slow, manual, and fragmented across SQL, notebooks, dashboards, and review meetings.

SubCore Agent trains models on your warehouse data and uses them to personalise the next action for every user — which offer, which nudge, which paywall, or no action at all. It discovers and joins the tables that matter, builds features, fits uplift and propensity models, assigns treatments under your guardrails, simulates the impact, and packages a rollout brief for your team. After launch, live outcomes feed back into the next loop.

1Discover
Agent maps your warehouse with your access
2Prepare
Joins facts and dimensions into a cohort table
3Train
Builds features and fits uplift / propensity models
4Personalise
Picks the best treatment per user under guardrails
5Simulate
Rolls assignments up into expected impact
6Learn
Live outcomes feed back into the next loop
Problem

The gap isn't the model. It's the decision.

Most subscription teams already sit on the data needed to answer their next growth or retention question. Some have models on top of it, some don't. Either way, the path from data to a decision someone is willing to ship is still slow, manual, and fragmented.

What teams already have
  • Subscription events in a warehouse or lake
  • Billing, plan, and usage tables
  • A dashboard or two, plus past experiments
  • Maybe a churn or pLTV model. Maybe not.
  • A growth, retention, or pricing question waiting on an answer
What's still missing
  • Which users get which treatment
  • Which cohorts to exclude
  • Business constraints and guardrails
  • Expected revenue, retention, or LTV impact
  • A rollout decision someone can sign off on

A churn score, a conversion rate, or a dashboard chart is not a decision. Someone still has to choose which users get which treatment, which cohorts to leave alone, what the expected revenue or retention impact is, and whether the plan is safe enough to roll out.

SubCore Agent works on the decision layer — turning the data you already have, with or without existing models, into a policy your team can review and ship.

Agent loop

Watch the agent think, ask, and personalise a token cap.

Runs in your perimeter on your warehouse credentials. Packages a rollout brief for your review before anything ships.

subcore-agent· Exploring your warehouse with your access
Looped demo
Discover
artifact · discoverLive
snowflake://prod · role: analyst_ro · auto-inferred fact dimension
dim_usersdim
  • user_id PK
  • billing_country
  • signup_ts
  • platform
dim_plansdim
  • plan_id PK
  • tier
  • price_usd
  • interval
fact_subscriptionsfact
  • sub_id PK
  • user_id FK
  • plan_id FK
  • trial_start
  • converted
fact_api_callsfact
  • call_id PK
  • user_id FK
  • endpoint
  • ts
fact_token_usagefact
  • usage_id PK
  • user_id FK
  • tokens
  • ts
Capabilities

Built for subscription trade-offs.

Every SubCore Agent workflow is grounded in a real subscription decision — not a generic ML metric.

Trial-to-paid conversion

Where should we personalize friction, and when should free-tier usage caps kick in to drive upgrades — without burning intent?

Churn reduction & save flows

Which at-risk users get a grace period vs. a personalized win-back offer, and at what depth — so we save revenue we'd otherwise lose?

Pricing & discount optimization

How do we lift conversion or retention without over-discounting users who would have paid full price?

pLTV & customer value targeting

Which actions should be reserved for high expected long-term value users, instead of optimizing short-term uplift?

Customer understanding & insight

What are the real behavioral segments, leading indicators, and decision drivers underneath the subscription funnel?

Rollout governance

Which policy is safe enough to test, expand, or ship — and when should we roll one back?

Who it is for

For teams turning subscription signals into decisions.

ML engineers and data scientists

Evaluate how model outputs change business policies, not just offline metrics.

Growth analytics and experimentation teams

Compare policies, guardrails, cohort risks, and rollout criteria before launching tests.

Revenue product teams

Make pricing, discounting, retention, and conversion decisions with clearer evidence.

Lifecycle and CRM teams

Translate predictive signals into actionable treatment logic across customer journeys.

Early access

Building with design partners.

Join the waitlist if your team works on churn, pLTV, pricing, offer targeting, lifecycle personalization, conversion, win-back, or ML-informed subscription growth.

SubCoreAI is in early development with design partners.