What it is
Hook is a London-based Customer Success platform built around machine-learning churn prediction and a set of AI agents that act on it. It trains an account-health model on your own user, subscription and product-usage data, layers in calls, tickets and external news, and flags at-risk renewals up to 180 days out. Hook’s own figures: 93% prediction accuracy in live environments with customer-specific training sets, and 90% of churn flagged as a risk before renewal. Customer hackajob reports 85% accuracy predicting renewals 180 days out. Treat those as vendor-reported until you backtest them on your own book.
The product is the agents, not a dashboard. Echo detects risk from usage and conversations; Activator runs onboarding milestones and queues next-best actions; Guardian watches adoption and nudges; Radar tracks external events such as acquisitions, leadership changes and funding rounds; Delta finds upsell candidates and hands them to sales; Anchor prepares renewal conversations. Each recommendation cites the ticket, usage pattern or call it came from, and actions go out through Slack, email or straight to a CSM for approval. Since July 2026 an MCP connector exposes the same signals and workflows to Claude, ChatGPT and other AI tools.
Named customers include Salesloft (25-35% CS efficiency gain, vendor-reported), hackajob, Superhuman, Planful and GWI. Founder Firaas Rashid previously ran Customer Success EMEA at AppDynamics; the company is backed by Lightspeed and LocalGlobe.
The use case it actually solves
The question Hook answers is “do we need a full CS platform, or a churn-prediction agent on top of the CRM we already have?” A concrete scoped case: a 6-CSM team managing 400 mid-market accounts in Salesforce, with usage data in Snowflake and calls in Gong, that keeps getting surprised at renewal. Hook connects to those three systems, has a first trained model in about four weeks (its stated 28-day, four-phase implementation), and replaces the hand-weighted health score with a model fitted to your actual churn history. The CSMs keep working in Salesforce and Slack; Hook tells them which 30 accounts to call this month and why.
Integrations
Hook’s integration list is unusually warehouse-heavy for a CS tool: Snowflake, Amazon Redshift, Google BigQuery, Databricks, PostgreSQL, Amazon Athena, Amazon S3 and Microsoft Fabric on the data side, plus Pendo and Mixpanel for product analytics. CRM is Salesforce and HubSpot. Conversation sources include Gong, Chorus, Jiminny, Clari, Salesloft and Outreach; support sources include Zendesk, Intercom, ServiceNow, Jira Service Management and Front. SSO via Okta. Hook is ISO 27001 certified and SOC 2 Type 2 attested. There is no public REST API documentation; the MCP connector is the programmatic surface.
Pricing reality
Quote-only. There is no public pricing page, and pricing scales with the number of customer accounts you manage. Third-party figures conflict: one 2026 roundup lists Hook from $500/month, a software directory lists a $399/month entry point, and buyer-side estimates put a full deployment at roughly $20K-$25K a year for the AI agents alone and $40K-$50K a year with the ML prediction models. Plan on a $20K-$50K annual band for a mid-market book. That sits below Vitally and ChurnZero mid-market deals ($25K-$70K) and well below Gainsight.
Best for
Heads of Customer Success at B2B SaaS companies with roughly 200-2,000 accounts, a working CRM and product data already landing in a warehouse, who want churn forecasting and next-best actions without migrating the team onto a new system of record. Strongest when CSM capacity is the bottleneck and the team needs a ranked call list more than another place to log activity.
Alternatives and when to pick them
- Gainsight — the market-share leader. Pick it if you need a full CS system of record with deep customization, a dedicated admin and enterprise governance, and you can absorb a six-figure contract and a long rollout.
- ChurnZero or Vitally — the mid-market default platforms. Pick one if your team needs a place to run success plans, tasks, in-app messaging and CSM workspaces, not just a prediction layer. Vitally fits HubSpot-centric teams better; ChurnZero is stronger on in-app engagement.
- Planhat — pick it if you want a configurable data model across CS, revenue and product data in one workspace and have the ops capacity to set it up.
- QuadSci — the closest like-for-like and the fastest-moving entrant in prediction layers. It forecasts 9-18 months ahead from raw telemetry and pushes output into Gainsight, Salesforce and Clari. Pick it if you already own a CS platform and want a longer forecast horizon feeding it; pick Hook if you want the agents that act on the prediction, not just the forecast.
Watch-outs
- The model needs churn history to learn from. A company with 40 churned accounts in its whole history gives a supervised model very little signal, and the 93% headline will not hold. The guard: ask Hook to backtest on your last four renewal quarters before signing, and get the accuracy figure measured on your data written into the evaluation criteria.
- It is not a full system of record. Hook ranks, explains and nudges; teams that need success plans, CSM task queues, in-app guides and customer-facing portals will still need a CS platform or a CRM build-out next to it. The guard: list the workflows your CSMs run daily and mark which ones Hook covers before you cancel anything.
- Warehouse-first plumbing assumes a data team. The richest signals come from Snowflake, BigQuery or Databricks tables, not a CRM sync. The guard: confirm which product-usage tables exist and who owns them before kickoff, or the 28-day timeline slips while you wait on analytics engineering.
- Name confusion in research. Several unrelated companies called Hook exist, including a music-licensing app that raised a Series A in 2026. The guard: verify any funding or news item against hook.co before quoting it in a business case.