Custify

Breadth across the classes a customer success team actually spans, with an API behind it. It connects to CRMs (HubSpot and Salesforce, both Sales Cloud and Service Cloud), billing (Stripe, Chargebee and Xero), support (Intercom and Freshdesk), product analytics and event data (Segment), communication (Slack, Gmail and Microsoft Outlook) and general automation (Zapier).

The vendor publishes an integrations directory and states it will build an integration free of charge where it makes sense for the wider customer base, and data also arrives through the documented API or manual entry. That covers CRM, billing, support, product usage and communication rather than depth in one.

Sourcecustify.com/integrations and custify.com/security-complianceread 2026-09-04

A chain that runs from signal to action without a person moving it along. The automation engine drives playbooks carrying AI decision steps, so a workflow can branch on a question like whether a customer is at churn risk rather than only on a static field, and it automates tasks, emails and notifications across Gmail, Slack, Salesforce and HubSpot.

The vendor's worked example runs end to end: login activity drops 62% at an account, the agent cross-checks against four similar accounts, suggests a re-engagement playbook, then the playbook runs with an email sent, a task created, the health score updated and the CSM notified. Agents divide by function around it, health monitoring, meeting preparation, market analysis and a Slack teammate that takes notes, creates tasks and drafts emails on request. Setup itself is orchestrated too, with uploads producing mapped lifecycle stages, segments, playbooks, scoring and thresholds.

Sourcecustify.com/product-automation and custify.com/ai-customer-success-featuresread 2026-09-04

Grounded on the customer's own material on two levels, and both are documented. The team drops in what it knows, files, pasted notes and call recordings, and the platform reads and learns from them, turning scattered documents into searchable knowledge; the vendor shows the result as lifecycle stages mapped, segments created, playbooks generated, health scoring tuned and risk thresholds set from the customer's own uploads.

That learned knowledge then shapes run time rather than stopping at setup: agents cross-check learnings against similar accounts before suggesting an action, and the conversational interface answers natural-language questions about accounts using health scores, risk and activity from the customer 360 view, which itself unifies CRM, support, billing and in-product usage. No source attribution or citation appears on generated summaries and answers, so a reviewer sees the output without seeing which document it came from.

Sourcecustify.com/ai-customer-success-features and custify.com/product-360read 2026-09-04

One real control and no gate. Admins can enable or disable CustifyAI features across the account, so a team can switch the agent layer off entirely.

Playbooks carry CSM follow-up as an explicit step, agents notify the lead CSM about concerning trends, and the vendor frames the division of labor as agents handling routine tasks while the team gets what needs a human.

No approval step, review queue, confidence threshold or hold is documented before an agent triggers a playbook, sends an email, creates a task or updates a health score, and the worked example shows a playbook running with the CSM notified after the fact.

Sourcecustify.com/ai-customer-success-features and custify.com/product-automationread 2026-10-01

Access controls and compliance are both published on a dedicated security page, linked from the site footer as Security and Compliance. Access surface: single sign-on is documented as an access control, and role-based access control lets admins fine-tune permissions so each user reaches only the data their role requires. Retention is customer-configurable, with a 24-month default the vendor states can be agreed differently and deletion within seven days of contract termination.

Compliance: SOC 2 Type 2 certified and ISO/IEC 27001 certified, both shown as badges with the AICPA SOC mark in the footer, alongside GDPR compliance with a published DPA and a separate AI policy. Supporting controls: AES-256 at rest and TLS in transit, enforced password policies with hashed credentials, annual third-party penetration testing, access limited to personnel under confidentiality agreements with anti-phishing and fraud training, a published data flow chart, and a 99.95% uptime commitment with service credits.

Sourcecustify.com/security-complianceread 2026-09-04

Scores a customer can read and adjust, and no record of what the agents did. Dashboards and reports track customer success KPIs and client interactions, calculated metrics combine events into custom measures, health scores are defined by the customer rather than fixed, and the AI layer publishes numbers rather than verdicts: churn risk scored 0 to 100 per conversation, sentiment tracked over time so a negative trend is visible before it escalates.

That is decision transparency at the prioritization layer. What is missing is a trace of the agents: no log of which agent acted on which account and why, no record of what a market analysis agent read before flagging a shift, no configuration change trail and no audit surface over the actions agents take automatically, which matters because the vendor describes agents implementing solutions rather than only suggesting them.

Sourcecustify.com/product-dashboards-and-reports and custify.com/ai-customer-success-featuresread 2026-09-04

A longitudinal customer record, not agent memory. The customer 360 view combines CRM, support, billing and in-product usage into one account history, health score history and lifecycle stage carry forward, and sentiment is tracked over time. Those are business records the agents draw on, not memory of their own. No memory construct the agents write and read back as their own state, no per agent recall across sessions and no retention or forgetting control at the agent level is documented.

Sourcecustify.com/product-360read 2026-10-01

Region is the customer's pick, and the whole chain can be pinned to it. The security page states that servers are hosted on AWS in the United States, Ohio, or in the EU, Germany, Frankfurt, and that on request data can be hosted only in the EU with the vendor ensuring its sub-processors do the same.

That is a choice between named regions with a subprocessor commitment behind it, an unusual extension of region choice. The pricing page repeats the German hosting commitment in answer to a data safety question. Residency is one of this vendor's stronger positions. EU only hosting is on request rather than self service, and no self hosted, private cloud or customer tenant option is documented.

Sourcecustify.com/security-compliance and custify.com/pricingread 2026-09-04

Named agents arrive built, and what would be a catalog is generated instead. Custify ships prebuilt AI agents as the product: meeting prep agents that build QBR agendas and prepare decks, customer health agents watching accounts and surfacing churn signals, agentic teammates reachable in Slack, and market analysis agents with news and research sub-agents scanning external sources. None of that is a catalog to browse and select from.

The candidates go the other way: playbooks are not chosen from a library but generated, either from uploaded documents during setup or from a scenario hint, which is authoring by machine rather than adoption; summary templates are configurable per audience rather than a supplied set; and no template gallery, play library or agent marketplace is offered. Custify's distinctive move is generating the configuration a template library would otherwise supply, which solves the same buyer problem by a different route.

Sourcecustify.com/ai-customer-success-featuresread 2026-09-04

Fires on lifecycle and behavioral events rather than on a dashboard someone remembers to open. Triggers documented: a drop in product usage, an account stuck during onboarding, an expired trial, a health score falling, sentiment turning negative in an email or message, a churn risk score crossing a threshold, and renewal dates approaching, with lifecycle stages and segmentation routing which team member is told.

Market analysis agents add external triggers, watching press releases, headcount changes, M&A activity and champion movement and matching a sudden shift against recent sentiment. Channels out: in-app tasks and alerts, Slack, email through Gmail, and writes into Salesforce and HubSpot, with the Slack teammate also serving as an inbound channel a CSM can ask questions through.

Sourcecustify.com/product-tasks-and-alerts and custify.com/ai-customer-success-featuresread 2026-09-04

The platform names where its models run, and the customer does not choose among them. The CustifyAI FAQ answers the question directly: all models run on AWS Bedrock, which the vendor cites as SOC 2 compliant, with no customer data stored in the models beyond generating a response.

Bedrock is a multi-model service, and the vendor's own phrasing is plural, so more than one model is in use behind a disclosed platform selected by the vendor. The vendor presents AWS Bedrock as running its inference rather than listing it as an infrastructure subprocessor. No model list, routing policy, per agent selection or bring your own key option is published. Admins can enable or disable CustifyAI features at the account level, a control rather than a choice.

Sourcecustify.com/ai-customer-success-featuresread 2026-09-04

A published developer surface. Custify links a Developer Center at docs.custify.com from the footer of every page under Resources, and the security page describes personal data arriving in the platform explicitly and voluntarily through our API, pulled from integrations or entered manually, so the API is a first-class intake route rather than an afterthought. More than twenty prebuilt integrations sit alongside it, including Zapier and Segment for onward reach. No SDK, client library or MCP server is published.

Sourcedocs.custify.com and custify.com/security-complianceread 2026-09-04

Tuning against outcomes, with nothing tested before it runs. Health scores, segments, lifecycle stages and calculated metrics are all customer-defined and can be tailored per product, plan and customer tier, so a team adjusts the model against what actually predicts churn in their book and reads the effect through dashboards and sentiment trends. Churn risk scores are numeric per conversation, which gives a readable result to calibrate against.

That tuning is aimed at the business outcome, and nothing is aimed at the agents: no controlled comparison of one configuration against another, no scored evaluation set, no benchmark, no regression check when the uploaded knowledge or generated playbooks change, and no sandbox or dry run before an agent triggers a playbook against a live account.

Sourcecustify.com/calculated-metrics and custify.com/product-healthread 2026-09-04

Everything the platform touches arrives through an authenticated connector or its own API. Data enters from CRM, support, billing and product usage through prebuilt integrations, through the documented API, or by manual entry, and actions leave the same way, as tasks, alerts, emails through Gmail and writes into Salesforce, HubSpot and Slack.

The market analysis agents research external sources such as press releases, news and headcount changes, which is retrieval rather than browser control. No browser extension, headless session, portal navigation, form filling or screen automation is documented.

Sourcecustify.com/integrations and custify.com/ai-customer-success-featuresread 2026-09-04