We raised €3M to grow your revenue on autopilot. Read the announcement
  • Sign in
  • Try for free
Back

Proactive onboarding

11 May 2026 · Uku Tammet

We've overhauled how Onboarding agents guide new users. Instead of sitting in the corner waiting for a click, the agent now opens itself at the right moment, learns who the visitor is, and walks them through the right onboarding flow for their situation.

Triggers that open the agent automatically

Each Onboarding agent can be configured with triggers that decide when the agent should proactively expand from the launcher into a conversation:

  • First login: the agent greets a visitor on their very first session.
  • Inactivity: after a configurable period of no activity (e.g. user came back two days later and hasn't done anything yet), the agent re-engages.

Triggers are evaluated against the visitor's actual history, so the agent doesn't repeatedly accost returning users who've already been onboarded.

Discovery before pitching

When the agent opens, it doesn't immediately fire a walkthrough. It first runs a short discovery (asking the qualification questions you've configured on the template) so it understands who's in front of it (role, goals, what they're trying to do). This shapes which workflow it offers next.

Workflows as the unit of onboarding

An Onboarding agent now owns an ordered list of workflows: named, step-by-step guides (e.g. "Set up your first project", "Invite your team", "Connect your data source"). Each workflow has a plain-text condition (e.g. "user is a team admin and hasn't invited teammates yet") that the agent evaluates from discovery answers and visitor history.

The flow is: greet → discovery → offer the next applicable workflow → walk it on confirmation → repeat with the next one. Conditions are never surfaced to the visitor. They just see a relevant suggestion.

Smarter walking

While walking a workflow, the agent now:

  • Adds a one-sentence orientation when the user lands on a new view ("This is the project settings page, where we'll connect your data source") instead of jumping straight to a click.
  • Skips form fields the user has already filled in.
  • For long forms (3+ empty fields), hands off ("fill in what you have, I'll pick back up") rather than dragging visitors field-by-field.
  • Lets the user lead. The agent never auto-marches through steps or auto-skips workflows it assumes are "already done."

Reliability fixes shipped alongside

Highlight targets now correctly ignore off-canvas drawer content, hidden elements, and zero-sized iframes, so hints land on the thing the user can actually see. First-turn behaviour was tightened so the agent reliably greets and takes the next action in the same turn, instead of stopping after hello.

Existing Onboarding agents pick all of this up automatically. Triggers and workflows are configurable from the backoffice.

For AI agents: a clean Markdown version is available at /changelog/2026-05-11-proactive-onboarding.md.

Handhold (“Handhold”) provides technology and AI-powered agents designed to help businesses engage visitors, qualify inbound leads, deliver personalized product experiences, and guide users through onboarding. Handhold is a software platform and does not provide sales, marketing, legal, or advisory services.

Any customer interactions, product information, or recommendations generated through Handhold agents are based on the configuration and data provided by the customer. Handhold does not control, verify, or guarantee the accuracy, completeness, or suitability of any information presented through its platform. By using this website or the Handhold platform, you acknowledge that all content is provided for informational and operational purposes only and agree to our Terms of Use and Privacy Policy.

Handhold

Sign inContact usCareers

Legal

Privacy policyCookie policyResponsible disclosure