As HubSpot shifted toward AI‑First Onboarding, we wanted to move beyond static checklists and generic “getting started” flows. The goal was to design an agentic intake experience that:

• Understands a new customer’s goals, context, and constraints

• Uses those signals to intelligently configure the CRM and onboarding surfaces

• Hands off into a more focused, opinionated onboarding system (User Guide, Grow Guide, Global Home) instead of overwhelming users with every possible task at once

This case study covers how I framed the problem, defined the Agentic Intake + Onboarding system, and created UX guardrails other teams could safely build on.

1. Framing the system

I started by mapping the end‑to‑end journey:

Signup → Agentic Intake → Smart Setup / Review → AI‑aware Guides → Global Home

For each step I captured:

• What the user is trying to accomplish

• What context we already know

• What new signals we should ask for (and when)

• Which team owns the surface and underlying systems

This mapping became the backbone for team charters and ownership docs, making it clear where the Agentic team stops and where onboarding / nav teams begin.

2. Defining requirements & guardrails for Agentic Intake

I then wrote a focused requirements doc for the agent, covering:

• Inputs:

  • Business type and goals (e.g., “book more meetings”, “move deals through a simple pipeline”)

  • Team composition (solo founder vs. sales team vs. agency)

  • Existing tools (e.g., Gmail, calendar, other CRMs)

• Outputs:

  • Recommended configuration (pipelines, stages, key objects)

  • Smart tasks and checklists tailored to their goals

  • Hooks into data synchronization and enrichment (see Case Studies 2–3)

For the UX, I set explicit guardrails:

• Always show what the agent is proposing to change, before it happens

• Provide simple controls: accept all, accept with edits, or skip

• Make AI choices legible and reversible – clear copy on “why this was suggested”, easy undo, and a log of major changes

• Avoid turning intake into an interrogation; keep interactions short, progressive, and visually calm

These guardrails were written so that multiple teams could implement agentic patterns without reinventing safety, copy, or review/confirm logic.

3. Connecting Agentic output to onboarding surfaces

Working with designers on User Guide, Grow Guide, and Global Home, I translated the agent’s output into surface‑specific requirements:

• User Guide / Grow Guide

  • Show a short, sequenced set of tasks tied directly to the goals captured in intake

  • Reduce generic “learn HubSpot” tasks; emphasize “do something real” (e.g., send a first email, import existing warm contacts)

• Global Home

  • Show high‑signal cards anchored in what the agent just set up (e.g., “Review your new pipeline”, “Check imported contacts”)

  • Treat the home experience as a live reflection of what the agent has configured, not a separate checklist

I documented these expectations so each surface could design within its own constraints while still feeling like one connected onboarding system.

4. Patterns & collaboration

Because Agentic Intake touched many teams, I invested early in reusable patterns:

• Intake prompt templates (how we word goals, trade‑offs between open vs. structured questions)

• Review & Confirm UI patterns – modals, summary views, change logs

• Microcopy and tone guidelines so the agent felt friendly but not chatty or opaque

I used design reviews, async Looms, and shared Figma libraries to socialize the patterns and keep new explorations aligned as other teams picked up agentic work.

Outcomes

A clear, written vision for Agentic Intake as part of an AI‑First onboarding system, not a standalone chatbot.

• Reusable UX guardrails and patterns (intake, review/confirm, AI attribution) that other teams could adopt.

• Better cross‑team alignment: onboarding, nav, and AI teams all working from the same journey map and expectations, which reduced overlap and conflicting designs.