How Custom GPTs Are Transforming Product Discovery and Lead Nurturing in HubSpot

Generative AI is rapidly reshaping the way customers discover products online — but the true breakthrough comes when AI is fused directly with CRM infrastructure. For brands using HubSpot, the most powerful development of 2025 is the rise of Custom GPTs: conversational agents that can understand a user’s needs, recommend tailored products, and automatically convert those interactions into structured CRM intelligence.

At the centre of this shift is a new architecture that connects a brand’s Custom GPT to HubSpot through a secure middleware layer. This configuration unlocks personalised product discovery, automated contact enrichment, and real-time lead nurturing — all within a single natural conversation.

This article explores how it works, why it matters, and how forward-thinking teams are implementing it today.

The New Model: Chat → CRM → Conversion

The traditional discovery journey is fragmented. Customers research on websites, get lost in endless product catalogues, and fill out forms reluctantly. Sales teams receive incomplete profiles, marketing teams rely on generic nurture sequences, and CRM data becomes inconsistent.

The Custom GPT / HubSpot integration solves this by reinventing the journey from the first moment of intent:

Old ExperienceNew GPT-Integrated ExperienceCustomers guess which products suit themCustomers describe needs naturally (“I have sensitive skin and redness”)Browsing takes time; overwhelm is commonGPT instantly generates personalised routines/productsManual lead captureGPT captures only what is necessary and requests consent clearlyManual CRM updatesContact/lead data is automatically populated and structuredGeneric email flowsNurture workflows are triggered based on actual stated needs

Within one conversation, the system simultaneously provides value to the user and builds a high-quality CRM record that powers the next steps in the sales/marketing lifecycle.

How the System Works

1. The User Speaks Naturally

Customers interact with the GPT as they would with a consultant:

“I’m looking for a skincare routine for sensitive skin with a budget under £60.”

The GPT asks a few clarifying questions — allergies, fragrance preferences, routine simplicity — mirroring an in-store consultation.

2. Secure Middleware Connects to HubSpot

A custom API layer sits between GPT and HubSpot. It handles:

  • Auth and OAuth token refresh

  • Contact creation and deduplication

  • Consent capture and audit logging

  • Product data retrieval

  • Deal creation and enrichment

  • Custom behavioural events

This layer ensures that all CRM operations respect consent, data minimisation, and GDPR/CCPA standards.

3. GPT Recommends Product Sets, Not Just Items

Using the product catalogue and RAG-indexed product content, the GPT builds structured routines:

  • Morning and evening usage steps

  • Ingredient rationale

  • Conflict/contraindication checks (e.g., retinoids + exfoliants)

  • Budget-aligned alternatives

This turns generic browsing into a personalised, high-trust consultation.

4. Real-Time CRM Enrichment

If the user asks for the routine to be emailed or saved, the GPT:

  • Upserts the contact

  • Updates preferences and concern fields

  • Logs the interaction

  • Captures consent at a granular channel-specific level

Everything is cleanly written into HubSpot under standard and custom property schemas.

5. Qualified Lead & Deal Creation

When a user accepts a routine or shows intent, the GPT can create a ready-to-work deal:

  • Deal name: “Sensitive Skin Routine – ChatGPT”

  • Stage: “Qualified – Recs Shared”

  • Pipeline: “D2C Consultations” or custom pipeline

  • Line items: Each SKU in the recommended bundle

  • Snapshot of routine and consent at the moment of creation

This ensures sales or CX teams have complete context.

6. Automated Nurture Sequence

With consent in place, the GPT/middleware enroll the contact into a tailored nurture program — for example, Sensitive Skin: 5-Day Routine Education Sequence. The emails include:

  • The routine

  • Usage tips

  • Ingredient education

  • Follow-up check-ins

  • Optional incentive on Day 5

This is more personalised and higher-converting than traditional lifecycle sequences.

The HubSpot Property and Workflow Pack

To operationalise this system, a structured HubSpot configuration is required. The recommended pack includes:

Contact Properties

  • Skin type (az_skin_type)

  • Primary concern (az_primary_concern)

  • Sensitivities/allergies (az_sensitivities)

  • Budget band (az_budget_band)

  • Last routine summary (az_last_routine_summary)

  • Consent status per channel (az_gpt_consent_email, etc.)

  • GPT lead score (az_gpt_lead_score)

  • Preferred channel (az_preferred_channel)

Deal Properties

  • GPT session ID (az_gpt_chat_session_id)

  • Routine snapshot (az_gpt_routine_summary)

  • Recommendation set ID

  • Consent snapshot at time of deal creation

Optional Custom Object: Consultation

Stores the full transcript summary, decision logic, routine JSON, and product SKUs — ideal for advanced analytics and omnichannel auditing.

Workflows

  1. GPT – Lead Scoring
    Calculates a dynamic score based on information richness and intent signals.

  2. GPT – Promote to SQL & Assign Owner
    Moves high-intent leads into the sales pipeline.

  3. GPT – Tag & Normalise Deals
    Ensures consistency when GPT-created deals hit HubSpot.

  4. GPT – Nurture (Sensitive Skin)
    Triggered automatically when consent and intent criteria are met.

  5. GPT – Consent Hygiene
    Ensures that opt-outs remove contacts from nurture sequences automatically.

These workflows ensure that the AI-driven journey remains compliant, consistent, and predictable.

Why This Matters for Brands

1. Frictionless Personalisation at Scale

Every user gets a bespoke consultation without waiting for a human expert.

2. Clean, GDPR-Safe CRM Data

No more messy contact duplication or unclear consent states.

3. Higher Lead Conversion

Routine acceptance is a powerful buying signal; deals created automatically capture it.

4. Richer First-Party Data

The GPT captures authentic problem statements in the customer’s own words.

5. Faster Time to Value

Because the AI does the discovery and qualification, sales teams engage only when there is real intent.

6. A Competitive Advantage That Compounds

The more customers interact with the GPT, the richer your CRM intelligence becomes — powering smarter segmentation, automated campaigns, and new product insights.

What’s Next?

The next wave will extend this architecture into:

  • Unified omnichannel consults (web, app, retail staff, WhatsApp)

  • Automated routine follow-ups (“How is your redness after 7 days?”)

  • Predictive reorder prompts using GPT + HubSpot custom events

  • Autonomous A/B testing of recommendations and nurture flows

Brands adopting this model now are getting ahead of a transformation that will soon become standard across e-commerce and retail CRM ecosystems.

HubspotFrancesca Tabor