Evidence and setup

This is a proposed workflow with an illustrative example, reviewed against the linked documentation on 5 September 2026. It is not a live HubSpot test or a claim about sales results. Start with the connection and permission checks and use the free worksheet to define the job.

Find the neglected queue

The first workflow should not score every contact or rewrite the CRM. It should find leads that already meet the business’s definition of waiting: no owner, no response after a named period, or an agreed next action that is overdue.

HubSpot scopes control which CRM objects an integration can access. Use the granular read scopes required by the selected contact, company and deal fields; HubSpot documents current app scope configuration.

OAuth app scopes can exceed the authorizing user's record visibility. A pipeline filter is a task boundary, not proof of enforced record isolation. Check the OAuth permissions model before using real records.

What the business provides

The exact HubSpot pipeline and stages in scope
A waiting threshold for each stage
Required contact, company and deal fields
Approved offer, pricing and scheduling facts
The rep who owns approval and the excluded accounts

The lead packet

Field	Requirement
Lead	Direct HubSpot record link and owner
Why now	Exact rule that placed it in the queue
Context	Last meaningful activity and known business facts
Gap	Missing information or unresolved objection
Next action	Proposed internal action and follow-up draft
Approval	Named sales owner before any send

Copyable first-task brief

Review deals in the approved pipeline that meet the waiting rules. For each, return the HubSpot link, owner, stage, last meaningful activity, known facts, missing information, recommended next action and one follow-up draft. Exclude named accounts and active opportunities with a future task. Do not send, change stages, merge records, promise terms or invent personalization. Stop after the review queue is written.

Worked example: prepare the queue

Synthetic input: the test clock is 5 September. D-104 was last contacted on 2 September, has no reply in the supplied history, and lacks a service location. D-105 has an agreed callback on 8 September. The business rule says follow up after two days, except where a later callback is agreed. Both records name the service coordinator as owner.

Expected output:

Deal	Decision and evidence	Preparation
D-104	Due: three days since documented contact; no later reply supplied	Ask for the service location; assign the draft to the service coordinator
D-105	Exclude for now: callback agreed for 8 September	Preserve the agreed date; do not generate an early nudge

The D-104 draft could read: “Could you share the service location so our team can confirm the next step?” This is expected wording for the example, not a Bot output or an approved send.

Failure case to test

Remove D-104's contact history. The Bot should say the waiting decision is unknown; a record modification date does not fill that gap. Remove its owner and expect an assignment request. Conflicting evidence should block the next action and identify the conflict.

Before scheduling, check ordinary, missing-data and conflicting-data cases with the real reviewer. A fixed number of successful runs does not guarantee reliability. Keep sends and CRM changes separately approved.

Measure the result

Track qualified waiting leads found, time from queue entry to rep review, drafts approved without factual changes and deals incorrectly surfaced. Precision matters more than the number of follow-ups produced.