01 / FIND THE FRICTION
Start with a person and a status.
In this example, staff spend 12 minutes per enquiry copying details and deciding who follows up. Before buying anything, name one enquiry owner, make service and location fields clear, and agree a shared status: new, assigned, awaiting reply, or closed.
Then inspect the tools you already pay for. HubSpot’s workflow options depend on subscription and action availability; check the actual account before proposing a build. If an existing workflow solves the problem reliably, configure it first. HubSpot’s workflow guide
This first pilot can use fixed rules and approved templates. A model that drafts quotes or interprets free-form enquiries is a separate decision, with separate review and testing.
02 / MAP ONE JOURNEY
From form to follow-through.
One website form, one CRM, one mailbox. The proposed sequence below keeps decisions visible and gives every exception somewhere to go.
Capture one good enquiry.
One website form collects business name, work email, service, location or site area, and requested timing. Missing or unclear fields go to a staff queue; the workflow fills in nothing on its own.
Keep sensitive customer records out of this first pilot.Create a record once.
Check the form submission’s unique reference before creating the CRM enquiry. Link repeat submissions for review instead of silently creating another deal or sending another message.
A repeated event must produce the same result, without another send.Give a person ownership.
Use an agreed service-area rule to assign a named team member. If no rule matches or the owner is unavailable, notify the backup owner and leave the enquiry visibly unassigned.
An exception is work for a person, not permission to guess.Review the acknowledgement.
Staff check the contact details and approve the message before sending. Use an agreed template that confirms receipt and the next step. It does not quote a price, promise availability, or agree a contract.
Customer-facing text stays under human approval in this pilot.Consider one follow-up.
After two business days, check the agreed timezone and working calendar, reply status, opt-out status, and documented permission. If eligible, put one follow-up in the owner’s approval queue. Recheck these conditions immediately before sending.
No reply is only one condition. It is not permission to contact someone.Stop and hand over.
A reply, opt-out, manual closure, or uncertain mailbox state cancels the pending follow-up. A person handles the conversation from there. If a system fails, pause sending and alert the owner with a clear recovery task.
One permitted follow-up is the limit, not the start of a campaign.
At any uncertain step: pause, queue, alert. The enquiry stays visible to staff. The system does not treat missing information, an outage, or an unanswered approval as success.
03 / KEEP THE CHECKS CLOSE TO THE ACTION
Before the first message leaves.
- A real approval step
- The owner approves each outbound message in this pilot. Approval is attached to the exact text and recipient, and expires if the enquiry or eligibility changes. Pricing, quotes, unusual requests, and promises always go to a person.
- Permission and a clear stop
- Agree the permitted contact purpose and channel rules before launch. Store the permission record, provide a clear opt-out, and apply suppression to every queued message. A reply or opt-out stops the scheduled follow-up; an ambiguous match goes to staff.
- A failure path you can use
- Set a named owner for alerts, a manual queue, and a pause switch. Record send attempts and outcomes. If a send result is unknown, reconcile it before retrying; never send again just because an acknowledgement was lost.
Reply monitoring needs maintenance. Google documents that Gmail notifications can be delayed or dropped and mailbox watches need renewal. Our design recommendation is to reconcile mailbox state and pause follow-ups when reply status cannot be confirmed. Google’s Gmail notification guide
04 / AGREE THE HANDOVER
A sprint should leave you with more than a demo.
- A written workflow and scope. Fields, owners, business-day calendar, approved templates, permission rules, stop conditions, and exclusions.
- A customer-owned setup. Agreed access, field mapping, duplicate prevention, approval queue, exception alerts, and a pause switch. No passwords in chat.
- A test record. Staging results, negative cases, known limitations, and a short controlled trial agreed with the customer before external sending.
- A usable handover. Account ownership and export instructions, a staff walkthrough, monitoring responsibilities, manual fallback, and rollback instructions.
- A defined support boundary. This example includes 14 days of in-scope defect support after acceptance. Coverage and any ongoing support are agreed in writing.
Keep this first scope small.
Exclude ad campaigns, WhatsApp, contact migration, quote pricing, automatic contracts, and repeated marketing sequences. Adding another channel or an AI drafting step needs a new scope.
The fictional build price is $1,500 before applicable tax and separate vendor costs. Ten business days is a planning assumption after access, funding, and specialist availability, not a delivery promise or an offer to purchase.
05 / TEST THE EXCEPTIONS, TOO
What counts as working?
Agree these checks before implementation. Run them in staging with synthetic contacts first, then review the evidence together. A real customer signs off on the accepted scope and results.
Ten representative form submissions
Required fields reach the right CRM record and the correct owner. Missing email or service goes to the exception queue.
Replay the same submission and delivery event
One enquiry record and, when approved, one message. No duplicate record or send.
No permission, or an existing opt-out
No proactive follow-up is sent. Staff can see why it was suppressed.
A reply or opt-out arrives before the send
The queued follow-up is cancelled, including after staff approval but before sending.
A weekend, holiday, or unavailable owner
The agreed business-day calendar and backup route work. No promise of a staffed response is invented.
Gmail or CRM fails; sending has an uncertain result
Sending pauses, the owner is alerted, and recovery checks the result before retrying. No blind resend.
Manual operation and rollback
Staff can pause the automation, find every open enquiry, continue manually, and restore the previous process.
During the controlled trial, measure staff time per enquiry including approvals and exceptions. Also record duplicates, suppressed messages, missed replies, and unresolved failures. Expand only after the team can operate and recover the workflow.
06 / CHECK THE VALUE BEFORE BUYING
Make the numbers earn their place.
Here is the arithmetic for the fictional cleaning company. None of these inputs has been measured. The four-minute target still includes staff review and remaining administration.
- Monthly enquiry volume
- 80 enquiries
- Assumed time today
- 12 minutes each → 16 hours/month
- Hypothetical time after the pilot
- 4 minutes each → 5.3 hours/month
- Potential capacity released
- 80 × 8 minutes ÷ 60 → 10.7 hours/month
- Gross time value at an assumed $30/hour
- $320/month
- After hypothetical $40/month extra tools
- $280/month in time value
- Illustrative $1,500 build ÷ $280/month
- About 5.4 months on time value alone
Calculations use unrounded hours. Recovered sales are assumed to be zero. The payback illustration excludes tax, extra maintenance, and costs beyond the stated assumptions. Released staff time becomes capacity; it becomes a cash saving only if actual paid costs fall. This is not a return guarantee.
Measure your own volume and handling time, verify tool costs, and obtain a funded implementation quote. A lower volume, more review, or an already capable CRM can change the decision. Optional support from $250/month could consume most of this example’s time value; buy it only when the agreed coverage is worth that ongoing cost.