Your Oppy Saved Ten Minutes. What Did You Spend?

If an oppy saves ten minutes preparing a leasing handoff, that is useful. If someone spends twelve minutes repairing it, the stopwatch has been selectively informed.

Before expanding the job, ask for a Usage Receipt: what work was accepted, what consumption was actually recorded, and what people spent getting it over the line.

This is a proposed internal accounting workflow, not a released Oppy billing feature. Start with approved records supplied by a human, not an imagined meter.

The mortgage headline is not your cost model

On November 6, 2026, HousingWire reported that Blend's Autopilot customer count had increased from six to 10. The same report put economic value per funded loan at $81. That is a company economic metric, not a lender's cost to originate a loan. The underlying earnings release was not independently verified in the supplied research. 1

Those figures describe adoption and company economics. They do not price your leasing handoff, mortgage intake brief, or maintenance summary. That calculation needs your own records and a clearly defined finish line.

Define one accepted task

For a property manager, choose one narrow job: preparing an internal leasing handoff from approved inquiry notes. Count it as accepted only when a human confirms that the brief identifies the customer's stated question, preserves unresolved facts, and points to its sources.

Do not count three drafts as three completed tasks. Do not count an accepted brief as a signed lease. Record any later customer outcome separately, with its own evidence.

Give every case a local task ID in a staff-maintained worksheet. Attach all attempts to that ID, including abandoned drafts and human repairs. Keep rejected and unfinished cases in the batch cost, even though they do not enter the accepted-task denominator.

Collect four inputs before asking Oppy to calculate

Have the operations owner approve the task definition and acceptance record. Have the person responsible for spending approve the cost-allocation method before seeing the result.

  1. Completion records: task ID, attempt IDs, acceptance or rejection, reviewer, and a pointer to the accepted output. These are records your team maintains, not a claimed native Oppy task log.
  2. Usage evidence, if available: approved plan and usage records, their billing period, and any reliable links between consumption and task IDs. If only account-wide consumption is available, leave task-level credits unknown.
  3. Human effort: minutes spent preparing inputs, operating the workflow, reviewing, correcting, and handling failures. Include the effort of preparing these receipts and the organization's approved labor rate.
  4. Cost rules: the actual subscription expense assigned to this queue, separately billed charges supported by records, and one-time setup or training costs. Name the allocation basis and retain the approval.

Supply only the information needed to reconcile the work. Use redacted copies or synthetic cases for the first test. Keep credentials, payment details, and unnecessary customer documents out of the packet.

A change in an account's remaining credits does not, by itself, identify which task consumed them. Other work may have happened between readings. The receipt should expose that gap, not divide it away.

Keep credits and dollars in different columns

As observed on November 7, 2026, Oppy's public monthly Pro plan card lists $400 and 13,000 credits per month. The reviewed evidence does not establish a credit-per-action schedule, overage terms, or a task-level billing API. 2

Dividing the plan price by its allowance creates an allocation ratio. It does not establish the cash charge for the next task. Unused allowance, shared work, and the actual billing terms still matter.

Oppy's April 3, 2026 analytics announcement documents five Account Concierge reports: Lead Flow, Communication Volume, Engagement, CRM Activity, and Deal Pipeline. It does not document per-task credit accounting. Do not substitute a communication count for a usage record. 3

Two useful answers: recorded credits per accepted task, when the evidence supports it; and allocated operating cost per accepted task. Neither is automatically the marginal cash cost of one more task.

Use the same billing period and task population throughout. Do not add a notional dollar value for included credits on top of the subscription expense. That charges the same allowance twice.

Give one internal oppy the reconciliation job

Oppy's setup walkthrough, displayed April 21, 2026, documents custom roles and supplied knowledge. Use those building blocks for an internal receipt-preparation role, after confirming that your configuration can handle the approved inputs. The source was updated May 7, 2026. 4

The Functions guide, displayed August 22, 2025, documents Public and Admin permissions. Admin-only limits who can invoke a function; it does not make a write function read-only. 5

An administrator must verify actual tool restrictions and isolate this role from customer messaging and production changes. If that boundary cannot be established, do the exercise in an approved isolated environment instead. A prompt is an instruction, not an access control.

Copy this starter prompt:

ROLE: Prepare an internal Usage Receipt for human review.
Use only the supplied, approved records. Do not fetch billing data.


INPUTS:


    

  • Task definition and acceptance rule
    • 

  • Task and attempt records for one stated period
    • 

  • Available usage evidence with source pointers
    • 

  • Human minutes and approved labor rate
    • 

  • Approved subscription allocation and separate charges
    • 

  • One-time setup costs and named review owner
    • 



RECONCILE:
Group attempts by task ID. Keep failed and unfinished work in
batch costs. Count each human-accepted task once.
Do not infer task credits from an account-wide balance change.
Keep recorded credits, allocated dollars, and setup costs separate.
Mark missing evidence UNKNOWN. Never fill a missing cost with zero.
Flag unmatched usage, duplicate entries, and period mismatches.


OUTPUT:


    

  1. Period, task definition, and source pointers
    2. 

  2. Started, accepted, rejected, and unfinished task counts
    3. 

  3. Attempt count, recorded credits, and attribution gaps
    4. 

  4. Human minutes and cost, allocation basis, and separate charges
    5. 

  5. Recurring cost per accepted task; setup-inclusive cost separately
    6. 

  6. Unknowns, reconciliation questions, and named human reviewer
    7. 



Return a draft only. Do not approve expenses, change a plan,
send messages, alter records, or declare customer outcomes.

The human reviewer checks the joins, acceptance decisions, arithmetic, and allocation. Only that person approves the internal receipt. The role does not certify an invoice or enforce a spending cap.

Work through one fictional batch

Every number below is invented for illustration, not an Oppy rate or customer result. Assume one queue has complete task-linked usage evidence and a human-approved $60 subscription allocation for the stated period.

Thirty leasing briefs were started. Eight extra drafting attempts brought the attempt count to 38. Twenty-four briefs were accepted; six were rejected or unfinished. The batch's recorded consumption was 900 credits, including those extra attempts.

Human work across the entire batch took 120 minutes at an approved $30 hourly rate. Separately billed charges were verified as zero. Setup took another two hours at the same rate.

Recurring human cost: 120 / 60 x $30 = $60
Allocated recurring batch cost: $60 subscription + $60 labor = $120
Recurring cost per accepted brief: $120 / 24 = $5.00
Recorded credits per accepted brief: 900 / 24 = 37.5
One-time setup cost: 2 x $30 = $60
Setup-inclusive pilot cost per accepted brief: $180 / 24 = $7.50

Read the result without double charging

The rejected work remains in the numerator. Setup remains visible. Credits do not become an extra invented cash charge. If task-linked usage were unavailable, the credit calculation would disappear and the attribution gap would remain.

Decide what the next batch earns

Test a small batch before expanding the role. Include one case with missing usage evidence, one with repeated attempts, and one with no accepted output. The receipt should report the gaps and avoid division by zero, not produce a reassuring average anyway.

Compare the assisted route with a human-only baseline using the same acceptance rule and comparable case difficulty. Measure all human effort, not just first-draft speed. Let the owner decide whether the task earns another batch after reviewing cost and quality together.

Time recovered is capacity. It becomes cash savings only when expenditure actually changes. Keep that distinction in the decision record.

The useful question is not whether your oppy looked busy. It is whether a specific, accepted piece of work justified what you spent.

References

  1. HousingWire: Blend posts smaller Q3 loss, warns of steeper mortgage revenue drop ahead. Published November 6, 2026. Company-reported metrics; primary earnings material was not independently verified.
  2. Oppy: Public product homepage and monthly plan cards. Undated page, observed November 7, 2026. Public pricing may change.
  3. Oppy: Real-Time Analytics Reports: Know Your Business at a Glance. Displayed April 3, 2026.
  4. Oppy: Watch: Creating Your First Oppy in 3 Minutes. Displayed April 21, 2026; updated May 7, 2026.
  5. Oppy: Functions guide and permissions. Displayed August 22, 2025; retrieved version updated May 7, 2026.