Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.duvo.ai/llms.txt

Use this file to discover all available pages before exploring further.

This guide helps you decide where humans belong in the loop, choose the right approval shape, and set escalation thresholds before you encounter an incident in production. For a reference on the mechanics of creating approval requests — how to write SOP instructions, respond via Slack, or manage the Activity Inbox — see Human-in-the-Loop.

When to Add an Approval Step

Not every step needs human review. Use this risk framework to decide where approval gates add genuine value and where they only add friction.

Risk tier framework

TierAction typeExamplesDefault stance
1 — Irreversible externalSends something that cannot be recalled or undoneOutbound email, publish to public channel, submit a paymentAlways gate
2 — Destructive internalRemoves or overwrites dataDelete records, overwrite fields, archive itemsAlways gate
3 — High-value transactionFinancial or compliance impact above a thresholdApprove purchase order, grant access, sign contractGate above threshold
4 — Ambiguous interpretationAssignment cannot reliably determine intentCategorizing a one-off case, routing to the right teamGate on low-confidence matches
5 — Routine and reversibleCan be corrected without consequenceTagging a record, updating a status, creating a draftSkip the gate
A practical starting point: start with an approval gate on all Tier 1 and 2 actions. Remove gates only after you have observed that a branch never produces surprises.

Decision criteria by action category

CategoryGate?Notes
Send email to external recipientYesInclude recipient, subject, and full body in the request
Post to a public Slack channelYesInclude the channel name and draft text
Submit a financial transactionYesInclude amount, recipient, and reference number
Modify or delete a customer recordYesInclude the record ID and the proposed change
Reply to an internal threadUsually notGate only if the reply is policy-sensitive
Tag or label a recordNoReversible; iterate without a gate
Create a draft (not sent)NoShow in the approval request instead of gating during creation

Choosing the Right Approval Shape

The Human-in-the-Loop connection supports three request types. Picking the right one keeps operators efficient and prevents ambiguous responses.

Approval (confirm or reject)

Use when the assignment has already determined what to do and just needs a go/no-go before acting. When it fits: The assignment is about to send an email, submit a payment, or delete a batch of records. SOP pattern:
Before sending the email, request approval. Set the title to
"Send to [recipient] — [subject line]". Include the full email body
in the description. Only send after approval. If denied, ask what to
change and revise.

Question (choose from options)

Use when the right next step depends on context the assignment does not have, and the options are well-defined. When it fits: The assignment needs to route a case, choose a tone, or decide between two valid policies. SOP pattern:
If the invoice currency does not match the vendor's default, ask:
(a) Convert to USD at today's rate and continue
(b) Flag for manual review by the finance team
(c) Reject and return to sender with a note
Design rules for good options:
  • Keep option labels short and action-oriented.
  • Make options mutually exclusive — if they overlap, the operator will guess.
  • Do not add an “Other” option unless you also tell the assignment what to do with a free-text answer.

When to use free-text instead

Free-text operator input makes sense when the operator needs to supply content, not just choose a path: dictating a reply, providing context about an anomaly, or overriding a specific field value. Risk: Free-text answers must be parsed by the assignment in its next step. If the assignment expects a decision but gets prose, it may misinterpret. Only use free-text when the assignment’s SOP explicitly handles open-ended input. Common mistake: Using a free-text question where a structured Question would do. Structured options are faster for operators and produce more reliable downstream behavior.

Designing Fallback and Escalation Behavior

Every approval gate needs an explicit answer to “what happens if no one responds?” Without a defined fallback, the assignment stalls indefinitely or makes an unsafe assumption.

Timeout → escalate or fail-safe

Decide up front: if the operator does not respond within a reasonable window, should the assignment escalate or default to the safe action? Escalate pattern:
Request approval with a 4-hour timeout. If not approved within 4 hours,
ask the account manager: [same options]. If still no response after 24 hours,
cancel the action and notify the original requester.
Fail-safe default pattern:
Request approval before sending the refund. If not approved within 2 business
days, cancel the refund and log the reason as "Approval timeout".
Use the escalate pattern when delay is costly (outbound communications, time-sensitive transactions). Use fail-safe defaults when the safe outcome is to do nothing and revisit manually.

Repeated rejection → halt

If the operator rejects the same action multiple times, the assignment has likely misunderstood the requirement. Continuing to loop wastes operator time.
If the approval is denied twice in a row for the same case, stop and
add a note: "Halted after two rejections — requires manual review."
Mark the case as Failed.

Multi-tier escalation

For high-stakes workflows, define a clear escalation chain so a request never hangs waiting for a single person.
Request approval from the team lead. If not answered within 2 hours,
escalate to the department head. If not answered within 24 hours,
escalate to the account manager and halt processing until resolved.
Keep escalation chains short. More than three tiers signals a process that is not well-defined.

Using Case Queue to Manage Exceptions at Scale

When an assignment processes many cases, the Activity Inbox becomes a bottleneck. The Case Queue is the right tool for managing exceptions across high-volume workflows.

Triage pattern

  1. Consumer assignment processes cases automatically.
  2. Cases that need a human decision surface as Needs Input in the queue.
  3. An operator reviews the Needs Input filter, responds in the case detail panel, and the assignment resumes.
  4. Cases the assignment cannot resolve at all land as Failed — the operator reviews, updates the case data, and retries.
This separates routine processing from exception handling: the assignment handles volume, the operator handles judgment calls.

Bulk delegation for surges

When a backlog of Needs Input cases accumulates, you can select multiple cases and delegate them to a specialist assignment — one designed specifically for exception handling — rather than asking one person to respond to dozens of individual HITL requests.

Aging and escalation logic

For cases that sit in Needs Input beyond an acceptable window:
After processing the case, if it moves to Needs Input, postpone it for 24 hours.
On retry, check whether the human has responded. If not, escalate to the team
lead by asking a question: (a) Proceed with the default option, (b) Assign to
a specialist, (c) Close the case.

Ramping Toward Safe Autonomy

The goal is not maximum oversight — it is the right level of oversight. As an assignment matures, you should expect to remove approval gates where they no longer provide signal.

Step 1: Start with gates on every risky branch

When you first deploy an assignment, add approval gates on all Tier 1–2 actions and any branch where you are unsure what the assignment will do.

Step 2: Measure the approval rate

After a few weeks of production traffic, review the pattern of approvals and denials in the Activity Inbox or the Jobs list. Estimate the approval rate per gate.
  • Approval rate above ~95%: The assignment is getting it right consistently. Consider removing the gate and trusting the output directly.
  • Approval rate below ~70%: The assignment is frequently wrong. Refine the SOP before removing the gate.
  • High denial rate with a consistent pattern: The assignment is doing the same wrong thing repeatedly. Update the SOP to correct the root behavior.

Step 3: Remove gates selectively

Remove gates one at a time, in order of confidence. Monitor the next two weeks of output. If quality holds, the gate can stay removed. Do not remove a gate if:
  • The action is irreversible and errors are high-cost.
  • Volume is too low to measure a meaningful approval rate.
  • You have changed something in the connected systems recently.

Step 4: Re-introduce gates when conditions change

Re-add approval gates when:
  • You modify the SOP in ways that could affect the gated branch.
  • The assignment gains access to a new system.
  • Volume increases significantly (edge cases appear at scale that were rare before).
  • Approval rate drops during a quarterly review.
Quarterly cadence: Review approval rates for all assignments in production. Prune gates with consistently high approval rates. Re-add gates on branches that have drifted.

Worked Examples

Expense report approval — threshold-based gate

The assignment auto-handles low-value expenses and escalates high-value ones through a structured flow.
Process each expense report:
— Under $200: approve automatically and update the status.
— $200–$1,000: request approval. Title: "Approve expense — [employee] — $[amount]".
  Description: employee name, department, line items. If denied, return to the
  employee with the reason.
— Over $1,000: ask whether to (a) approve, (b) deny, or (c) escalate to the
  finance director. If no response within 1 business day, escalate automatically.
See the full tutorial: Expense Report Approval

Customer response email — tone check before send

The assignment drafts a reply and holds it for review before sending.
Draft the reply email. Request approval before sending.
Title: "Send reply to [customer name] — [ticket ID]".
Description: the full draft email body.
If approved, send immediately.
If denied, ask: (a) Revise the tone to be more formal, (b) Revise the tone
to be more empathetic, (c) Escalate to the account manager.
Regenerate based on the selected option and request approval again.
See a related example: Reviewing Drafts Before Sending

Supplier follow-up — timeout escalation

The assignment follows up on outstanding purchase orders and escalates if no response arrives.
Send the follow-up email to the supplier. Request approval from the operations
lead before sending.
Title: "Follow-up: PO [number] — [supplier] — [days overdue] days overdue".
If the approval request is not answered within 4 hours during business hours,
escalate to the account manager with the same request.
If still no response after 24 hours, send the email using the default template
and log: "Sent without explicit approval after escalation timeout".