Separate Permission to Investigate From Permission to Commit
Prepared with AI assistance by Alactic Editorial. Examples identified as hypothetical are illustrative, not customer results.
“See what we can do for this customer” can mean collect facts, build alternatives, or recommend an exception. It should not silently become permission to offer a discount, accept a condition, or tell the customer that a decision is final.
Separate the work into four stages: fact gathering, option development, recommendation, and commitment. Give the assistant a defined responsibility at each stage, name the person who can advance the decision, and specify which customer-facing statements remain off-limits. Permission to investigate should produce a decision-ready brief—not an implied promise.
This is a boundary of business responsibility. An assistant might be able to send a message while lacking authority to make the promise in it. Conversely, it can investigate a proposal thoroughly without owning the decision. The worksheet below makes that distinction usable in an ordinary customer-exception request.
1. Define the decision before delegating the investigation
An open-ended task invites the assistant to infer what a helpful outcome looks like. If the instruction is to keep a customer happy, granting the requested exception can appear to be the most direct route. That is an objective, however, not a delegation of commercial authority.
Start with the decision that someone will eventually need to make. Describe the requested departure from the normal arrangement, the information needed to assess it, and who owns the answer. Then say explicitly whether the assistant can only prepare the decision or also communicate an approved outcome.
A useful delegation contains four elements:
- Question: What exactly are we investigating?
- Permitted work: Which facts, alternatives, and assessments should the assistant develop?
- Boundary: What must it not offer, accept, waive, or represent as decided?
- Handoff: Who receives the result, and what decision do they need to make?
The boundary should cover meaning, not just particular words. Prohibiting the word “approved” is insufficient if the assistant can still say something that reasonably implies the customer will receive the exception. A courteous reassurance can become a business promise even when it contains no formal approval language.
2. Give each stage a different output and authority boundary
Fact gathering: establish what is known
The assistant collects the request, relevant account history, existing terms, operational constraints, and missing information. Its output separates confirmed facts from estimates and unresolved questions. A customer saying they were promised something is a fact about the customer's report, not confirmation that the business made that promise.
The boundary is straightforward: asking whether an arrangement is possible does not reserve capacity, grant an exception, or acknowledge an obligation.
Option development: describe feasible choices
The assistant constructs alternatives, including keeping the existing arrangement. Each option should show its assumptions, resource implications, customer impact, and dependencies. An option is not an offer merely because it is detailed enough to become one.
Keep this work internal unless a human has authorised a specific exploratory discussion. Presenting tentative terms to a customer can shape expectations before anyone has chosen them.
Recommendation: advise without deciding
The assistant identifies a preferred option and explains the trade-off. It can make a strong recommendation without claiming authority to adopt it. The output should name the decision needed and any unresolved dependency that could change the recommendation.
Do not treat a confident recommendation as a substitute for a decision. Confidence concerns the quality of the assessment; authority concerns who may commit the business.
Commitment: approve defined terms and represent the decision
The authorised person chooses the terms, conditions, scope, and validity of the offer. Communicating that choice is a separate responsibility. A person may approve an exception but still want to send the message themselves.
Where an assistant is authorised to communicate an approved offer, it should convey that offer—not negotiate further concessions. Customer acceptance, a counterproposal, and an internal approval are different events and should remain distinguishable.
3. Filled responsibility worksheet: a customer requests free onboarding
Fictional worked example: All businesses, amounts, estimates, and circumstances in this worksheet are hypothetical. A business-services supplier has completed the onboarding included in a customer's package. The customer requests six additional onboarding hours at no charge because more staff need training. An account manager asks an assistant to investigate the request.
The following is a recommended operating worksheet, not a description of an existing product control.
Assignment and decision ownership
- Decision to prepare: Whether to provide additional onboarding, how much to charge, and under what scope and scheduling conditions.
- Assistant's assignment: Gather facts, develop alternatives, and recommend an option. Do not make an offer or accept customer conditions.
- Case coordinator: Account manager. Resolves missing account information and receives the investigation.
- Business decision owner: Commercial manager. Approves or rejects the exception and its exact commercial terms.
- Operational contributor: Service lead. Confirms whether the proposed work can be accommodated. Capacity confirmation does not itself approve free work.
- Customer communication owner: Account manager. Sends the final offer unless a separate, explicit instruction authorises the assistant to send the approved wording.
Stage A: fact gathering
- May explore: What training the customer needs; who will attend; what onboarding has already been provided; documented package scope; indicative staff availability; and whether there is evidence of an earlier promise.
- Confirmed for this hypothetical case: Included onboarding is complete. The customer requests six additional hours. The standard additional-service price is $100 per hour.
- Planning estimate: Internal delivery cost is $60 per hour, making the estimated delivery cost of six hours $360. This is not a full profitability calculation.
- Unresolved: The customer has not identified all attendees. The service lead has not confirmed dates. No earlier promise of free additional hours has been found in the records reviewed.
- Permitted customer contact: Ask about training topics, attendees, and preferred timing. State that the request is under review.
- Requires a human decision: Acknowledging that free work is owed, reserving a delivery date, or suggesting the exception is already agreed.
- Exit output: A fact summary with sources, estimates, and unresolved questions clearly separated.
Stage B: option development
- Option A: Six additional hours at the standard hypothetical price of $600. Estimated internal delivery cost: $360.
- Option B: Two hours at no charge and four paid hours, for a hypothetical customer charge of $400. Estimated internal delivery cost remains $360.
- Option C: All six hours at no charge. Hypothetical customer charge: $0; estimated internal delivery cost: $360.
- May explore: Whether a shorter session could address the main training need, how each option affects service capacity, and what scope would make the work manageable.
- Must not do: Present any option as available to the customer, imply a concession will follow if they agree to something else, or remove scheduling dependencies.
- Exit output: An internal options brief showing the customer charge, estimated effort, uncertainties, and trade-offs.
Stage C: recommendation
- Recommended option: Option B, subject to service-lead capacity confirmation. It shares the additional cost while offering a limited accommodation.
- Reasoning: The reviewed records do not establish a prior promise of free additional training. A partial concession may address the relationship concern without granting the entire request.
- Reason to choose differently: New evidence of an earlier promise, a different training need, or unavailable capacity would require reassessment.
- Decision requested: Commercial manager to approve, amend, or reject the precise concession. Service lead to confirm a workable scheduling basis.
- Boundary: The recommendation remains internal and does not change the customer's arrangement.
Stage D: commitment
- Hypothetical human decision: The commercial manager approves Option B after the service lead confirms capacity.
- Approved terms: Six additional onboarding hours covering the agreed training topics; two hours free and four hours charged at $100 each; session dates to be mutually agreed within the next 30 days.
- Offer validity: The account manager may communicate these terms as available for acceptance for five business days. These periods are hypothetical business choices, not external requirements.
- Communication responsibility: The account manager sends the offer. The assistant may prepare a draft but has no authority to send it under this worksheet.
- Return for a new decision: The customer asks for more free hours, a lower charge, additional topics, or guaranteed dates not already confirmed.
- Completion evidence: Record the approved terms, the message actually sent, and the customer's response separately. Do not mark the exception accepted merely because the internal approval exists.
4. Keep exploratory language from becoming a soft offer
The risky moment often comes before formal approval. An assistant trying to reassure the customer may describe an internal possibility as something the business can provide.
In the fictional onboarding case, “We can include the extra training” crosses the boundary. “We are reviewing your request for additional training” describes the current state without predicting the outcome. Likewise, asking which dates would work is different from saying a trainer will be available on those dates.
Use a simple test before customer-facing communication: Could a reasonable reader understand this sentence as a decision, entitlement, price, or promise? If yes, either obtain the relevant authority or rewrite it as a question or status update.
Conditional language is not automatically safe. In this hypothetical case, “We can provide the free hours if you renew” would introduce a proposed bargain, even though it contains an “if.” Exploring the customer's priorities does not authorise the assistant to create that bargain.
5. Make approval specific enough to survive a handoff
A short approval such as “fine” can leave unresolved which option, scope, and conditions were accepted. Before treating it as authority to commit, resolve the ambiguity with the decision owner.
A useful approval record identifies:
- The selected option and exact concession.
- The work included and excluded.
- Any capacity or timing dependencies.
- How long the offer remains available.
- Who may communicate it.
- Which customer changes require another decision.
For the hypothetical worksheet, approval of two free hours does not authorise six free hours. Approval of a customer charge does not confirm specific delivery dates. Approval to draft a message does not authorise sending it.
This precision adds work at the approval point, but reduces interpretation later. Keep the record short enough to read quickly; the goal is a clear decision boundary, not a lengthy approval ritual.
6. Balance investigation speed against customer expectations
Requiring approval for every factual question makes investigation slow. Letting every investigator discuss possible concessions freely creates a different problem: the business can appear committed before it has assessed the request.
A practical middle ground is to predefine the questions and status updates the assistant may send. Leave exploratory commercial proposals internal unless someone explicitly authorises discussing them. This gives the assistant room to make progress without making the customer negotiate against unapproved possibilities.
Also distinguish information authority from decision authority. The service lead may be the best person to assess effort, but that does not mean they own the commercial concession. The account manager may understand the relationship, but may still need another person's approval for free work.
Where routine exceptions are frequent, the business can choose to delegate a narrow class of decisions. That is a new, explicit commitment mandate with defined conditions—not an interpretation of a general instruction to investigate.
7. Recover when the boundary has already been crossed
If the assistant has implied approval, stop treating the case as an ordinary investigation. Preserve the exact statement, identify who received it, and bring it to the decision owner. Do not quietly change the internal status and assume the customer's expectation has disappeared.
- Establish what happened. Separate the customer's request, the assistant's statement, any human approval, and the customer's response.
- Assess the expectation created. Identify the price, scope, timing, or condition the customer may now believe was agreed.
- Get a human decision. The owner decides how to address the mismatch. The assistant should not invent a second concession to repair the first.
- Clarify with the customer. A human-approved message should acknowledge the discrepancy and explain the actual position without falsely suggesting the earlier statement was authorised.
- Repair the delegation. Add the missing boundary or handoff to the worksheet and review similar open cases.
For an ambiguous internal approval that has not yet reached the customer, recovery is simpler: pause the commitment stage and ask the approver to specify the intended terms.
8. Test the boundary on the next exception
For the next customer-exception request, complete the worksheet before the investigation begins. Name the decision owner, state what the assistant may explore, and write down what would count as a commitment.
After the case closes, review whether each stage produced its intended output. Could someone locate the fact summary, alternatives, recommendation, approval, and customer response without inferring one from another?
Across the next few cases, track unclear-approval handoffs, customer-facing statements made before approval, and cases reopened because the customer requested different terms. Also note where investigation stalled unnecessarily. These observations show whether the boundary is both understandable and workable.
The target is not more approval for its own sake. It is useful investigation that reaches the right decision-maker without making the decision on the way there.
Put the next conversation to work.
Explore Alactic Communications for WhatsApp conversations, campaigns and AI-assisted workflows. Choose the product and plan that fit your team.
Explore Communications