Responding faster to customer requests

Specialty retailer with online sales

Decision studied

Reduce first-response time without lowering satisfaction.

The retailer, which has 55 employees, received 2,400 requests a month. Order tracking and product questions overwhelmed six advisors, and a contractor charged for part of the handling on demand.

Approach

  1. 01

    Frame

    The director wanted to cut first-response time by 25% without lowering satisfaction. The project covered three frequent request types; disputes and refunds stayed with the advisors.

  2. 02

    Investigate

    I reviewed 800 requests, the contractor’s invoices and the service rules. I measured the full handling time, including corrections, separating simple requests from sensitive cases.

  3. 03

    Explain

    Two causes were compared: too few staff, or scattered information. Repeated searches explained delays more than the number of new requests did.

  4. 04

    Build

    We compared an improved FAQ, an assistant that prepares sourced answers and an autonomous agent, with the existing setup as the financial baseline.

  5. 05

    Test

    A six-week test split 400 comparable requests between two groups. The thresholds were 95% of answers accepted without major correction and a 25% reduction in full handling time. The test reached 96%, and full handling time dropped from 7 to 4 minutes.

  6. 06

    Decide

    The director chose drafts reviewed by the advisors. The pilot was run on a budget of $16,500, including $5,500 for strategy. The drop in subcontracting must be confirmed on the invoices before any expansion.

  7. 07

    Steer

    Each week, the manager tracks errors, satisfaction and the requests still billed by the contractor. A request type is paused if major corrections exceed 5% for two weeks.

Test results

Threshold: 5.25 min (25% drop)7 min4 min−43%Before the testDuring the testAnswers accepted without major correction: 96% (threshold: 95%).
Full processing time per request, corrections included, across 400 requests split between two groups. Threshold set: 25% reduction, or 5.25 minutes.

What AI contributed

AI helped me group the request types and prepare the test. In the service, it suggests answers based on approved documents; the advisor validates each response before it goes out.

Business model (projection)

At steady state, 9,000 requests a year leave the contract, at $4 each, for $36,000 in avoided spending. The staff time freed up is not counted as savings. In the prudent scenario, only 4,500 requests leave the contract.

Annual projection for this case
Initial investment$16,500 CAD
Annual gains$36,000 CAD
Costs to subtract$9,600 CAD
Net annual gain$26,400 CAD
Simple payback7.5 months (at steady state)
Prudent scenario$8,400 CAD net, payback in 23.6 months

Annual projection built from the results of the engagement. It is not a guarantee of performance.

A comparable challenge?

Describe the decision to make and the data available. I will suggest an approach to prepare it and the evidence to gather before deciding.

Let's talk strategy