Soundly RunSound familiar?

The support bot got it wrong

One confident wrong answer about a refund, and the customer posts the screenshot before anyone inside has seen it.

InsteadCustomer replies are checked before they go out.

The video

Video planned53 secondsHow every video is made
The script
The one point
Refund replies are checked against current policy, by a check outside the AI, and wait for a person before they send.
Why a video
Time passing: a clock sweeps from a wrong reply to a public post before anyone inside sees it; then the same reply shown held, and a person’s send click.
  1. Intro0:00-0:03sting

    Picture
    Brand mark and sound, then the heading ‘The support bot got it wrong’ on the site’s background, fading straight into the guide with no black.
    On screen
    The support bot got it wrong
  2. The moment0:03-0:10person

    Picture
    Guide in the usual place and framing, speaking to camera. Lower third on first appearance.
    Voice
    One confident wrong answer about a refund. The policy changed in one team while the bot still read the old one, and nothing stood between draft and send.
    On screen
    Guide’s first name, Soundly Run guide
  3. The cost0:10-0:17motion

    Picture
    A clock hand sweeps fast. A reply ‘Full refund within 90 days’ leaves; a screenshot of it appears in a public feed and shares stack up; only then does the support inbox mark it opened. The account’s renewal date dims. Example badge, no outside figure.
    Voice
    The customer posts the screenshot before anyone inside has seen the reply.
    On screen
    Example: posted before anyone inside saw it
  4. The mechanism0:17-0:34screen

    Picture
    Drawn Needs you screen, Example tag top left, context ‘Company crew’. Step band prepare, review, decide. Cursor walks in order: Customer asks, refund on a $14,400 plan; Drafted reply flagged; Check citing refund policy of 1 Sep; Corrected reply $9,600; Why held. Side panel: renewal in 47 days.
    Voice
    A helper drafted: full refund within 90 days. The check sits outside the AI and reads the refund policy of 1 September: prorated after 30 days. Corrected reply: $9,600 for 8 months. Refunds wait for a person.
    On screen
    Example: refund reply held. It quotes the old policy.
  5. The decision0:34-0:42person and screen

    Picture
    Send corrected reply, Change and Pass visible; guide in picture-in-picture. Cursor on Send corrected reply, click held two seconds. Rows turn green: ‘Sent: $9,600 prorated refund, policy of 1 Sep’, ‘Old wording retired from the sources’.
    Voice
    The support rep decides: send corrected reply, change it, or pass. The account owner is told when they decide.
    On screen
    Sent with your approval.
  6. The difference0:42-0:49person

    Picture
    Guide back in the same place and framing as The moment.
    Voice
    Nobody has to remember to tell the bot when a policy changes. The helper drafts; a person decides on costly topics. It does not watch, score or replace anyone.
    On screen
    Customer replies are checked before they go out.
  7. Outro0:49-0:53sting

    Picture
    Standard end card, the same in every video: the page title, the brand mark and one small line, ‘Example screens’.
    On screen
    The support bot got it wrong

On screen as proof

  • Example tag on the drawn Needs you screen throughout the mechanism
  • Drafted reply ‘Full refund within 90 days of purchase’ flagged against the refund policy of 1 Sep
  • Send corrected reply click held two seconds, then ‘Sent with your approval.’
  • No outside figure exists for this page, so the cost is a sweeping clock, tagged Example, with no outside figure

Before and after

Before
The refund policy changed; nobody owned telling the bot
After
Replies cite the current policy, with its date
Before
Escalation only after the customer complains publicly
After
Refunds, pricing and terms go to a person first
Before
One bad answer becomes silent churn risk
After
The account owner sees it the same day

How it works

Every customer reply is checked against the written policy before it sends; costly topics wait for a person.

  1. The head of support decides which topics are costly; those replies wait for a person.
  2. Each reply is checked against the current policy from outside the AI, so a cleverly worded message meets the same check.
  3. If a wrong reply gets out, the old wording is retired and more replies wait for a person; the account owner hears that day and decides the follow-up.
Needs youExample

Company crew

Refund reply held. It quotes the old policy.

Customer asks
Refund on a $14,400 annual plan, month 4
Drafted reply
Full refund within 90 days of purchase
Check
Refund policy of 1 Sep: prorated after 30 days
Corrected reply
$9,600 prorated refund for 8 months, paid within 5 days
Why held
Costly topic: a person approves

Source: help desk and refund policy, 14 Oct

Send corrected replyChangePass

Built to work in the email, calendar, chat, finance and sales tools you already use.

What it never does

  • Acts alone only inside limits a person sets; asks first for anything else.
  • Reaches only what its job needs.
  • Never watches, scores or replaces anyone.