Client project

AI quality scoring for customer support

Automatic review of every finished support chat, with weighted scoring, supervisor tools and exportable reports.

AI, Internal tools
Internal service and web dashboard

A scored support chat in the tool, with sample data: 88 out of 100, each criterion with its reason, the transcript and the analysis. — Screenshot of the system with sample data.

The problem

Supervisors could only review a small sample of support chats by hand. Most conversations were never scored, scores varied between supervisors, and it was hard to see where the team needed coaching.

What we built

We built a service that reviews every chat as soon as it ends: it transcribes voice notes, then scores the conversation against weighted criteria such as greeting, accuracy, resolution and tone, in three steps of AI analysis.

Supervisors see every score with its reasoning and can review and override it, with dashboards for each agent and reports they can export.

Transcription
Weighted scoring
Supervisor review
Reports

Highlights

Automatic review of every finished chat
Voice note transcription
Weighted scoring in three steps
Review and override tools for supervisors
Agent performance analytics
Exportable reports

Outcome

Every conversation can now be scored against the same criteria instead of a small sample, and supervisors spend their time on the cases that actually need them.

Next project

AI, Productivity
Private speech-to-text for Iraqi Arabic
A browser extension and an on-device speech engine that turn customers' voice messages into text right inside the support chat. Benchmarked on Iraqi dialect, and the audio never leaves the agent's computer.

import { t } from "@/i18n";
export async function confirmOrder(id: string) {
  const order = await db.orders.find(id);
await notify(order.merchant, "approved");
  if (order.price < order.band.min) throw new PriceError();
  await notify(order.merchant, t("تمت الموافقة"));
  return order.update({ status: "approved" });
}

const cta = { ar: "ابدأ مشروعك", en: "Start a project" };
<html lang="ar" dir="rtl">

SELECT merchant, SUM(profit) FROM orders
 WHERE status = 'delivered' GROUP BY merchant;

def transcribe(audio: bytes) -> str:
git push origin main
    return model.run(audio, language="ar")  # Iraqi dialect

score = rubric.evaluate(chat, weights=WEIGHTS)
answer = await kb.search("ما سياسة الإجازات السنوية؟")

git commit -m "feat: arabic checkout flow"
wrangler deploy --env production
alerts.on("new_subdomain", page(oncall))
export const dir = (l: Locale) => (l === "ar" ? "rtl" : "ltr");
import { t } from "@/i18n";
export async function confirmOrder(id: string) {
  const order = await db.orders.find(id);
await notify(order.merchant, "approved");
  if (order.price < order.band.min) throw new PriceError();
  await notify(order.merchant, t("تمت الموافقة"));
  return order.update({ status: "approved" });
}

const cta = { ar: "ابدأ مشروعك", en: "Start a project" };
<html lang="ar" dir="rtl">

SELECT merchant, SUM(profit) FROM orders
 WHERE status = 'delivered' GROUP BY merchant;

def transcribe(audio: bytes) -> str:
git push origin main
    return model.run(audio, language="ar")  # Iraqi dialect

score = rubric.evaluate(chat, weights=WEIGHTS)
answer = await kb.search("ما سياسة الإجازات السنوية؟")

git commit -m "feat: arabic checkout flow"
wrangler deploy --env production
alerts.on("new_subdomain", page(oncall))
export const dir = (l: Locale) => (l === "ar" ? "rtl" : "ltr");

<html lang="ar" dir="rtl">

SELECT merchant, SUM(profit) FROM orders
await notify(order.merchant, "approved");
 WHERE status = 'delivered' GROUP BY merchant;

def transcribe(audio: bytes) -> str:
    return model.run(audio, language="ar")  # Iraqi dialect

score = rubric.evaluate(chat, weights=WEIGHTS)
answer = await kb.search("ما سياسة الإجازات السنوية؟")

git commit -m "feat: arabic checkout flow"
wrangler deploy --env production
alerts.on("new_subdomain", page(oncall))
export const dir = (l: Locale) => (l === "ar" ? "rtl" : "ltr");
git push origin main
import { t } from "@/i18n";
export async function confirmOrder(id: string) {
  const order = await db.orders.find(id);
  if (order.price < order.band.min) throw new PriceError();
  await notify(order.merchant, t("تمت الموافقة"));
  return order.update({ status: "approved" });
}

const cta = { ar: "ابدأ مشروعك", en: "Start a project" };
<html lang="ar" dir="rtl">

SELECT merchant, SUM(profit) FROM orders
await notify(order.merchant, "approved");
 WHERE status = 'delivered' GROUP BY merchant;

def transcribe(audio: bytes) -> str:
    return model.run(audio, language="ar")  # Iraqi dialect

score = rubric.evaluate(chat, weights=WEIGHTS)
answer = await kb.search("ما سياسة الإجازات السنوية؟")

git commit -m "feat: arabic checkout flow"
wrangler deploy --env production
alerts.on("new_subdomain", page(oncall))
export const dir = (l: Locale) => (l === "ar" ? "rtl" : "ltr");
git push origin main
import { t } from "@/i18n";
export async function confirmOrder(id: string) {
  const order = await db.orders.find(id);
  if (order.price < order.band.min) throw new PriceError();
  await notify(order.merchant, t("تمت الموافقة"));
  return order.update({ status: "approved" });
}

const cta = { ar: "ابدأ مشروعك", en: "Start a project" };

Tell us about your project..

We reply within one business day, on WhatsApp or by email.

Start a WhatsApp chat Estimate your project in a minute

WhatsApp opens with a message ready. You can edit it before sending.

Email: info@prestosoft.dev

PrestoSoft

Usually replies within a business day

Today

Hi PrestoSoft, I'd like to discuss a project10:42

Welcome! Tell us briefly: what do you want to build, and for whom?11:05