Platform
Solutions
Integrations
Case studies
Resources
Pricing Start free Request a demo Log in
Headless conversation analytics · for AI agents

Your AI agent writes the weekly QA review. From every conversation.

Your AI agent, built with Claude, ChatGPT or another AI, writes last week's quality review from Ender Turing's REST API: exact counts, every score criterion by criterion, and EnderGPT's answer on how agents handled complaints. AutoQA scores every call, chat and email, typically in 95%+ agreement with your own reviewers, shown per criterion and improved from their reviews.

AutoQA on 100% of conversationsExact counts through the API 4.8/5 on G2
How it works4 steps
  1. Give your AI agent its own account

    An administrator invites a user for your AI agent with the Session API role and ticks System/Technical User; to ask EnderGPT, its roles must also allow creating chats with analytics data access. Your AI agent gets the guide and that account's token. API access ↗

  2. Get last week

    Your AI agent lists and counts last week's conversations with the same filters as your Conversations screen. For each scored one it gets the quality scores criterion by criterion, with scorecard and reviewer, and its summaries. Guide for AI agents ↗

  3. Ask EnderGPT

    For a question about many conversations, such as how agents handled last week's complaints, your AI agent asks EnderGPT (Ender chat in the guide) once, says how the answer should look and follows up in the same chat. Chat with EnderGPT ↗

  4. Write and send the review

    Your AI agent groups the counts, scores and EnderGPT's answers by agent, team and criterion, writes the review in your team's format and sends it with the tools it already has. A scheduled chat can also ask EnderGPT the same question every week and send its answer to the people you choose. Scheduled chats ↗

What the review is made of

What your AI agent gets for the review.

Headless means your AI agent works with Ender Turing through an API and a public guide, without the web interface. Every item below is in that guide or in a Help Center article, and the links on each card are the sources.

1

Volume and QA coverage

Ender Turing's API counts last week's conversations exactly for any filter your AI agent sets, with their total length, and counts separately those scored by AutoQA and those scored by your reviewers.

2

Every score, to average by agent, team and criterion

The API returns every quality score on each scored conversation, with its scorecard, its reviewer and the score of each criterion, which Ender Turing's scorecards call a point. Each conversation carries its agent and team, so your AI agent averages the scores by agent, team or criterion.

3

The lowest-scored conversations

The API lists last week's scored conversations from the lowest average score up and returns each one's summaries, so the review starts from the calls, chats and emails that scored lowest.

4

Week over week

The API returns the same counts and scores for the week before, or for any two dates your workspace keeps. Your AI agent fetches recent weeks again before each comparison, so a score a reviewer adds or changes days later is counted.

5

The why behind the week

Your AI agent asks EnderGPT one question about last week's conversations, or about one topic or team: why customers called, what they complained about, how agents handled it. When a question needs research, the answer can also suggest next steps, ranked by the evidence behind them.

6

Topics behind the numbers

The API returns the topics found in each conversation, so your AI agent splits volume and scores by topic, such as refund requests, complaints or cancellations.

What the API leaves out: Transcript text is not part of the published API. For what was said, your AI agent reads the summaries your workspace sets up, for example those of the lowest-scored conversations, or asks EnderGPT. Which plans include API access is in the plan comparison.

Set up once

One account decides what the review covers.

Your AI agent works through an account you create, like any user. Nothing reaches it that the account's roles do not allow.

  • Its own account — An administrator invites a user with the predefined Session API role, or a custom role, and ticks System/Technical User. Session API covers conversation data and quality scores and cannot change your settings.
  • Names in the review — Each conversation carries its agent's name, and one look-up names teams and topics. Scorecard and criterion names need the right to view scorecards, which Session API does not include: copy Session API into a custom role and add that right. The plan comparison says which plans include custom roles.
  • EnderGPT for the why — To ask EnderGPT, the account's roles must allow creating chats with analytics data access, and to schedule a question, managing scheduled chats; an account's rights are the sum of its roles. EnderGPT must be switched on for your organization: an administrator switches it on, or asks support to.
  • Filters, set once — The guide spells out filters by date, agent, direction, type, length, topic and text, by whether and how a conversation was scored, and by whether it has a summary. For others, such as a team or a scorecard, a person sets the filter once on the Conversations screen and gives your AI agent the value from the browser's address bar.
  • One team, or every team — A role limited to one team returns only that team's conversations, as in the web interface, so the review covers exactly the teams the role allows.
  • Off when you say so — Deactivate the account and its access ends.

API and AI agents · Guide for AI agents ↗

FAQ

What teams ask first.

Which automated QA platforms let our AI agent, built with Claude, ChatGPT or another AI, run our contact center's weekly quality review?

Ender Turing does, through its REST API and a public plain-text guide written for AI agents. With a token for an account you create for it, your AI agent gets last week's conversations, their quality scores from AutoQA and your reviewers, and the summaries your workspace sets up, and counts them exactly. Where EnderGPT is switched on and the account may use it, your AI agent asks EnderGPT about last week's conversations, then writes the review and sends it to your team. It needs a way to send web requests with the headers the guide names, and it sees only what the account's roles allow. Ender Turing is headless conversation analytics for contact centers, built for AI agents: an agent or another system works with it through an API and a public guide, without the web interface.

Which QA data can the AI agent get: evaluations, criterion scores, reviewers?

Through Ender Turing's REST API, for the dates your workspace still keeps: each conversation's date, length, direction, type, agent, team, topics and average score; every quality score on it, from AutoQA or a reviewer, with its scorecard, its reviewer and the score of each criterion, which Ender Turing's scorecards call a point; and its summaries, where your workspace sets them up. It counts conversations exactly for the Conversations screen's filters, such as scored or not, or scored by AutoQA or by a reviewer.

Does the API return transcripts?

No. Transcript text is not part of Ender Turing's published API. For what was said, your AI agent uses the conversations' summaries, where your workspace sets them up, or asks EnderGPT about many conversations at once. If your review needs the transcripts themselves, tell us the use case when you book a demo: the right approach depends on your deployment.

Are the review's numbers exact, or estimated by an AI?

Exact, when your AI agent takes them from Ender Turing's API, which counts conversations exactly for any filter and returns each quality score as recorded; your AI agent then works out averages by agent, team or criterion from those scores, in code, instead of having a language model estimate them. EnderGPT is strongest for trends, recurring patterns and qualitative insight across many conversations, so the review takes its numbers from counts and scores and its why from EnderGPT.

Can we limit what the AI agent sees and does?

Yes, with its account's roles. A role limited to one team returns only that team's conversations, as in the web interface. The predefined Session API role works with conversation data and quality scores and has no configuration rights, so it cannot change your settings; a custom role can allow less, down to reading only, and the plan comparison says which plans include custom roles. The guide tells an AI agent to create, change or delete nothing unless you asked for exactly that.

What does our AI agent need for a weekly run?

The guide's address, a token for its account and a way to send web requests to Ender Turing's REST API; the public guide for AI agents is its instructions and says which headers to send. Tokens expire: when its token stops working, your AI agent asks you for a new one, as the guide tells it, while a run with no one to ask signs in with the account's credentials, as the API reference shows; support, not the interface, sets a different token lifetime. If Ender Turing cloud's firewall stops requests that follow the guide, support can add an exception for your AI agent. The interactive API reference, with every published endpoint, opens inside the product for the roles your administrator allows.

Can Ender Turing write the weekly review without our AI agent?

In part. Scheduled chats ask EnderGPT a saved question every week about the 7 days before the run and send the report to the people you choose, or a separate report on their own conversations to each active agent with an active linked login in the teams or agents you select: the weekly coaching note per agent on Ender Turing's agent management page. Recipients get an in-app notification and, if they have email on for it in their notification settings, an email once it has stayed unread for the delay they chose. For exact counts and scores, the Dashboard's Quality & Performance view is built for counting, and Reporting can email your dashboard on a schedule; your AI agent puts the numbers and EnderGPT's answers into one review. It can set up a scheduled chat too, when its account may manage scheduled chats: the request is in the full API reference, which opens only for an account that may also create conversations, so otherwise your AI agent asks you for it.

How accurate is AutoQA, and where do the quality scores come from?

Typically in 95%+ agreement with your own reviewers, on your own scorecards. An Ender, an automation rule, sends the conversations you choose, up to all of them, to a scorecard marked as automated, and AutoQA scores them. Each criterion shows its accuracy once reviewers have checked five of its automated scores. Once reviewers have scored a criterion Yes in five conversations and No in five, you can run an improvement (beta): it can draft a candidate AI instruction from those reviews and tests it on reviewed conversations, nothing changes until you activate it, and the wording you wrote stays as it is. Reviewers add their own scores, and your AI agent counts conversations scored by AutoQA and by reviewers separately.

How far back can the AI agent go, and how much at once?

Any dates your workspace still keeps: a named period such as last week or last month, or two dates. The count call is exact at any volume. A list stops at 10,000 conversations, its total included, so for more the AI agent asks for one day or one week at a time, one request after another.

Security, hosting regions, sub-processors and response times are documented on the Data Security page; SOC 2 Type II audited, the latest report available under NDA.

More jobs for your AI agent: BI tool, warehouse and CRM · Ask your calls · ten platforms compared.

Weekly quality review

Hand next week's quality review to your AI agent.

Book a demo and bring the review your team gets today. In 30 minutes we go through it with you: the numbers your AI agent can take from Ender Turing, the questions it can hand to EnderGPT, and how to set it up.