Getting Started with Automations
What is an automation?
An automation is an AI analyst that works through your team's calls on a schedule you set and delivers what it finds to the people who need it. You tell it what to look for in plain language, such as a recurring objection, a missed compliance step, a competitor mention, or a great call worth celebrating, and it sends a short, evidence-backed report with quotes and links to the exact call moments.
Who can access Automations
Access follows your teammate role:
Editor: Full access. Can create, edit, run previews, and delete Automations.
Reviewer: View only. Can see Automations and their results but cannot create or edit them.
Team Member: No access. The Automations section isn't shown to them at all; they only ever receive individual reports addressed to them.
Template vs. custom automations
Templates are pre-built automations for the most common use cases, with the analysis already written for you. Start with a template when one fits, since you only need to pick the playbooks, schedule, and recipients.
Weekly coaching goals: Use this to give every teammate a focused weekly goal with midweek progress check-ins, and give managers a Monday digest of their team's goals.
Customer praise digest: Use this to share the moments customers go out of their way to praise your product, service, or team, in their own words.
Perfect calls: Use this to surface calls with a 100% playbook score so the whole team can learn from what excellent execution sounds like.
Playbook feedback alert: Use this so managers hear about new grading feedback right away and can keep your playbook accurate.
Playbook score movement alert: Use this to see which criteria scores moved week over week, which teammates drove the change, and what to do next.
Lead burning alert: Use this to catch teammates moving through live leads without meaningful conversations so managers can step in quickly.
For any use case a template does not capture, choose Custom automation. A custom automation looks for whatever you describe, on the calls you choose, and delivers the result in the format and to the people you specify.
Intro to custom automations
A custom automation is built from five settings. Each one is set on the automation's configuration page, and you can change any of them later.
Setting | What it controls |
|---|---|
Which calls should it analyze? | The playbook or playbooks whose calls the automation reviews. Pick more than one to look across them in a single result. |
Check frequency | How often the automation runs: hourly, daily, weekly, or monthly, at the time you choose. Hourly automations only run when there is new call activity. Each run covers the calls since the previous run. |
What should this automation look for? | Your instructions. This is the part that decides the quality of the report. |
Result type | Whether everyone gets one shared result, each person gets their own, or your instructions define the results. |
Delivery channel and recipients | Where results are sent (Email, Slack, or Microsoft Teams) and who receives them. These options change with the result type. |
What data a custom automation can use
A custom automation can use the same data your team sees in Alpharun for the playbooks it watches. The more precisely your instructions point at this data by name, the sharper the report. Everything below can be used both to narrow which calls the automation looks at and to shape what goes in the report.
Data | What the automation can do with it |
|---|---|
Playbook scores | Read section and criteria scores on each call (Yes/No and 1 to 5), filter to calls that passed or missed a criterion, see team-level scores, and rank teammates by performance. |
Custom fields (contact and customer interaction) | Filter and group calls by any custom field on the linked contact (for example state, product, or lead source) or on the customer interaction itself (for example campaign, queue, or disposition). Text, number, and date fields are all supported. |
Data extraction fields | Use the values Alpharun's AI extracts from each transcript, such as quoted price, competitor named, or appointment date, as filters and as data points in the report. |
Signals | Find calls where a signal you have configured was detected and read the moment that triggered it. Reference signals by their exact name in your instructions. |
General | Transcripts, call date, call duration, outcomes, topics, customer experience (CX) score, vocal intelligence, knowledge base documents, and coaching notes. |
Teammates | Limit the analysis to specific teammates, compare performance, and name who a finding is about. |
Managers | Roll teammate results up to each manager's team and draw on manager coaching notes. |
Teammate groups | Limit the analysis to a group, or segment findings by group, such as by site, shift, or team. |
Playbook feedback | Read grading feedback submitted by reviewers, including open feedback and its history, and the criteria each piece of feedback applies to. |
A custom automation only sees data that already exists in your workspace. If your use case depends on a pattern that no signal or field captures yet, create and apply that signal or field first, then build the automation.
Writing "What should this automation look for?"
The automation sees your calls and your instructions, and nothing else, so write the instructions so they stand on their own. A strong write-up has these sections:
What to look for: The pattern, risk, opportunity, or theme to watch for, described in terms that can be heard on a call, such as a phrase, an objection, a missed step, or a signal by name. Say what does not count too, so a quiet period produces nothing instead of filler.
What I want to receive: What each result should include, such as the teammate, a short explanation of what happened, a quote from the call, and the action the reader should take. If you want the results presented in a specific format, spell that out too, for example a summary table followed by a list of customer interactions grouped by rep.
Who receives it (Custom result type only): Who each result is about, who gets it, and where it is sent. See the Custom result type below.
For example:
What to look for: Calls on the Sales playbook where the customer mentions a competitor by name. Only include mentions where the customer is actively comparing options. Skip passing references and calls where the teammate brings up the competitor first.
What I want to receive: For each call, the teammate's name, the competitor mentioned, the customer's quote, how the teammate responded, and whether the customer moved forward. End with the two or three most common comparison points and one suggestion for how the team can handle them better.
Choosing a result type
The result type decides who each report is about and who receives it, and the delivery options below it change to match.
Result type | What it produces | Delivery options |
|---|---|---|
One shared result | One report that everyone receives. Best for team-wide digests. | Email the teammates and groups you select under Notify, or post to a Slack or Microsoft Teams channel, or send a direct message to selected recipients. |
One result each | A report for each selected teammate about their own calls. Managers and groups get an aggregated result across their teammates' calls. Best for coaching and performance insights. | Private delivery only: Email, Slack DM, or Microsoft Teams chat, to each person you select under Result recipients. |
Custom | Whatever results your instructions describe, routed however your instructions say. Best when you need a mix, such as a team digest plus individual results. | None to configure. Delivery channel and Notify are removed, and your instructions define the routing. |
To see how the choice plays out, take a competitor mention automation. If product marketing wants a weekly roundup of every competitive comparison, choose One shared result and post it to your competitive intelligence Slack channel. If you want each rep to see how they handled competitor comparisons on their own calls, with managers getting a view across their team, choose One result each. If you want both at once, a team digest in a leadership channel and a private summary for each rep, choose Custom.
Writing delivery instructions for the Custom result type
With the Custom result type, the Delivery channel and Notify settings are no longer shown, so your instructions must say who receives each result and where. You can be as detailed as you want, for example "send each rep their own summary and copy their manager" or "post one weekly digest in #revenue and DM each manager a summary of their team."
For Slack, include the exact channel name (for example #sales-leadership) or the Slack username of each teammate who should get a direct message. Run a preview after saving to confirm the automation read your routing the way you intended.
Learn how to connect with Slack
Best practices for building and testing custom automations
Build and iterate with Ask Alpharun
Your Ask Alpharun automation is trained on the best practices for building automation automation prompts. Describe what you want in your own words and it translates your idea into a well-structured prompt, taking you from idea to draft in minutes. It can also recommend the right template, check your call volume against the schedule you picked, and find the Slack or Teams channel you want to deliver to.
Want to make tweaks? Just tell Ask Alpharun what you want changed on any automation automation and it will spin up a draft for you to review. Nothing changes until you apply the draft, and applying it never turns the automation on, so enabling it is always your decision.
Test your custom automation with Run preview
Once you have a draft, click Run preview on the automation's page to generate a sample report from your historical calls. A preview runs the automation exactly as it would on schedule but notifies no one, so you can review the output and refine the instructions before your team sees anything.
A preview may not produce a sample report when there is not enough data for the automation to work with. For example, if you build an automation around a signal you just created, the preview has nothing to report on because that signal has not been applied to any conversations yet. Give the signal time to run on new calls, then preview again.