These are the automations that run when a phone call is completed within the CRM, provided the calls are being recorded, and there is a configuration for the sub-account where the call occurred.

• Determining which user answered the call, using the transcript
• Extracting the name of the contact who called if they’re unnamed in the CRM
• Fixing issues with the transcription using an LLM
• Summarizing the call, and grading the sales performance if applicable

These are automations that run regularly in order to perform a number of automations for each contact in the CRM:

• Grading the sales performance for the conversation overall
• Summarizing the conversation overall
• Generating followup tasks if necessary
• Determining the value or pipeline status of the contact
• Etc.

There are a few ways that evaluations can be triggered, to ensure that your data stays as current as possible:

• When a call is completed over X seconds long (default 60), the call will be processed and then the contact will be re-evaluated to account for any new information that call provided.
• After X hours (default 4) have passed since the last activity, an evaluation can be triggered. This is useful for ensuring contact evaluations always end up being accurate in the long run.
• When there are X (default 6) messages sent/received. This is useful when you do a lot of your communication over messaging, but want contact/sales scoring to stay up to date.
• After a specific (default “run-eval”) tag is added to the contact. Useful when you want to evaluate a contact immediately if they just filled out a form or additional context was added.

In an effort to make evaluations more efficient and reduce AI usage, there is a “middleware” system that runs before an evaluation.

In this system, an LLM will inspect the differences between the last time the contact was evaluated and now, in order to determine if a re-evaluation is necessary.

This AI usage is not rebilled to you.

When evaluating a contact, the LLM can see basically everything you can think of:

• Sub-account details like name, timezone, and any additional context
• Contact details like name, email, and phone number
• All custom fields and notes on that contact
• Appointments, opportunities, and tasks
• All communication including emails, phone/video call transcripts, and messages
• The LLM sees this on a chronological layout, meaning it can see how long it takes for the contact or sales rep to respond to messages

The first way we solve this is by grading the quality of our leads using AI.

In this system, you’re able to define your own criteria for what makes a qualified lead:

The AI then grades the contact from 1-5 in each of these criteria, to produce an overall grade.

You can see these grades in the dashboard, using the “Contacts” table. You’re also able to sort this table by the quality grade, to surface your highest quality leads at the top:

Sales grading works in a similar way to contact grading, however sales grading applies to both individual phone/video calls, as well as the conversation with a contact overall.

With sales grading, you can paste in your sales framework:

Then, you define the criteria you’d like the AI to use when grading the sales rep:

These criteria result in a very similar grading result:

As you’ve seen above, the way you define your grading system is through a list of 3-6 criteria that you’d like the AI to grade the contact on.

There are four levels of importance you can choose from:

• Critical (10x)
• High (3x)
• Normal
• Low (0.5x)

In order to score the lead/rep, the AI will assign a score from 1-5 for each of your criterion.

Then, a weighted average is taken. For example:

• Framing: High (3x) → Scored 2 → +6 points
• Discovery: Critical (10x) → Scored 3 → +30 points
• Closing: High (3x) → Scored 5 → +15 points

The total points possible is (5 x 3) + (5 x 10) + (5 x 3) = 80 points.

Since the rep scored 51 points, the overall score is 51/80 = 0.6375 or 63.75% (displayed as 64%).

The AI is also capable of deciding to skip lead/sales grading, in cases where there isn’t enough information to justify it.

You can define your criteria for when to skip grading:

When the AI decides to skip grading, you’ll see a reason for this in the dashboard:

Firstly, we can use the context around the contact to determine where they should be in the pipeline.

All you need to do is define a few statuses that you’d like the AI to choose from:

When the AI makes a decision, you’ll see the result and thought process in the dashboard:

We can also make it so this status decision will update a custom field on the contact:

Then, all we need is a simple workflow to translate this result into actual pipeline updates:

The next piece of the puzzle is updating the opportunity value to reflect the actual value of the lead.

For services with high price variability, this can give you a much clearer picture of actual revenue and ROAS.

First, we define a set of examples that the AI can draw from to estimate the value:

Then, we only need to enable the custom field updates, and create a simple workflow to save the results:

As discussed in the pricing section, call transcripts are generated using a high-quality speech to text model, then improved in several ways:

• The correct user is assigned (only necessary for inbound calls)
• Diarization errors (who’s speaking, and when) are fixed
• Obviously incorrect word transcriptions are fixed

This allows for a more accurate call transcript, which results in more accurate AI analysis results.

Once the call processing is complete, you’ll have a page like this for your phone call:

With this app, we actually have two kinds of summaries:

• Call summaries, as you’d expect
• Contact summaries, which generates a summary of the entire contact and conversation overall.

You simply need to define a prompt for the summary:

Then you can define your options for how summaries will appear:

Summaries will always show up in the dashboard, and can optionally be added to the notes or a custom field of the contact.

I usually recommend that you only have the “Full Contact Summary” added to the notes, as this prevents the notes from becoming clogged with old call summaries.

Once a new contact summary note is created, the previous (now outdated) note will be deleted.

In addition to regular phone calls, video calls can also be saved to the system.

For this to work, you need to be able to copy the call transcript from your note taker, such as with Fathom:

When you paste a Fathom transcript, it’ll look like this:

Streamlined Founder Meeting for Jeff Evans - August 14
VIEW RECORDING - 30 mins (No highlights): https://fathom.video/share/g_n9etyxet_UB-yVxbrh...

---

[...full transcript here]

Now, if we save that to the notes section on the contact, the system knows what to look for:

Then, it’ll generate a summary and a sales grade just like it would for a call:

When you first sign up for Streamlined, you’ll be given four default dashboards:

There are three primary dashboard filters that you can use:

• Sub-account or configuration: Filter all data down to a single account or a group of accounts.
• User: Filter all data down to a single user to view only their performance.
• Date: Pick from a list of presets, or select a custom date range for the data.

For user filters, in many cases you’re able to select what the user filter should define:

Custom stats are the most developed dashboard feature, allowing you to calculate almost any metric you’d like.

When you click on the blue settings icon at the top right of a fresh stat card, you’re able to choose from a list of templates:

You have a choice of either using a default template provided by me, or copying one of your own metrics.

Selecting “Appointment Show Rate” from the list of templates, we’re presented with our formula:

There are a few key things to point out here:

• “Previous” in the bottom left, means that we’ll compare the selected date range’s result to the previous result.
• To the right of that, “Higher is better” means that when our value increases over the previous value, the indicator should be green.
• On the far right, the “Percentage” means that this number will be displayed as a percentage.

In this screenshot, where we’re hovering over the green indicator:

• 56.25% of our appointments in the selected “Last 7 days” date range are marked as shown
• That’s 29% higher than it was in the 7 days before the selected date range
• Because we selected “Higher is better”, this indicator is green

We could also compare this to a KPI that we set:

In this case, we set our target to 75%, which means the indicator turns red because we’re under our target.

We can also set monthly goals, for numeric targets:

Here, we’ve set our monthly goal for appointments shown to be 25.

Since we’re looking at the “Last 7 days” date range, our target becomes (7 ÷ 30) × 25 = 6.67 for this period.

To calculate appointment show rate, we filter each of the metrics:

For the first “Appointments” metric, we’ve filtered it so “Appointment Status” includes “Showed”.

Thus, dividing showed appointments by total appointments gives us our percentage of appointments marked as showed.

You can use several different metrics, filters, and aggregations to create almost any metric:

Here are some examples that you’ll have in your “Default Templates”:

If you’re unsure of whether this app is capable of tracking a specific metric you have in mind, please 💬Get in Touch.

To simplify things, there are two “versions” of charts available:

• Square boxes, which can have either pie charts or radar charts
• Rectangular boxes, which can have either line charts or bar charts

Bar, line, and pie charts all display sums, and so they have the exact same target and grouping options:

There are countless permutations of values and groupings you can select here. I encourage you to play around and create as many charts as you like.

Radar charts are different, since they display a percentage:

These are useful for comparing different marketing channels or sales reps:

To add a new table, we’ll add a full-width table section as shown here:

You’re able to sort data however you’d like, and show/hide/reorganize the columns:

In order to share your dashboards, you have two options:

• Share the dashboard with specific user types in your CRM
• Create a publicly accessable link to view the dashboard

When you create a new dashboard, you can choose which user roles have view/edit permissions:

Note that users will need to sign in to the dashboard in order to view their data with the correct permissions.

Only the creator of the dashboard is able to edit the name and permissions of the dashboard.

For the default dashboards, the creator will be the user signed in when the app was installed.

You can also create a public share link for the dashboard like so:

After copying the public share link, your clipboard will contain a link like this:

https://streamlined.so/interfaces/public/dashboards/b17293b0-e997-4eb0-8d7e-4678065d65d7/{{location.id}}

If you just want to send this directly to someone, you’ll need to swap out the location ID with a real location ID:

https://streamlined.so/interfaces/public/dashboards/b17293b0-e997-4eb0-8d7e-4678065d65d7/flj4Xi8hjECt8FVkAAJA

Alternatively, you can create a custom menu link using the default link, which will auto-populate the location ID:

The dashboard has three theme settings you can choose from:

• System: Uses the same color scheme as your device (default)
• Light: Light mode, always
• Dark: Dark mode, always