Its own calls, behind its own login
Your platform charts the calls it handled, for whoever logs into it. The phone team and the next vendor aren’t in it, and the person carrying the number never opens that dashboard.
Palesa Independent measurement layer
Any platform, one scale
You answer for every conversation you run.
Palesa scores every one of them, voice or text, AI agent or human, on any platform, against one scale. When someone asks whether it is working, you have the number, and the quote sitting underneath it.
Inbound voice · 3m 12sVoice AI
AGENT 07
Our policy is a full refund on a duplicate debit. I’ve raised it.
- Resolution
- Resolved
- Compliance
- Compliant
- CX score
- 5/5
The evidence
A conversation in. A record out.
Your platform can pull fields out of a call. What it can’t do is hand that record to someone outside it, graded against your standards, with every field showing the turn it came from. The person who doesn’t trust you can go and check.
Inbound voice · 3m 12sVapi
- Agent01
Thanks for calling Acme Energy. You’re speaking with an automated assistant. How can I help?
- Customer02
I’ve been charged twice for September. Same amount, two days apart.
- Agent03
I can look into that. Can you confirm the postcode on the account and the last four digits of the card?
- Customer04
2481, and the card ends 4417.
- Agent05
That’s verified, thank you. I can see two debits of $49.00, on the 3rd and the 5th.
- Customer06
So can I get one of them back?
- Agent07
Yes. Our policy is a full refund on a duplicate debit. I’ve raised it. You’ll see the $49.00 back within three business days.
- Customer08
Great, that’s all I needed.
Support template
Duplicate September debit, refunded.
Billing · duplicate debit · 8 turns · 374 words
CX score
5/5
resolved · low effort · ended positive
Resolution
Resolved
turn 7 · “I’ve raised it”
Customer effort
Effortless
turn 8 · “that’s all I needed”
Sentiment
Positive
whole conversation
Compliance
Compliance
Compliant
whole conversation · no gaps
Identity verification
Attempted
turn 3 · “confirm the postcode”
AI disclosure
Given in turn 1
turn 1 · “an automated assistant”
Custom fields
Policy stated before commitment your field
Yes
turn 7 · “our policy is a full refund”
Refund promised your field
$49.00 · 3 business days
turn 7 · “within three business days”
The gap
Your platform reports on itself.
We measure the whole business.
Every source grades differently
Each vendor scores against its own scale. Two dashboards can both be right and still disagree, so no comparison between them holds.
The record leaves with the contract
Switch platform and the baseline resets, exactly when the comparison is worth most. The evidence you’d want later belonged to the vendor.
How it works
One layer above every platform you run.
Collect
Every conversation, wherever it happened
Connect Palesa with your voice AI, telephony or note-taking provider, e.g. Vapi, ElevenLabs, Bland, LiveKit, Twilio, Zoom Phone or Granola, or simply send your conversation data to a webhook. AI or human, phone call or meeting in a room, it all lands in one schema on the way in, so nobody has to ask which platform a number came from.
Grade
One definition of a good conversation
We score resolution, customer effort, sentiment, conduct and compliance the same way on every conversation, against your standards and your own fields. The CX score rates each one 1 to 5 from whether it was resolved, how hard the customer had to work and how they left, so when it moves you can see why. The numbers reconcile because one scale produced all of them.
Account
What you hand to the board
Scheduled reports reach the people who never log in, and the events you care about raise an alert in the channels your team already works in, e.g. email and Slack. Extracted fields push into your CRM, e.g. HubSpot, and a webhook carries the rest into the warehouse you report from. Ask the same questions from inside your own AI tool over MCP, or from the command line.
How your data is held. What GDPR asks of you is a setting here, not a support ticket. You set retention, you export, you delete, and we can strip personal data out of transcripts before any of it is stored. Transcripts, the analysis drawn from them and every personal field are each encrypted with AES-256-GCM. Recordings open only from inside the app, on a link that expires, and you choose whether we keep them at all.
The comparison
Is the AI doing it as well as the people were?
The highest-stakes question in the business, and every platform answers it for the conversations it hosts. Not the phone system nobody replaced. Not the vendor you are weighing it against.
Palesa puts both sides on the same scale, in one report, and keeps the baseline when you change your mind about a vendor.
Same scale · Q3
Human team AI agent
Illustrative figures, every metric scored 0–100. One scale, both sides, and the platform that ran the call is not the one grading it. Small samples get no verdict at all.
Independence
We don’t sell AI agents.
So measuring them fairly costs us nothing. Every platform that grades its own calls has a side to take, and a record of its own good behaviour is worth nothing to an auditor, a client or an insurer.
Launching its own analytics product, an enterprise voice AI provider reported that standard resolution metrics had overstated performance by twelve percentage points.
Who it’s for
Built for whoever answers for the outcome.
Not for whoever tunes the prompt. They already have a dashboard, and it lives where the prompt does. We measure conversations that already happened, not agents before they go live.
One platform, one channel, nobody asking you for the number? Your built-in analytics are enough. We’ll tell you so on the call.
See what the record looks like.
Twenty minutes on a call. We take a graded conversation apart score by score, then talk about yours. That is enough to tell whether this is for you.