Perpetio

02/10/2026

What does a dented bumper have in common with a suspected fraud case?

No, that's not the start of a joke 😅 Both can end up in the same claims queue.

Not because they need the same level of attention. The problem is that someone still has to review the photos, policy documents, claim forms, and adjuster notes before deciding what happens next.

AI can help with this first step, so it doesn't take weeks.

Here are 4 ways how:

📸 Photos: a vision model identifies visible damage, estimates severity, and flags images that need a closer review.

📄 Policy and claim documents: a document parser extracts coverage limits, exclusions, and important dates.

📝 Adjuster notes: an LLM turns free-text notes into useful signals, including prior losses, possible pre-existing damage, and uncertainty in the assessment.

A scoring layer then combines the results and routes the claim to:

• straight-through processing
• a specialist queue
• further investigation

There are already some strong results in this area.

McKinsey reports that Aviva improved routing accuracy by 30% and reduced complex liability assessment time by 23 days.

PwC estimated 29% efficiency savings for estimators using AI photo analysis.

Low confidence, conflicting information, or unusual cases should always be sent for review.

More details: https://buff.ly/3ZD8fgf

16/09/2026

Satellite imagery alone can’t tell an insurer whether a crop insurance claim is valid.
A drop in NDVI, for example, could indicate crop damage. But it could also be a normal change as the crop moves through a different growth stage 🌱

So how could a claims workflow handle this?

Imagine a farmer reports hail damage on June 18. The system could:
1️⃣ Compare several satellite images from before the event to establish the field’s normal vegetation condition.
2️⃣ Check weather data around June 18 to see whether conditions at that location were consistent with the reported hail event.
3️⃣ Compare post-event imagery with the baseline to identify significant changes in vegetation.
4️⃣ Add context from crop type, growth stage, historical yield, irrigation records, and other farm data.
5️⃣ Check whether the reported date, location, crop, and damage match the available evidence.

From there, the claim could be prioritized for a faster review, standard review, or closer investigation.

This kind of multi-source approach is already being used for agricultural disaster assessment. For example, USDA’s National Agricultural Statistics Service combines satellite imagery, including Sentinel-1, Sentinel-2 and Landsat data, with crop information to assess areas affected by disasters.

We break down this workflow in more detail, including the limitations of satellite data and what the adjuster could actually see in the interface.

Read the full article: https://perpet.io/blog/automating-agricultural-insurance-claims-using-satellite-data-weather-and-farm-records/

24/08/2026

Today we celebrate our Independence 🇺🇦

Our biggest wish is simple: a peaceful sky over Ukraine.

Our people are our biggest gift, our greatest strength, and sadly, our biggest loss too. The price of freedom is paid every single day, by those who defend us and by civilians who simply want to live in their home.

Happy Independence Day, Ukraine 💙💛

07/08/2026

The Apple Heart Study monitored over 400,000 participants using an Apple Watch algorithm designed to detect irregular heart rhythms. When the algorithm flagged a potential issue, 84% of those notifications were confirmed as actual Atrial Fibrillation (AFib) upon follow-up ECG monitoring.

That’s a massive milestone for continuous, preventive health AI 🫀

But it also highlights a critical boundary for healthtech product teams: Even with an 84% positive predictive value, the algorithm didn't diagnose the patient. It simply shortened the distance between a subtle risk signal and a doctor getting involved.

For healthtech founders and CPOs, drawing that exact line between "flagging elevated risk" and "making a clinical diagnosis" isn't just a UX choice. It directly influences your entire regulatory path:

⚠️ Diagnosis ➡️ Triggers FDA Software as a Medical Device (SaMD) oversight and expensive clinical validation.
✅️ General wellness/Risk trends ➡️ Stays outside SaMD oversight, but requires strict claim language discipline.

Understanding where your feature sits on this line dictates everything downstream: how you handle messy multi-sensor data normalization, how you architect HIPAA-compliant pipelines, and how you design the UI to avoid user alarm fatigue.

We wrote a complete guide breaking down these exact trade-offs for healthtech teams building predictive AI features.

Give it a read here:

Healthcare has always run on a simple loop: something goes wrong, a person notices, and only then does treatment begin. This is AI preventive healthcare in practice, and it depends on AI health risk scoring models that turn raw physiological data into something a person can actually act on.

03/06/2026

We built the same app twice using Base44 and Claude Code, and the difference was very real 👀

We kept everything identical to make the comparison fair, and what we saw was a clear gap in how each tool behaves once things get real.

👉 Base44 gets you to a working product fast, and it’s genuinely impressive at the start, but as soon as you need automation, debugging, or deeper control, the limitations become obvious

👉 Claude Code feels closer to working directly in your codebase. More flexibility, more transparency, and fewer moments where you are guessing what went wrong.

We broke the whole experiment down step by step, what worked, what broke, and where each tool actually fits in real product development

🧠 Read the full comparison here:

Base44 is one of the most talked-about vibe-coding tools right now. You describe what you want, and it generates a working web app: database, UI, and logic included. For the right use case, it’s genuinely impressive.

21/05/2026

Сьогодні у вишиванках, як і багато років поспіль 💙💛

Це вже наша маленька традиція в Perpetio: навіть коли працюємо онлайн, ми щороку збираємось разом у вишиванках.

З Днем вишиванки ✨

Today, we’re wearing vyshyvankas, just like for many years in a row 💙💛

It has become a small tradition at Perpetio: even while working remotely, we still come together every year in vyshyvankas to celebrate Ukrainian culture.

Happy Vyshyvanka Day ✨

06/05/2026

We gave AI design tools a task we know very well… and let them try to keep up 🙌

We tested Figma Make and Google Stitch on the same real design scenarios we work with every day.

What are the results? Clean layouts, logical UX, and surprisingly solid starting points. But also some gaps when it comes to polish and control.

We also looked into Claude Design to see how this new player compares 🤔

Curious whether the AI tools did well with the design tasks? We broke it all down 👇️

We’re probably all familiar with AI tools for writing by now, like ChatGPT or Claude. That’s where it started for most people. Then AI got pretty good at working with code. And now, it’s moving into design.

12/04/2026

Світлого і спокійного Великодня 🐣🤍

Нехай у серці буде тепло, а навколо – тепло рідних і близьких!

31/03/2026

Thank you, Family F.C., for supporting Ukrainian children 🤍 ⚽️

Our client, Family F.C., made a donation to Voices of Children, one of the largest organizations offering psychological help to Ukrainian children affected by war. They provide online and offline therapy, run support programs, and help children cope with trauma and rebuild a sense of safety.

Family F.C. also gave families free access to their app, helping children find some moments of joy, distraction, and quality time through play and learning.

It honestly makes us really happy to see our clients taking action like this. It's a true match in values for us, and we’re very grateful for their support of this important Ukrainian cause 🙏

We invite you to join too, whether by spreading the word or making a donation if you can:
https://buff.ly/TflOiBh

Family F.C. is a team that we’ve been collaborating with for quite a while. It’s an innovative football coaching app for parents of kids aged 4–11, offering 90 expert-led training sessions, including ones created with former England national team player Matt Jarvis. It helps families build skills, confidence, and teamwork together 🙌

If you’re looking for ways to get children more active through sports, this is a great place to start:
iOS: https://buff.ly/e4ha9cr
Android: https://buff.ly/xamJuEV

────────────
Дякуємо команді Family F.C. за підтримку українських дітей 🤍 ⚽️

Наші клієнти Family F.C. зробили донат у фонд «Голоси дітей» – організацію, що надає психологічну допомогу українським дітям під час війни. Вони проводять терапевтичні сесії онлайн та офлайн, реалізують програми підтримки та допомагають дітям долати наслідки травм і відновлювати відчуття безпеки.

Family F.C. також поділилися безкоштовним доступом до застосунку, щоб допомогти дітям знаходити моменти радості, відволікатися і якісно проводити час під час тренувань.

Ми щиро тішимося, коли бачимо, як наші клієнти підтримують такі ініціативи. Ми раді, що вони поділяють наші цінності і дуже вдячні їм за підтримку такої важливої організації🙏

Долучайтеся й ви, якщо маєте таку можливість: словом або ж ділом, тобто донатом:
https://buff.ly/qBO1tWX

Ми вже досить довго співпрацюємо з командою Family F.C.. Це інноваційний застосунок для футбольних тренувань, який створений для батьків дітей віком 4–11 років. Він пропонує 90 професійно розроблених тренувальних сесій, зокрема, створених разом із колишнім гравцем збірної Англії Меттом Джарвісом. Він допомагає родинам розвивати навички, впевненість і командну роботу 🙌

А якщо ви думали про те, щоб додати більше спорту в життя дітей, цей застосунок може стати чудовим початком:
iOS: https://buff.ly/e4ha9cr
Android: https://buff.ly/xamJuEV

08/03/2026

Today is International Women’s Day 💙

At Perpetio, women are behind every project we make. They work as project managers, QA specialists, HRs, sales managers, designers, and developers, helping ideas turn into real products every day.

They ask the right questions, notice the small but crucial details, solve problems, and keep the work going along the way.

Today, we just want to say thank you to the women on our team and to women across the tech community who keep learning, building, and sharing their knowledge🌷

────────────

Сьогодні Міжнародний день боротьби за права жінок 💙

У Perpetio жінки є частиною кожного проєкту, який ми створюємо. Вони працюють як проєктні менеджерки, QA-спеціалістки, спеціалістки з продажів, HR-менеджерки, дизайнерки та розробниці, щодня допомагаючи перетворювати ідеї на реальні продукти.

Вони ставлять важливі запитання, помічають невеликі, але важливі деталі, знаходять рішення та рухають роботу вперед на кожному етапі.

Сьогодні ми хочемо подякувати жінкам у нашій команді та всім жінкам у сфері IT, які продовжують навчатися, створювати нове й ділитися знаннями🌷