Case Studies

Using AI to simplify carbohydrate counting: JAKALA for Theras

4 min read

Published on October 1, 2026

Solutions: App Development Services, Creativity & Content
Pharma e Lifescience
Using AI to simplify carbohydrate counting: JAKALA for Theras
Theras logo

Carbohydrate counting is a demanding part of everyday life for people who need to monitor their carbohydrate intake, including those living with type 1 diabetes. It requires attention and accuracy at every meal, whether at home or on the go.

In collaboration with JAKALA, Theras, an Italian biomedical company, addressed this challenge by developing T.conto, an AI-powered app that recognizes food from a simple photo and estimates carbohydrate content in seconds.

Developed through a co-design process involving users and healthcare specialists, and validated through testing with more than 200 volunteers across three specialized centers, T.conto was created as an educational tool designed to support, rather than replace, users’ judgment and their ongoing relationship with healthcare professionals.


At a glance

+13,000

Total app downloads (as of September 2026)

20-30%

User retention rate, measured 4–5 weeks after first use

3 workstreams

R&D, feature development and AI governance: the three areas of ongoing collaboration between JAKALA and Theras since launch

Context & challenges

Theras, an Italian biomedical company, set out to address carbohydrate counting by listening to the needs of its users. For more than 15 years, the company has maintained an ongoing dialogue with healthcare professionals and people who need to carefully monitor their carbohydrate intake, including those living with type 1 diabetes.

Carbohydrate counting requires constant attention and presents several practical challenges. Foods contain different amounts of carbohydrates, while portion sizes can vary significantly.

At home, people generally have greater control over ingredients and quantities. Eating out, whether at a restaurant, while traveling or in social settings, makes accurate estimation considerably more difficult, as nutritional information is often unavailable or imprecise.

Mixed dishes, such as complex salads or complete meals, add another layer of complexity, making it harder to estimate the carbohydrate content of individual ingredients.

Time pressure, fatigue and distractions can also lead to errors. Beyond the practical challenges, the constant need to calculate carbohydrate intake can create stress and a sense of restriction, affecting the enjoyment of shared meals.

Theras had a clear objective from the outset: to help users approach carbohydrate counting as an educational practice, reduce average estimation errors compared with the tools typically available to them, and significantly simplify their daily experience.

This approach also acknowledged that a certain margin of inaccuracy is inherent in virtually any estimation method, whether technology-based or human.

For JAKALA, the challenge was twofold: designing an AI-powered solution capable of recognizing dishes and estimating nutritional values, while creating a user experience that kept people firmly in control.

The goal was to prevent AI from becoming an unquestioned source of truth, preserving the role of users’ judgment and the educational journey shared with healthcare professionals (HCPs).

JAKALA's approach

The partnership between JAKALA and Theras began in May 2025, bringing together technical, design and domain expertise.

Built around experimentation and continuous user validation, the collaboration has evolved into an ongoing program focused on improving the solution and ensuring effective post-launch governance.

May 2025 — From prototype to MVP

The project began by transforming an early-stage prototype into a robust, usable minimum viable product (MVP).

During this phase, the technical solution and user experience were developed in parallel as part of a fully integrated design process.

The result was a two-step analysis workflow:

  • Food recognition: an AI algorithm, progressively refined to identify dishes and ingredients.
  • Nutritional analysis: carbohydrate estimation based on a certified nutritional data source.

A critical design decision was to give users the ability to review and adjust ingredients and quantities between these two steps.

This ensures that AI accelerates and simplifies food recognition, while users retain the ability to validate or correct the information before the final calculation.

July 2025 — Real-world validation

The solution was tested with real users during a dedicated event in Milan.

The feedback and results collected, together with internal testing, marked a turning point and informed the decision to develop a native mobile application.

August–September 2025 — Native app development

The team moved into full-scale development, establishing the technical foundation for the solution’s public release.

January 2026 — App store launch

T.conto was officially released on app stores.

In the following months, adoption was supported through a low-profile approach, primarily driven by word of mouth, internal company presentations and meetings with stakeholders connected to Theras, including healthcare professionals.

Beyond launch — Continuous improvement and AI governance

The collaboration did not end with the app’s release.

The joint JAKALA–Theras team continues to work across three dedicated streams:

  • R&D: continuous improvement of the technical solution and user experience, including benchmarks developed with nutritionists and dietitians using established carbohydrate reference values, alongside ongoing monitoring of AI models available on the market.
  • Feature development: progressive expansion of the app’s capabilities.
  • AI governance: oversight of application metrics, security, algorithms and costs.

Technologies

  • Artificial Intelligence for food and ingredient recognition

  • Integration with the certified CREA nutritional database

Results

  • 01

    Reducing the cognitive burden of carbohydrate counting

    Users now have access to an app designed to support everyday meal management. By simply taking a photo of their food, they can receive a carbohydrate estimate within seconds, reducing the cognitive effort associated with a task that needs to be repeated at every meal.

  • 02

    Keeping users in control

    T.conto allows users to review and correct ingredients and quantities before confirming the final estimate. This reflects the project’s original objective: creating an educational tool that makes carbohydrate counting faster and easier while ensuring that users retain the final say over every estimate, within the broader educational journey shared with healthcare professionals.

  • 03

    A long-term commitment beyond launch

    The partnership between JAKALA and Theras continues beyond the app’s release. The joint team remains focused on continuous improvement through benchmarks developed with nutritionists and dietitians using established carbohydrate reference values, monitoring of emerging AI models, analysis of real-world usage data and user feedback, feature expansion, and ongoing governance of security, algorithms and costs.

     

“This app was born from listening. For more than 15 years, we have built an ongoing dialogue with people who need to pay particular attention to carbohydrate counting, their families and healthcare professionals. We wanted to provide a tool that is simple, intuitive and genuinely useful.”

Andrea Scaranello

R&D Product Owner, Theras 

Andrea Scaranello
TECHNORETAIL_HERO_INSIGHT_2200x1222_OWNED-copy

Discover more about JAKALA's unique offering

Solution

  • App Development Services
  • Content, Creativity & Experience

Industries

  • Pharma & Lifescience

Insights

Inspirational IAB 2026
Article

Inspirational IAB 2026

9 min read
XXIII Open Space Edition: Pamplona
Article

XXIII Open Space Edition: Pamplona

5 min read
JAKALA Connect: Back to School, Back to Adobe
Article

JAKALA Connect: Back to School, Back to Adobe

5 min read
Meet Eugenio Aluffi Valletti, this year’s Step Challenge champion
Article

Meet Eugenio Aluffi Valletti, this year’s Step Challenge champion

6 min read