What Can Go in Lomi?

Helping customers compost confidently through AI-powered guidance.

About Lomi

Lomi is a suite of digital and physical products that reimagines how people manage food waste at home. Beyond the flagship Lomi countertop composters and accessories, the Lomi Ecosystem consists of e-commerce experience, web portals, and multiple mobile apps. Together, these create an ecosystem that seamlessly blends hardware and software to motivate and influence customers in reducing their household food waste.

The Problem

One of the biggest challenges customers faced was understanding what could safely go into their countertop composter. With millions of possible food items, our team previously felt this was an "unsolvable" problem overwhelmed by the impossible task of manually testing and documenting each and every possible food item. As our customer base grew, so did our support tickets, fuelled by misinformation spreading in our community forums and social media groups.

Top help article top searches were related to what can go in Lomi and warranty details.

• Users: Confused and hesitant, unsure what could go in Lomi

• Support team: Backlogged with repetitive inquiries, often taking 2–3 days to respond

• Business: Warranty claims rising due to improper usage, hurting brand trust
  

Problem statement:“How might we help customers quickly and confidently know if any food item can go into Lomi — without requiring human intervention?”

A few extracts of customer comments from a survey. Wanting to better understand food suitability was constantly the top theme in feedback channels.

The Opportunity

By championed the use of groundbreaking artificial intelligence and machine learning technology we could transform this problem into an opportunity: giving users a searchable, AI-powered guide that could predict food suitability based on known qualities like starch, sugar, or fat content.

We designed an AI-powered search feature inside the Lomi app that allowed users to type in almost any food item and instantly receive guidance on:

• Whether it’s safe for Lomi

• Any exceptions, e.g. pits

• Recommended portion size

• Tips to optimize compost quality
  

Instead of relying on solely a static list, the ML model assessed items dynamically based on their food composition properties (like sugar, starch, or acidity), making the system scalable and future-proof.

The vague and limited guidance caused frustration and hesitation in using their Lomi.

My Role

• Built the vision, campaigned for leadership buy-in and solicited resources for the project.

• Generated project enthusiasm and built cross-team trust.

• Partnered with engineering and science team to conceptualize and validate the ML model.

• Oversaw the designs of the search experience, results interactions.

• Mentored designers on AI best practices.

• Guided set-up of impact measurement
  

  
  

Process

1. Empathize: Research & Insights

• Tracked warranty claims and conducting discovery on causes which showed a strong link to unintentional misuse

• Tracked consumer sentiment and explored themes related to declining brand reputation

• Collaborated with warranty and product teams to understand cost implications and business impact

  

  Outcome of our Journey Mapping workshop with leaders across the org identified "onboarding" and "first 30 days" as the lowest user experience points but team was suck on how to solve the primary issue.

  
  

2. Problem Definition: Framing the Opportunity

• Manual testing was unsustainable — too many possible inputs and time consuming process

• We had strong evidence that unintentional misuse and customer knowledge was impacting warranty claims and overall brand reputation

• Built internal education around AI capabilities and ease of use

• Success metric: reduce helpdesk tickets and warranty claims while increasing app engagement
  

Flow diagram of the algorithmic framework that would be used and the hierarchy of process.

3. Prototyping the Experience

• Created lightweight proof of concepts to demonstrate viability of the technology and build enthusiasm for the capabilities of AI technologies.

• Tested early with customers to refine terminology and confidence scoring

• Iterated with engineering to balance accuracy and user clarity

  

  Early wireframes

4. Launch & Learn

• Released the first iteration of the feature within the Lomi app only included amounts and cycle suitability. Received positive impact but solidified customers desire for additional knowledge

• Released second iteration which included additional exemptions (e.g. pits or bones), and bucket balancing.

• Based on technical complexity, we decided to wait till a future release to include searches for multi-ingredient foods (e.g. ravioli, cake).

• Monitored adoption, ticket volumes, and compost quality complaints.
  

  

  Loading interaction for AI generated answers.

  
  

Results & Impact

• 4x increase in weekly active app usage from about once every two weeks to twice a week. Feature was most visited page on the app, demonstrating cx enthusiasm for the feature and high engagement.

• Percent of support ticket topic dropped from 7% to <1% of all inquiries, preventing backlog from growing and reducing wait time on all tickets.

• Provided customers Instant answers instead of 2–3 day delays, and reducing misinformation from peer to peer groups.

• Team was confident in the future reduction of warranty claims, based on improved customer education.

• Complaints about poor compost quality nearly disappeared, building trust, improving Lomi reputation, and reducing frustration from customers around cleaning their Lomi.

• Reduced additional food waste. Multiple customers commented that the feature increased the amount of food they composted by providing them clarity on items they previously would have thrown in the trash.
  

High fidelity mockups used for prototyping and usability testing.

Personal Learnings & Take-aways

This project demonstrated how AI and design together can scale knowledge at a rate which humans can’t. By reframing the problem from “document every food” to “predict based on food qualities,” we unlocked a solution that was:

• User-centric; confidence, speed, clarity

• Business-positive; reduced warranty costs, fewer complaints, improved customer sentiment

• Future-proof; scalable beyond manually curated lists, and amplifying our impact
  

For me, this reinforced the importance of:

• Building an understanding of the newest technologies and their usage.

• Resilience in taking multiple approaches to breakdown assumptions and bias

• Cross-functional collaboration with data science and support teams

• Building a "better than" feature rather than a perfect solution.

• Using AI as a bridge between overwhelming complexity and human usability