Accord Mediation Engine

Inspiration

I’ll be honest: I’m a pretty sensitive person and I pick up on emotions easily. Because of that, texting can be a nightmare for me. Tone completely drops out over text, and I can't count how many times I’ve gotten into stupid arguments with my friends just because a message felt cold or sounded mean.

I've always thought that it was a 'me' problem, but one random day I actually looked it up to see if others feel the same way as well and I found out that this a very real problem in the corporate world, they are losing billions of dollars in lost productivity over managing friction and arguments instead of doing actual work.

When you look at remote teams working on Slack, Discord, or WhatsApp, the same thing happens but on a bigger scale. People misunderstand each other, take things personally, and suddenly a whole project stalls because everyone is upset. I built Accord to act as a circuit breaker. I wanted something that could take a messy, heated chat, strip away the angry noise, and show what people actually need so they can just fix it.

What it does

Accord takes messy, heated group chats from platforms like Slack or WhatsApp and strips away the angry noise. It calculates a quick escalation score to show how bad the fight is, highlights what each person is actually upset about, and builds a step-by-step checklist of compromises. Instead of a confusing layout, you just see pure, logical facts or the emotional sides of the argument.

How I built it

I wanted to build a workspace that didn't feel cluttered. You just upload or paste a messy chat log from Slack or WhatsApp, and the app goes to work.

I used Supabase Edge Functions to handle the heavy lifting. Instead of splitting the screen up and making it look confusing, I put a simple toggle switch right at the top. You can click it to flip between Objective Mode (just the facts and logical issues) and Emotional POV Mode (which shows everyone's actual hurt feelings and core needs). It also creates a step-by-step resolution list where people can check boxes to agree on compromises, and the app updates the status automatically. Ofcourse with MeDo doing all the heavy lifting.

Challenges we ran into

People talk in incredibly messy ways. When they're arguing, they use typos, weird slang, and they talk over each other. Getting the AI to actually tell the difference between a hard fact, a personal opinion, and someone just lashing out emotionally took a ton of trial and error with prompt writing.

Another headache was making the dashboard update smoothly. If a user is chatting with the AI helper to fix a specific part of the analysis, I had to make sure it updated that single piece of the page without wiping out the agreements people had already clicked.

What we learned

This project forced me to learn how to bridge the gap between unpredictable human emotions and strict code. I had to figure out how to take a completely chaotic conversation and format it into clean, reliable data structures like arrays and objects so the UI wouldn't break. On a personal level, it taught me that most text arguments aren't actually about people being mean; it's just that tone gets lost, and people start missing each other's point or they get too caught up in the moment and their emotions that they don't understand the other person's point of vew. But looking at the analysis from a 3rd POV really opens your eyes to hard facts

What's next for Accord Mediation Engine

Right now, Accord relies on pasting text or uploading files manually, so I want to build direct integrations next. The goal is to let users connect it straight to a Slack channel or Discord server with a quick command. I also want to add a feature that flags historical patterns, so team leads can see if the same two people keep running into the exact same communication issues over and over again.