An LLM turned Jira tickets and customer chats into one clear mission.
Finding the Mission That Actually Matters
I once worked with a team that had more than 1,200 Jira tickets. Some were five years old. Nobody remembered who wrote them. Many of the people who had created those tickets didn’t even work at the company anymore. The ones who were still around couldn’t explain what half of them meant.
Every Friday, the Product Owner sat in a “grooming” session, dragging cards up and down the backlog. Meanwhile, developers had stopped reading the tickets months ago.
On the surface, the team looked busy. The backlog was endless, and there was always something to discuss. But in reality, nothing important was moving forward. That’s not high performance. That’s noise.
And that’s the question I keep hearing at meetups, on social media, and even within teams themselves: how can AI help us cut through the noise and build real performance and how can we be product driven ?
In one retrospective, I asked the team a simple question:
“What is your mission?”
Some of the answers that came back were like:
- “Make the user experience better.”
- “Add more payment options.”
- “Reduce support calls.”
All good ideas. But none of them matched. The team was busy, but not aligned. There was no shared mission.
Instead of guessing, we decided to look at what customers were actually saying. But there were tens of thousands of messages: support tickets, chat logs, surveys, emails, app reviews. No human could read them all. The data we needed wasn’t just scattered—it was in different systems: support tickets in Zendesk, live-chat transcripts, NPS surveys, emails from the website, and even a few App Store reviews. We pulled five months of it together. There were around 30,000 messages in total.
When we first looked at it, it was chaos. Some people wrote long letters: “Dear Support Team, I’ve been a loyal customer for years…” Others just typed a single line: “checkout broken.” Many complaints were repeated dozens of times. Some messages were full of order numbers and email addresses we didn’t need.
Before the AI could help, we had to clean. We stripped out greetings, IDs, duplicates, and kept only the customer’s raw words. What remained were short, simple lines:
- “checkout too slow”
- “payment failed three times”
- “form too long on mobile”
Now we had something real: the voice of the customer. But 30,000 lines were still too much for humans.
Letting AI find the patterns
We used a large language model to help us make sense of it. Nothing advanced was needed a mid-size model like GPT-3.5 was fine. The steps were straightforward:
- Group similar sentences – Each complaint was turned into a vector (a list of numbers representing its meaning). Sentences with the same meaning landed close together, like puzzle pieces of the same color.
- Cluster them – The AI grouped nearby points into clusters, each one representing a recurring theme.
- Label the clusters – We gave the model a simple prompt:
You are an analyst helping a product team. Here is a list of similar customer complaints.Give the cluster a short label (max 5 words).Summarize the pain in one clear sentence.Count how many complaints fit this theme.Complaints: {{list_of_complaints}}
The model came back with themes that felt immediately familiar:
- Too many form fields — “Checkout asks for too much information.”
- Payment errors — “Payments fail or time out, especially on mobile.”
- Checkout takes too long — “Customers abandon carts because the process is slow.”
One theme dwarfed the rest. Almost 40% of all complaints pointed to checkout speed.
Turning complaints into a mission
We didn’t want just a list of problems. We wanted one line that the whole team could aim at. So we asked the model again:
The answer was short and sharp:
Based on the top pain point, write a mission in this format: “Our mission is to [verb + measurable outcome] before [timeframe].”
“Our mission is to reduce checkout time by 30% before Christmas.”
Why this was a mission
At first, someone asked: “Is that really a mission?” And yes, for this team, it was.
- It was specific — not “make checkout better,” but “reduce time.”
- It was measurable — 30% is a clear target.
- It was time-bound — before Christmas, a date that mattered for the business.
- It was customer-driven — it came from actual complaints, not from a manager’s slide deck.
- And it was shared — every role in the team could see how to contribute. Designers cut form fields. Developers optimized queries. Testers tracked speed. Support expected fewer angry calls.
That’s the difference between a vision and a mission. A vision might say: “We want to be the best online shop.” That’s inspiring, but vague. A mission is grounded: “Cut checkout time by 30%.”
From there, the team could later break it down into goals, smaller, testable steps that showed progress.
And just like that, the team finally had one destination. Designers, developers, testers, and support staff, all pointing at the same problem.
Missions in Startups and Scale-Ups
In startups and scale-ups, the mission changes often. One week the focus is on customers, the next on investors, then on growth or hiring. Teams can feel like the goal is moving all the time.
That is normal. Small companies grow fast, and priorities shift. What matters is not to find one mission forever, but to agree on what the mission is right now.
AI can help by looking at feedback, sales talks, or investor updates and showing what comes up most. The mission will change again later, but at least the team has clarity for today.
One startup I worked with had a different problem: they were looking for investors. Instead of spending hours on LinkedIn, X, and Google, they built a small AI agent with Make.com. They gave it clear specifications: industry focus, funding stage, geography. The agent searched online for matching investors, gathering names, websites, and contact details into one list.
The mission was still changing every few months, but for that stage, it was simple: find the right investors. The AI agent didn’t replace the team’s work, but it saved them weeks of searching and gave them clarity on who to approach first.