An AI-powered Slack Monitor so Lyre Studios never misses a message
Across many teams and Slack channels, important messages were getting buried. Even “I’ll get to it later” replies looked the same as no reply, and slipped through just as often.
Build partnership
They felt like an extension of my own team. They understood the AI workflow design immediately and shipped production-grade code that slotted straight into how Lyre Studios runs.
The challenge
Alexander Popov is the founder of Lyre Studios, a software studio that builds AI-augmented products. As the studio grew, important conversations across its Slack (client escalations, time-sensitive requests, team mentions) were getting buried in the firehose.
- Too many channels, too many messages
Across dozens of channels and hundreds of daily messages, important threads were getting buried in scroll history within hours.
- “I'll get to it later” is invisible
Messages with delayed-reply intent (“I'll handle this tomorrow”, “circling back later”) looked the same as unanswered ones, and were just as likely to slip.
- No prioritization beyond chronology
Slack surfaces the newest messages first, whatever their importance. A frustrated client message and a casual thank-you ranked the same.
How we built it
Alexander brought us in as Lyre Studios’ engineering build partner. Together we shipped the Slack Monitor, an AI-triaged inbox that surfaces what needs a reply, ranked by urgency and sentiment.
- Scheduled message fetch: a background job runs on a schedule, pulling all relevant messages and conversation history from Slack.
- Reply detection: every message is classified as replied, no reply, or delayed reply (where someone said "later" without resolving). Only the unanswered and the delayed surface in the monitor.
- AI sentiment + urgency scoring: every surfaced message is scored on Urgency (Low / Medium / High) and Sentiment (Happy / Neutral / Mad) by an AI workflow. The score determines its rank.
- Sort that puts what matters at the top: worst-sentiment + highest-urgency messages always rise to the top. A "Mad / High" client escalation surfaces above a casual "Neutral / Low" team mention every cycle.
- Filter the noise: by channel, by person, or by date range, so each reviewer sees just their own slice.
See it in action
A walk through the three core screens, reconstructed to mirror the live tool. It advances on its own, or step through it yourself. Data shown is illustrative.
| Channel | Customer | Volume | Tracking |
|---|---|---|---|
| #client-acme | Acme Corp | 142/wk | |
| #cs-tier-1 | Support | 320/wk | |
| #internal-escalations | Internal | 88/wk | |
| #feature-requests | Product | 54/wk | |
| #billing | Finance | 30/wk | |
| #general | — | 410/wk |
Choose what gets watched
- Pick exactly which Slack channels and customers the monitor tracks
- Map each channel to a customer, with message volume in view
- Set the scan cadence and turn AI sentiment + urgency scoring on
Results
Sentiment + Urgency scoring on every unanswered or delayed-reply message
Worst-sentiment + highest-urgency messages surface first, every cycle
Background fetch runs continuously, so nothing is missed and no one refreshes by hand
By channel, person, or date. Every reviewer can slice the monitor their way
If important messages keep slipping through
We build the AI-augmented operations tools that catch what matters before it gets buried.