Yash Tekriwal, Clay's AI use case

Head of Education at Clay

Built a custom AI-powered Slack dashboard that turns 150+ daily notifications into a focused Kanban-style inbox organized by urgency: Action Required, Need to Read, and FYI.

The problem

What was broken before AI

Yash was waking up to more than 150 Slack notifications, and the hard part was not simply reading them. The hard part was deciding which ones deserved action, which ones were worth reading, and which ones could be safely ignored. Slack treated everything like it mattered equally, so triage became its own job before the real work could even begin.

What changed

What the use case made possible

Yash first built a text-based digest with OpenClaw, using mostly deterministic code to pull and group the right Slack notifications while AI handled the more subjective judgment of whether each message was Action Required, Need to Read, or FYI. That worked, but it was still a wall of text. He then used Perplexity Computer to turn the digest into a clean Kanban-style dashboard with direct links back to Slack and an Archive All action for the low-priority column.

Why this matters

Why this use case is worth studying

Yash’s workflow is useful because it shows the difference between summarizing information and reshaping work. A summary can still leave you with a long list to process. His dashboard changes the interface: messages are grouped by urgency, FYIs can be cleared in one move, and the inbox begins to match how his brain wants to work. That is where small personal software gets interesting — it does not need to replace Slack, only make Slack usable for one person’s workflow.

Use this when

When this pattern applies

Use this pattern when an important work stream has become too noisy to trust at a glance. It works especially well when the tool already contains the information you need, but the default interface treats everything like it deserves the same level of attention.

Exponential Builder analysis

Start with the decision you keep repeating.

Yash did not treat Slack volume as the core issue; he targeted the daily judgment call of “do I need to act, read, or ignore?” AI is most useful when it takes over a recurring classification that already exists in your head.

Let code handle facts and AI handle ambiguity.

The workflow separates deterministic retrieval from subjective urgency scoring, which makes the system easier to trust. Builders should resist handing the whole pipeline to AI when timestamps, unread state, grouping, and links can be handled more reliably with normal logic.

A better interface can be the real automation.

The shift from text digest to Kanban dashboard matters because it changes the work surface: act here, skim here, clear this pile. Many AI tools fail because they produce more reading; this one reduces the number of decisions needed to start the day.

Who this is for

Best fit

Operators and team leads buried in Slack

Founders managing too many communication channels

Customer-facing teams triaging messages and requests

People who start the day with notification overload

Teams that need better prioritization without replacing their existing tools

Builders who want a small personal dashboard instead of another full app

What to avoid

Mistakes and warnings

Where this pattern can go wrong if you copy it too literally.

Do not ask AI to summarize everything if the real problem is prioritization.

Avoid building the interface before the underlying data is reliable.

Keep urgency categories simple enough to trust at a glance.

Be careful with tools that mark messages as read or archive items in bulk.

If the dashboard becomes another inbox to check, simplify it instead of adding more features.

Public workflow preview

The shape of the workflow

A high-level look at how the use case works, with the reusable pattern made clear.

Separate the signal from the noise

Yash starts by deciding which notification types matter and which ones usually do not.

Define urgency categories

Messages are sorted into Action Required, Need to Read, and FYI instead of one undifferentiated feed.

Pull the right Slack data

The system checks timestamps and unread state so it only grabs messages that actually created notifications.

Let AI judge intent

Deterministic code organizes the data, while AI decides which messages require attention.

Turn the digest into an interface

Perplexity Computer converts the text digest into a Kanban dashboard with links back to Slack and bulk archiving for FYIs.

Copy the pattern

The reusable idea

Pattern in one sentence

Build a thin AI layer on top of a noisy tool so messages are organized by what you need to do, not just where they came from.

Reusable idea

Yash’s use case is a good reminder that the best personal AI tools often begin with a mental model, not a feature request. Before asking AI to automate something, decide how you wish the work was organized. In Yash’s case, the important distinction was not channel name or timestamp; it was whether a message needed action, awareness, or no attention at all. Once that structure was clear, AI and code could build a small tool around it.

Steal this workflow

Build a triage dashboard for one noisy feed:

Pick the feed that creates the most morning drag: Slack, email, support tickets, task updates, or internal alerts.

Define three urgency buckets in plain language. For example: “Action Required,” “Need to Read,” and “FYI.”

Collect 10–20 real examples for each bucket so the AI has a clear sense of your threshold.

Build a text digest first. Include source, sender, timestamp, short message body, unread/read state, and original link.

Use deterministic logic for collection and grouping. Use AI only to classify each item into the urgency buckets.

Review the digest for a few days before designing the dashboard. Fix bad inputs and fuzzy category definitions first.

Turn the validated digest into a simple board with one column per urgency bucket.

Add deep links back to the source and one safe bulk action for low-priority items, such as archive or mark read.

Test it in the exact routine where the clutter appears, such as the first 10 minutes of the workday.

Suggested prompt

“I want to turn my noisy [Slack/email/ticket/task] notifications into a triage dashboard. Help me design and build a first version. Use deterministic logic to pull only items that actually need review, including source, sender, timestamp, unread/read state, message text, and original link. Then classify each item into three buckets: Action Required, Need to Read, and FYI. Use Action Required only when I need to respond, decide, assign, or do something. Start with a text digest so I can validate the data and categories, then propose a simple Kanban-style dashboard with deep links back to the original items and one safe bulk action for FYI items.”

Field notes

Get new AI use cases in your inbox

A short weekly note on how real people are using AI to save time, make money, build tools, and run their lives.

Related use cases

Keep exploring nearby systems.

Browse all

Marketing & ContentPersonal Productivity

Josh Belanger

Financial entrepreneur and newsletter writer at Belanger Trading

Finance newsletter editing workflow where human-written market thoughts are streamlined with ChatGPT, copyedited with Claude, and refined with custom GPTs to preserve trading-specific language.

ChatGPTClaudeCustom GPTs

Use case

View use case

Personal ProductivitySales & Outreach+1

Hiten Shah

Serial founder at Nira / Crazy Egg / KISSmetrics

Hiten Shah uses ChatGPT Projects as context-loaded assistants for managing up, self-coaching, and turning sales frameworks into custom discovery scripts.

ChatGPT ProjectsChatGPTGoogle Search

Use case

View use case

Life SystemsPersonal Productivity+1

Lucas Werthein, Cactus

Founder / operator at Cactus

Uses AI to turn personal routine data and business context into a lightweight reflection loop, while also using synthetic customer profiles to stress-test ideas, messaging, and assumptions.

ChatGPT

Use case

View use case