I build AI agents,and transform the business.

Pre-meeting research

Automatically researches the other party and logs reviews and facility details into the record.

Automatic deal logging

Reads every Zoom sales call and automatically writes the activity record into the CRM for you.

Auto-attached recording links

Attaches each meeting's recording link to the record automatically, so no one has to go hunting for it.

Automatic attendee detection

Identifies who joined each call from the recording and logs it, separating data entry from reporting.

Automatic deal-stage updates

Updates each deal's stage and date by fixed rules, keeping reports from drifting out of sync.

Sales win-pattern analysis

Breaks down calls, appointments and wins by rep and period to show which actions actually close deals.

Target-account ranking

Ranks target accounts on one screen by expected revenue and likelihood of winning.

Public-data collection

Gathers public government and Hello Work data every night and compiles it into a ledger.

Client-change detection

Catches name changes, relocations and closures from public data and flags them with sources.

Neglected-account check

Lists each rep's last contact with every client, so neglected accounts are obvious every morning.

Neglect-risk ranking

Ranks neglect risk across all reps, making it easy to decide who needs help right away.

Churn-signal diagnosis

Cross-checks every record with AI to flag likely churn, each one backed by evidence.

Call & complaint detection

Scans every call transcript for outcomes and early signs of complaints, quoting the exact lines.

Dormant-client outreach list

Builds an outreach list by contract-end date and double-checks for anyone missed.

Standing-meeting logging

Logs each client meeting's conclusions and next steps automatically as calendar events and to-dos.

Client-email logging

Picks out client correspondence from the shared inbox and files it into the record.

Contact clean-up

Moves orphaned contacts to the right account and groups them by client.

Automatic meeting notes

Fills each deal's minutes field with the meeting's key points, cutting down on manual notes.

Automatic applicant intake

Logs every new application into the CRM the moment it arrives, with no duplicates, so reps can call right away.

Email-triggered applicant intake

Logs applications from job boards with no API, triggered by their notification emails.

• 
• Airwork
• Job Medley
• 

Daily first-response tracking

Counts calls, connects and interviews set up for each application every day, surfacing anything left untouched.

Job-listing update proxy

AI updates job listings with no update API, operating the client's own computer to do it for them.

• 
• Playwright

Recruiting-site migration

Moved the site from a paid no-code platform to our own build, cutting the fixed monthly cost.

• STUDIO
• 

Dropped-task monitoring

Constantly checks that every automation is still running and flags only genuine problems.

Daily to-do sheet

Rebuilds each person's to-do list every morning, so all they do is fill in their own row.

Automatic fix pointers

Flags mismatches between systems down to the exact field, current value and the fix.

Meeting-recording sweep

Matches calendar events against recordings every day and passes only complete ones to the note-taker.

Weekly logging check

Cross-checks calendar events against records every week and reports only the gaps, skipping manual audits.

Automatic record matching

Identifies the right client and deal from the meeting title and attendees, and asks the team when unsure.

Duplicate & empty-record clean-up

Checks for duplicates before every entry and clears out empty records left by cancelled meetings.

Direct-to-attendee follow-up

Asks unclear points only to the person who was actually in the meeting, without stopping anyone else's work.

Midday & evening alerts

Reports only when something needs action, at midday and in the evening — silent otherwise.

Decisions & action items from recordings

Transcribes meeting audio entirely on-device, then checks the decisions and action items the following week.

• Plaud
• whisper.cpp

Deployment-gap monitoring

Finds, every day, any change that was pushed but never reached production, and flags only real problems.

Automatic board-number rollup

Rolls CRM numbers into the board report every night, and halts if anything swings too far.

Standardized plan-vs-actual rules

Maps a fixed counting rule to every metric, so month-to-month comparisons never drift.

Target collection & transcription

Moves field targets into the record through planning, approval, writing and read-back, in that order.

Automatic revenue import

Pulls in sales every night from a storefront with no sales API and shows them on the dashboard.

• 
• Playwright

Daily M&A deal collection

Pulls deals from news and US filings every morning and extracts the parties and amounts.

Company fact ledger

Gathers facts about a company from its own site and filings into a ledger with cited sources.

Parallel sourced research

Runs several AI researchers in parallel to gather material and build a ledger checked against the original text.

Second opinion on hard calls

Before any irreversible decision, a separate AI is always required to raise an independent counter-argument.

Automatic competitor benchmarking

Enter a facility name and area, and it turns competitors' going rates into a sourced report.

• Indeed
• Kyujin Box
• 
• 

Baseline hiring-strength diagnosis

Examines the market, competitors and the client using only public data, and produces an evidence-backed diagnosis.

• 
• 
• Firecrawl

Auto-posting with no API

Posts automatically into an admin panel with no API, then reads the published content back to confirm it.

• 
• Playwright

Company-research radio at scale

Automatically produces an audio episode for each company and links from its description to the full article.

• YouTube
• 

YouTube publishing proxy

Enters the title, description and thumbnail exactly as briefed, then verifies them after publishing.

• YouTube

X analytics & reply drafts

Diagnoses performance from X's numbers and drafts replies, with a person approving before anything posts.

Daily search-ranking watch

On any morning search rankings drop, it points to exactly what needs fixing in the admin panel, right down to the page.

Automatic video QA

Catches misreads, restarted lines and missing rights credits automatically before the final video export.

• whisper
• ffmpeg

Pre-publish QA gate

Always cross-checks every number against its source, and blocks approval if anything doesn't match.

Design-vs-build comparison

Checks every outsourced screen against the design by capturing and comparing them one by one.

• Figma
• Playwright

Full-site screen mapping

Lays out every existing screen to find dead ends and pages with no way in.

• Figma
• Playwright

AI mystery-shopper UX check

An AI acts as a fictional customer walking through the screens and reports confusing spots before launch.

• Playwright
• 

Document-review admin screen

Built a flow that approves identity documents only once every condition is met.

• 
• Next.js

LINE job-listing AI maintenance

AI handles the server upkeep, so job listings can be fixed with nothing but a LINE chat.

• LINE

Never logs into ended accounts

Checks the contract before every login and never lets anyone into an account outside it.

• Airwork
• Job Medley
• 

AI off-limits action list

Deletions, external sends, billing and production changes never run without a human's approval.

Secrets stay hidden

Shows only the count and location of keys and tokens, never a single character of the actual value.

• macOS Keychain
• 

Multi-stage PII screening

Catches secrets and personal data at several checkpoints — before sending and at the end of every task.

Secret-commit prevention

Spots anything that looks like a key before it's committed and blocks the commit outright.

Quarantine instead of delete

AI never deletes during clean-up — it quarantines with a log and provides the steps to restore.

Automatic bot-config backups

Backs up every AI bot's configuration automatically each week, so a break can be undone right away.

Internal AI rulebook

Keeps one master set of rules, so every AI tool we use runs by the same standard.

Independent-AI review ledger

Never takes an AI's "done" at face value — a separate AI checks the work and logs it.

Reinvention prevention

Always makes the AI search for existing work before building anything, so no past effort goes to waste.

Milestone procedures & resume points

Turns kickoff, progress, handoff and completion into set procedures, so any task can resume mid-way.

Feedback memory

Logs every piece of feedback received and re-reads it before each task to stop repeat mistakes.

Staged review for external writing

Separate AI reviewers check tone, depth and whether it reads as AI-written, one stage at a time.

AI bot operations audit

Reviews AI bot activity by reading logs alone, flagging wasted runs and gaps in a summary.

Request routing desk

Takes every request at one window and routes it to the right AI, so no one has to guess who to ask.

AI-settings rollout

Pushes the leadership-approved AI configuration to every employee's PC in a single action.

In-house AI training

A complete in-house training package built for people who aren't engineers, slides and script included.

AI-tools directory app

Searches every in-house AI procedure and tool in plain language and launches it in one click.

Ask AI over LINE

The CEO just sends a LINE message, AI returns the finished work, and a confirmation closes it out.

• LINE
• 
• 

AI-queryable data endpoint

Exposes our own data through an endpoint that Claude or GPT can query directly.

• 
• API