Skip to main content
Navigation
Home EXPERIENCE

EXPERIENCE

What AI can do now, and what we have done.

Our in-progress work with today's AI, and past consultation and development engagements, shown separately as in progress and past work. Each case lists its purpose and our actual scope.

EXPERIENCE

What we do now, what we have done.

AI WORKSPACE / AGENT OPSIn progress

A&A in-house operations

AI does the work, people decide. How we run it ourselves.

We combine AI development tools with assistants we can instruct from chat to form the workflows we run in our own work. We keep documents and the reasoning behind decisions, divide the work, and include human review.

Scope: In-house design, implementation, operation, and improvement

Purpose

Try what the latest AI can do on our own work first, shape it into something usable, and use it as the base for applying it to customers' work.

What we are building

  • Set up development environments such as Claude Code and Codex and connected them to daily work
  • Reused knowledge and procedures, divided work among multiple agents, and resumed work midway
  • Recorded documents and decision reasoning, with human review and verification built in

This describes our own operations. It does not claim deployments for customers or a production validation of all products combined.

AI AGENT / SALES AUTOMATIONIn progress

A&A in-house / in development

Telephone outreach, connected to voice AI and a CRM.

We are building telephone outreach automation in-house: voice AI placing the calls, integration with a CRM, and stage-by-stage tracking from call to conversation, appointment, and business meeting. We build it together with a team of AI agents.

Scope: In-house design and development

Purpose

Run the path from first call to business meeting while watching each stage in the numbers.

What we are building

  • Building the voice AI calling system
  • Integrating with a CRM
  • Tracking each stage: call, conversation, appointment, and business meeting

This describes a system we are building in-house. It does not claim customer deployments or results such as call volumes or appointment rates.

LINE × AGENT / CRMIn progress

A&A in-house / in development

Never keep a LINE inquiry waiting, never let one slip.

We are building in-house a system that connects LINE conversations to a customer-management CRM where AI agents do the work. The design aims to cut reply waits, missed replies and handoff gaps between staff, and to remove manual copying into the ledger. We expect staff to focus on judgment and sales conversations, with first-line response continuing through nights and holidays.

Scope: In-house design and development

Purpose

Aim to respond to every LINE inquiry without delay or omission, so staff can focus on judgment and sales conversations.

What we are building

  • Designing first-line response to LINE inquiries and handoff to staff
  • Designing a structure in which AI agents handle ledger records and history upkeep
  • Building it step by step in our own work, with nights and holidays in mind

This describes an effort in progress. It does not claim a running service, customer deployments, or results.

LINE × WEB / PRODUCTIn progress

In-house product / in development

Take requests on LINE, finish the site on the web.

We are developing in-house a service that connects intake and interviews on LINE with design selection, payment, and generation and publishing on the web. The experience of choosing, confirming, and redoing is designed together with the AI processing and data management behind it.

Scope: In-house planning, design, and implementation

Purpose

Join the customer's steps of choosing, confirming, and redoing with the processing behind them as one experience.

What we are building

  • Designed and built intake and interviews on LINE
  • Connected design selection, payment, and generation and publishing on the web
  • Designed the choose, confirm, and redo screens along with AI processing and data management

This describes a service we are developing in-house. It does not present the number of paying customers or conversion rates.

AI TOOL / WORKFLOWPast work

Real estate / anonymized

Connect document review, research, and reporting.

We developed a workflow tool in which AI prepares sourced background research and a draft report from property documents and public data. We designed a step where people first confirm what was read from the documents, screens for review and approval, and a revision-request flow, and we handled improvements based on feedback from the client's staff, functional testing, and handover documentation.

Scope: Contracted design, implementation, improvement, testing, and handover

Purpose

Carry document review, research, and report preparation as one connected workflow, with people making the decisions.

Work performed

  • Designed and built document input and human review and approval screens
  • Implemented the connection between AI research and report output
  • Revised the processing and added checks based on feedback from the client's staff
  • Tested functionality and prepared handover documentation

This describes the development work performed. It does not claim the accuracy of valuation results, current operation, or measured time or revenue effects.

AI ADOPTION / KNOWLEDGEPast work

Advertising company / anonymized

Planning AI adoption around internal knowledge.

We reviewed workflows and internal documents, clarified what AI could handle, and proposed how to use a shared knowledge foundation. The proposal connected each person's AI with a process for returning lessons from daily work to that shared knowledge.

Scope: Paid consultation, workflow review, requirements, and proposal design

Purpose

Clarify how scattered internal documents and staff knowledge could become useful in everyday work.

Work performed

  • Reviewed workflows, internal documents, and current issues
  • Designed roles for individual AI assistants and the shared knowledge foundation
  • Proposed an adoption sequence and a process for capturing new knowledge

This case covers consultation and proposal work. It does not claim a completed rollout or improvements in business performance.

WEB SERVICE / AUTOMATIONPast work

Store marketing / anonymized

A marketing service connecting store operators and stores.

We developed a web service for managing marketing across multiple stores. The work covered an operator admin screen, a store-facing screen, reporting, automation of routine tasks, and a flow for people to review AI-generated output.

Scope: Contracted requirements, design, and implementation

Purpose

Let operators and stores each work in their own role, with people reviewing AI output before it is used.

Work performed

  • Designed and built the operator admin screen and the store-facing screen
  • Implemented data management and reporting
  • Designed routine-task automation and a human review flow for AI-generated output

This describes past development work. It does not claim ongoing operation or improvements in sales, customer acquisition, or workload.

INDUSTRY FRONTIER

How far AI organizations have come elsewhere.

Based on figures and primary sources published by companies outside Japan. These are not A&A results.

Every

  • Every runs six products, a media company and a consultancy with about 25 people
  • Four AI agents handle priorities, meeting tasks, goal setting and growth reporting
  • Its AI coworkers each have a name and a manager, shown in a public org chart

Published 2026-04

Sierra

  • Sierra merged its role-specific AIs into one internal agent, Pinecone
  • Since its first commit in March: 75,000+ sessions for 600+ people
  • Reports that 70% of its pull requests are opened through Pinecone

Published 2026-07

Stripe

  • Stripe runs unattended coding agents called Minions
  • Over a thousand pull requests merged each week are entirely Minion-produced
  • A run starts in a Slack message; humans review instead of writing code

Published 2026-02

Vercel

  • Vercel runs more than a hundred agents in production
  • Its most-used internal tool is a data agent answering 30,000+ questions a month
  • The sales lead agent costs about $5,000 a year; one engineer maintains it part-time

Published 2026-06

WorkOS

  • WorkOS is mid-transition to an AI-native organization
  • Reports 62 team agents built internally across every function
  • A recruiter now builds their own sub-agents

Published 2026-08

Lovable

  • Lovable migrated its own website platform with AI
  • One developer and a pack of agents moved a site of about 400 routes
  • The author notes it would have been a multi-team effort two years ago

Published 2026-08

Cognition

  • Cognition builds Devin, its AI developer, with Devin itself
  • Reports merging 659 Devin PRs into its own codebase in a single week
  • Adding a label to a bug ticket starts the investigation automatically

Published 2026-02

TextYess

  • TextYess rewired the whole company around AI agents in twelve months
  • AI reads an issue, fixes it and opens a pull request on its own
  • As development sped up, the bottleneck moved to deciding what to build

Published 2026-05

Anthropic

  • Anthropic published a survey of 132 of its engineers and researchers
  • Self-reported: Claude is used in 59% of their work, with +50% productivity on average
  • 27% of Claude-assisted work is tasks that would not have been done otherwise

Published 2025-12

Ramp

  • Ramp wants every team member to be a builder, not just a "button pusher"
  • Reports 90% monthly active usage of Notion AI
  • Ramp Assist is described as resolving over 90% of requests

Published 2025-09

STRATEGY TO IMPLEMENTATION

Business understanding, carried through to delivery.

The goals and context shared in our conversations carry through requirements, design, and implementation. Keeping that understanding together reduces repeated explanations and coordination, helping us build something aligned with your purpose.

  1. UNDERSTANDYour business
  2. DESIGNRequirements
  3. BUILDDevelopment
  4. ADVISEOngoing advice

Move into the next decision and implementation with your business context intact. We keep the purpose connected across advice and development.

Understand the contextSet prioritiesMake it useful

LET’S TALK

Let’s start with your business.

Use the button to book a time that suits you, right away. You do not need AI expertise or a ready-made plan.

Book a free assessment

Free, online, pick a time that suits you