Agents built for the company you already run.

We start from agents that run our own business every day, then tailor them to your systems, your data and your approval rules.

Talk to us
An operator at her workstation in a control room, with two AI agents drawn as red mesh figures at the workstations beside her. Labels: You, approve what matters. Agent, reads the reports. Agent, drafts the update.
You: approve what matters
Agent: reads the reports
Agent: drafts the update

In use

Inside Dosan, Oct 2026

Our agents serve 9 brands. Seven of them work alongside our marketing team every day: 2,384 requests in the last 30 days, about 340 video ads since August. We sell what we use.

Agents

Starting points

Every engagement starts from an agent we already run. We tailor what it reads, what it produces and where your people approve.

Content production team

In production at Dosan

Turns a product brief into finished video ads, stage by stage: brief, research, script, production, verification, upload.

Reads:
Product info, brand rules, past results
Makes:
Scripts, videos, upload-ready files
You approve:
Script and pilot

Marketing copilot with team memory

In production at Dosan

Handles marketer requests and saves their corrections as shared memory, so every agent works from the same rules.

Reads:
Requests, chat feedback, brand rules
Makes:
Drafts, analyses, saved rules
You approve:
Each result, as the requester

Audience pre-test

Research-validated

Estimates click-through for new creatives by simulating audience segments, before budget is spent. Presented at the RecSys 2026 GenAI-eCommerce workshop; accepted at the NeurIPS 2026 SocialAgent workshop.

Reads:
New creative, target segments
Makes:
Estimated click-through by segment
You approve:
What runs

Industries

Use cases

Work we see owners and operators doing by hand, that an agent could carry. Each one starts from your own data.

Apartment keys over a calculator and rent paperwork

Multifamily

US owners and property managers

  • Read each monthly manager report and flag variances
  • Track loan covenants and upcoming tests
  • Draft LP updates for you to review and send

Starts from:Owner oversight agent

A guest checking in at a hotel front desk

Hotels

US owner-operators

  • Pull the numbers out of each property system
  • Bring them together in one owner view
  • Flag variances against budget, early

Starts from:Owner visibility agent

A technician in safety glasses working on a production line

Manufacturing

Korean manufacturers and their US plants

  • Write the daily production and quality report
  • Summarize it for HQ in Korean and English
  • List the exceptions for the morning meeting

Starts from:Plant reporting agent

How we work

One workflow first

Discover

We sit with your team, map the work and pick one workflow with a clear payoff.

You give:
Time with the people who do the work.
You get:
One workflow, scoped.

Build

We tailor an agent to your systems and data, with checkpoints where your people approve.

You give:
Access to the systems you choose.
You get:
An agent shaped to that workflow.

Run

The agent works every day. You see what it did and where it needs a decision.

You give:
Approvals at the checkpoints.
You get:
The work, done daily.

Improve(then back to Run)

Feedback and results become memory the agent uses next time.

You give:
Corrections, as you would to a colleague.
You get:
An agent that learns your rules.

Your data stays yours.

Agents reach only the systems and files you approve.

Contact

Reply within one business day

Pick one workflow.

Tell us what takes your team the most time. We will show you the agent we would start from.

Thank you.

We will reply within one business day.

Industry
Goes to the Dosan founders.