What we have made.
What we have connected.

We are one lead, one producer, an intelligence, and a bench of over 100 agents. Every line here is something we made or connected, for a client or for ourselves.

It is a list, not a portfolio. Search it for a platform, a channel, or a kind of work.

What we have made

For clients.

• A site rebuilt for a services company

  Three working versions shown at the first meeting. Six rounds after that. Twenty nine pages, moved onto WordPress so their team can edit it.

• A pitch won with a working prototype

  Built for the prospect’s real problem and shown live, in place of a deck.

• A marketing operations stack, connected

  Five platforms and twelve workflows. Project tracking, notifications, and reporting run by themselves, and about ten hours a week came back.

• A data warehouse with dashboards

  More than eight systems feeding one warehouse, with dashboards leaders can read on one screen.

• An AI search audit

  We asked the engines a company’s own customer questions and recorded who got named. The company was named once in sixteen answers.

• A portal that carried an audit to a client call

  The findings, the questions, and the next steps on one private page, ready before the meeting.

• A boutique’s catalog moved into Shopify

  A point of sale export turned into 261 products, with theme directions built on a development store and a cheat sheet so the owner can run it.

• An AI workspace for an agency

  Separate client workspaces, specialized agents, and connected business systems. Daily workflows for briefings, time reports, and proofing, with an operations guide and training for the team.

• A tracking and consent audit

  Every tag on a site found, and checked against what fires before a visitor agrees.

• A media mix model

  The data checked, the model built, and scenarios planned for where the next dollar should go.

• A survey research pipeline

  From the raw survey to statistics, segments, coded written answers, and the finished report.

• Monthly business reviews from the warehouse

  The numbers pulled and the review built without a person assembling it.

• Estimates from real hours

  Project estimates built from what past work took.

• Three ad seats, opened and run by machine

  Google, Microsoft, and OpenAI. Each account opened, connected, and given a written runbook.

• Google Ads, audited

  An audit against about ninety rules, and a loop where every change is proposed, approved, run, and checked.

• Meta, set up and audited

  An audit against about seventy rules, with the pixel and the server side connection checked.

• Ads inside AI answers

  A first paid campaign live on a new kind of ad seat.

• Mail sent by data

  Print ready pieces mailed by machine, triggered by a list.

• Social publishing on four platforms

  Drafted, approved, and scheduled to X, Instagram, TikTok, and YouTube from one place.

• Instagram comments, read and answered

  Through the platform’s own connection, for two accounts.

• A brand run every week

  Scripts, cards, and video, posted every Friday on four platforms, with each platform’s proof kept.

• A weekly journal

  One piece a week, written, designed, scheduled, and checked.

• Tracking which AI engines name a brand

  The same questions asked of five AI answer engines on a schedule, with every answer recorded.

• This site

  Built with the method it describes: a draft, a conversation, and the next draft.

• Sites for a family of brands

  Splash pages and brand sites, each with its own mark and voice.

• A brochure site for a boutique retailer

  Live on their own hosting, with a second location added later.

• Private portals with a passcode

  A gate, a record of who visited, and a kit so the next one takes an hour.

• Traffic across seventeen sites

  Counted on every site and pulled into one report every morning.

• A company renamed

  The name, the mark, the mail, the accounts, the sites, and every agent moved to the new name.

• Brand books and marks

  A mark, colors, type, and voice, written down so anyone can apply them.

• An icon system

  One set of icons with rules for size and placement, drawn once and reused everywhere.

• Images and video made by AI

  Product and brand images, and short motion pieces, made to a brief.

• Hats and business cards

  Designed, proofed, and ordered through the printer’s own system.

What we have connected

Getting AI and your systems to talk, so people can do things faster and easier than before.

• Trello

  The boards report to the team. What is due, what moved, and what is stuck arrives every morning.

• Slack

  Updates and alerts arrive where the team already talks.

• Webvantage

  Hours, revenue by client, and staffing, read straight from the system of record.

• Paycom

  Who is on the team, read from the HR system, so staffing questions get a real answer.

• Snowflake

  One warehouse an agent can ask in plain words, and get the number back.

• Domo

  Dashboards fed from the warehouse, and read back by an agent for reports.

• Google Workspace

  Mail, calendars, files, documents, sheets, and slides, read and written by an agent inside set limits.

• Google Ads

  Every account under one manager account, read by machine for reporting and audits.

• Microsoft Ads

  The account run by machine, with a written runbook.

• OpenAI ads

  Campaigns in AI answers, created and read by machine.

• Meta

  Instagram comments and ad data, through the platform’s own connection.

• Metricool

  One approval schedules a post to four platforms.

• Google Analytics

  Site numbers read by an agent, so a question gets an answer without opening a report.

• Google Places

  Listings and reviews for every location, pulled on demand.

• WordPress

  Content managed by machine, and plain pages moved into fields a team can edit.

• Shopify

  Products, themes, and photos managed by machine.

• Klaviyo

  Email results read straight from the platform into the report.

• Salesforce Marketing Cloud

  Email programs, audiences, and results read without an export.

• SurveyMonkey

  Survey answers pulled and analyzed without a spreadsheet in between.

• Figma

  The design file read directly, so what gets built matches what was drawn.

• Gamma and PowerPoint

  Decks built from the data, and old decks read back in.

• BugHerd

  A client clicks the thing they mean, and the note goes straight into the build.

• Lob

  A list becomes mail, printed and sent.

• Printful

  A design becomes a product, proofed and ordered.

• Vercel

  Sites go live on approval, and traffic is read every morning.

• Telegram

  The producer and the lead talk by chat, from a phone.

• Crystal Knows

  A reader’s profile shapes how a message to them is written.

• draw.io

  System and process diagrams made from a description.

• Public economic data

  Public numbers fetched and joined to a client’s own.

• Search data

  Rankings, keywords, and links pulled for audits.

• Systems with no connection of their own

  An agent works in the logged in browser, the way a person would.

What we run on

The department behind all of it.

• An AI producer on his own server

  He has email, chat, a calendar, and a schedule of his own. He takes the brief, staffs the work, and brings it back.

• A bench of over 100 agents

  Each one built for a single job, in four teams: strategy, creative, execution, and operations.

• A record of every job

  Each piece of work leaves a record. Twenty four records across twelve jobs in its first month, swept up and summarized without anyone asking.

• A monitor that texts a phone

  It checks every unattended job every thirty minutes, and alerts a phone when one stops.

• A weekly review that writes itself

  Assembled from the week’s records and sent before the Monday meeting.

• A library of 1,700 public APIs

  Mirrored on two machines and updated every day, for scoping builds that need outside data.

Nothing in the log matches that yet. Ask us. We may have done it and not written it down.