Brigham.ai

The AI chief of staff

Companies run on follow-through.Brigham.ai does the work, and answers for every step.

An AI chief of staff that knows your business, works across your systems, and stops for your approval before anything that matters.

● 06:02 / Brigham.ai, today

Overnight: 2 new leads scored, 1 deploy failed and rolled back.

Chief of staff

A business is never one task. It’s the hundred handoffs between them.

Brigham.ai holds those handoffs: the thread nobody answered, the deal that went quiet, the deploy that failed at 3 a.m. It notices, does what it’s allowed to, and brings you the rest.

One operator. Every system you already run.

Founders · Sales · Marketing · Engineering · Ops · Finance

Founders
07:00 · Briefing

Good morning. 4 meetings, 3 follow-ups, 1 deal gone quiet.

Prep attached →
Operators
Infrastructure
Needs approval

Send 3 follow-ups to quiet deals

Approve
Sales
Signal, not noise
Memory · decision

“Push to main doesn’t deploy. Trigger it and confirm the image.”

source: ops notes · verified 3d ago
Holding companies
Every connection, scoped
Estate health

38 sites up · 1 rolled back · 0 down

● all clear by 06:14
Your inbox, private
Teams
Executives
01

It knows the business.

Companies, brands, people, deals and decisions, held as memory with a source and a last-verified date. Nothing becomes fact without a person or a verified source.

02

It works where you work.

Gmail, Calendar, the CRM, Drive, GitHub, your CMS and your analytics. Each connection is yours, revocable, and starts read-only.

03

It does the work.

It plans, executes and checks: briefings, follow-ups, reports, fixes, drafts. When something needs judgment, it brings you the decision rather than the task.

04

It answers for itself.

Every step is audited and costed. Budgets are checked before each model call, not after. You can see exactly what it did and what it cost.

Morning briefing · 07:00

Wake up briefed.

Before you open your inbox, Brigham.ai has read it. Your meetings come with prep, your follow-ups are drafted, and anything that went quiet or went down overnight sits at the top. Each briefing is built as you, so it only contains what you’re allowed to see.

  • Today’s meetings, with attendees matched to people, companies and deals
  • Follow-ups you owe, cross-checked against the CRM
  • Deals you own that have gone quiet
  • New leads, and what Brigham.ai already found on them
  • Estate health: what’s up, what’s down, what changed
Brigham.aiTue · 07:00

Good morning. Here’s your day.

Meetings · 4
09:30Pipeline sync2 deals at risk
11:00Intro call, Northwindnew lead, researched
14:00Board prepdraft ready
You owe · 3
↩Reply to Dana re: renewal termsdrafted
↩Send the SOW to Halvorsendrafted
Overnight
●1 deploy failed, rolled back at 03:12fix on a branch

A live run

Every step, on the record.

Ask once. Brigham.ai plans, calls the tools it’s allowed to, and stops at the line you drew. Every call has an audit row, every run has a price.

run_8hd2k · standard tierworkspace: holdco · actor: you
You

What’s gone quiet in my pipeline? Draft follow-ups for the top three.

  1. 01Plan

    Read the pipeline → check last touch in mail → draft three replies → stop before sending.

  2. 02crm.pipeline_summaryreadauto · in scope412ms

    6 deals idle 14+ days · $412k open

  3. 03gmail.search_threadsreadprivate to you690ms

    Last touch: 19, 22 and 31 days ago

  4. 04gmail.draft_reply ×3reversible writeenabled by admin1840ms

    3 drafts saved to Gmail · nothing sent

  5. 05gmail.sendsensitiveneeds approval

    Bound to payload sha256:9f2c…e41a

    Change one word and it asks again.

It doesn’t suggest the work. It hands it to you finished.

Reports & board packsInvestor updates and weekly reviews where every figure cites its source.
DashboardsPipeline, traffic and operations, across every brand you run.
Code & pull requestsBranch, fix, test and open the PR. You approve the deploy.
Drafts & outreachReplies and follow-ups saved as drafts, ready for one look.
CRM hygieneNotes logged, tasks created, stages moved, all reversible.

Use cases

Briefed for the seat you sit in.

The whole business, in the time it takes to pour coffee.

Brigham.ai reads across pipeline, traffic, deploys and inbox so the first thing you see is what changed and what needs you.

Ask it

  • “Brief me on the week ahead, and what I’m behind on.”
  • “Draft the investor update. Cite every number.”
  • “Which brands lost search traffic this month, and why?”
  • “What did we promise customers last week that isn’t done?”

Trust

Secure by code. Not by prompt.

01

Policy is code, not prompt

Every tool call runs one pipeline: look up, validate, check policy, approve if needed, execute, verify, redact, audit. The model never decides authorization.

02

Three action classes

Reads run inside granted scopes. Reversible writes run only where an admin switched them on. Sensitive actions always wait for a person.

03

Approvals bound to the payload

An approval covers the exact action, hashed. Change a word in the email and it asks again. Links in email open a confirm page and act only on a click.

04

Untrusted content stays data

Email, web pages, documents and notes are labelled as data. Nothing that arrives through a tool can approve an action or widen a scope.

05

Your inbox is yours

Each person’s mail and calendar are private to them, workspace admins included. Sharing a thread is an explicit act.

06

Protected domains

Finance is granted to a named person, never to a role. Anything that touched it inherits the label: summaries, memories, briefings, subtasks.

07

Credentials never in context

Encrypted at rest, referenced by id. Keys never enter the model’s context, the logs or a tool result.

08

Delivered means delivered

A provider accepting a message isn’t delivery. Brigham.ai reports “sent” when the delivery event arrives.

Built where it runs

Proven on our own company first: dozens of brands, several legal entities, one workspace.

40+brands run as entities in one workspace
1security boundary per company. Nothing crosses it
0emails sent without a person approving
100%of tool calls audited, with arguments redacted

How it works

Connect. Brief. Delegate.

1

Connect

Sign in with your company account and connect your own inbox, calendar and tools. Everything starts read-only.

2

Brief

Ask anything about the business, or let the morning briefing come to you. Answers cite where they came from.

3

Delegate

Turn on the actions you trust. Brigham.ai does them, and holds anything sensitive for your approval.

Everything included · Full capability map →

Core

The chief of staff

Sign in through erp.io. One agent that answers from your live systems, with every step audited and costed.

Your day

Mail, calendar and CRM

Meeting prep, follow-ups owed, replies drafted or sent on approval, and a morning briefing for each person.

Memory

Business memory

Documents, decisions and relationships with citations, conflict flags and last-verified dates.

Control

Automations & approvals

Schedules, triggers and multi-step workflows that pause for a person before anything sensitive.

Agents

Specialist agents

Sales, Marketing, Dev, Ops and Research agents, delegated by the chief of staff. Delegation never widens permission.

Scale

Portfolios & compute

Parent companies across subsidiaries, sandboxed builds and apps, and proactive reviews with evidence.

Every company deserves a chief of staff. Most will never hire one.

The best operators have someone who sees the whole board, remembers every promise and keeps the small things from becoming big ones. Brigham.ai makes that person available to every team, with the guardrails a real business needs.

Questions, answered.

What is Brigham.ai, exactly?

An AI chief of staff. It keeps knowledge of your business across sessions, connects to the systems you already run, and does authorized work: it plans the work, does it, checks the result and leaves a record.

How is it different from a chat assistant?

A chat assistant answers what you paste in. Brigham.ai reads your systems directly, remembers what it learned, takes actions you’ve allowed, and stops for approval on anything sensitive.

Can it send email or spend money on its own?

No. Sending email, deleting, publishing, deploying to production, changing ad budgets, moving money and changing permissions are all sensitive actions. They always wait for a person, and the approval is bound to the exact payload.

Who can see my inbox?

Only you. Your mail and calendar are private to your own runs, and workspace admins don’t get access. Sharing a thread is an explicit, per-thread choice.

How do you keep finance data contained?

Finance is a protected domain. Access comes from a grant to a named person, never from a role, and every layer checks it in code. Anything derived from finance data inherits the label, and admin views show counts, not rows.

What does it connect to?

Gmail, Google Calendar, Drive, the CRM, GitHub, your CMS, Search Console, Analytics and Google Ads to start, plus any service with an MCP server as a governed connection.

Which models does it use?

Brigham.ai runs a provider-neutral agent loop, so memory and orchestration don’t depend on one vendor. Cheap, fast tiers handle classification and monitoring; deeper models run only when needed.

How do I get started?

Sign in through erp.io and connect your own accounts. Everything starts read-only; you switch on the actions you trust, and anything sensitive waits for your approval.

Hire Brigham.ai.

Give your company a chief of staff. Leave your work email and we’ll set up your workspace.

Get started

Sign in with erp.io. Read-only to start. No spam, ever.