The management operating system for the agent-human workforce.
philter is the system your company runs on — and the AI agency that installs it. We design your operating model, build your workspace, and put agents to work beside your people.
Leadership sync
Recorded Tuesday · 68 min
Confirm the security review scope
Awaiting you“Let’s confirm the security review scope before the board pack goes out Thursday.” — Leadership sync, Tuesday
Owner: Maya — she offered to run it. Date: Thursday, from the deadline named in the same sentence.
Team board
12 open- Confirm the security review scopeThu
- Rewrite the onboarding briefFri
Undo is one click, from a record that keeps who decided what.
Always current
Your team keeps working. The record keeps up.
Whatever your team already uses — the call someone recorded, the thread someone forwarded, the group chat, the deck, the sheet somebody maintains — goes in as it is and comes back as tracked work. Nobody retypes it, nobody summarises it, and nobody writes a status update. You stop asking where things stand, because asking is what people do when the record cannot be trusted.
- Recorded meetingsconnected, attached or pasted
- Forwarded emailthe whole thread, not just the last reply
- Group chat exportsthe export file — message by message for WhatsApp today
- Pasted transcriptsfrom anywhere at all
- Documents, decks and sheetscomments, speaker notes, cell values
- Shared file linksa link instead of the attachment — Dropbox and Drive today
- The tracker sheetthe one somebody already maintains
Files are read in place — a deck’s speaker notes, a document’s comments, a sheet’s actual cells. Not just the filename. New channels join this list without anyone changing how they work.
What stops being expensive
Then the expensive things stop being expensive.
Being current is not the point on its own. The point is what it makes possible once the record is true: knowing where everything stands, telling each person what they owe, and producing the documents that used to need a team.
Open it and know where everything stands
Across every project, and what is yours this week. You can act on what you see without checking it first, because the context underneath it is current.
The brief nobody had to write
Each weekday morning, in your own timezone and your own language: what you owe today, what moved, and who is waiting on you. No status meeting feeds it.
Board-grade documents, without the month of interviews
A strategy memo or a board pack normally starts with weeks of interviews to reconstruct what everyone already knows. Here it starts from a record that is already current, already cited and already agreed — which is most of what a consultancy bills for.
And it compounds. Every source that lands is context the next proposal, the next brief and the next document can draw on — so the longer it runs, the less of your company lives only in somebody’s head.
What we do
We build the system. We rewire the work.
Three jobs most companies split across a consultancy, a software vendor, and an internal task force — done as one motion, so the design, the tooling, and the habits match.
Rewire how you run.
Working sessions with your leadership redesign decision rights, structure, and the project-management backbone around what AI actually changes. The output is an operating model your team owns — not a recommendation deck.
Build your workspace.
A management OS shaped to your company on the philter platform, live from week one: meetings, notes, and threads become decisions, tasks, deliverables, and deadlines — with briefs that land before anyone asks for a status update.
Put agents to work.
Agentic workflows with named jobs, defined outcomes, and a human review gate — wired into your stack, with your people trained to delegate to them, review them, and raise the bar.
What you get
What you walk away with.
- An operating model your team runs — decision rights, owners, and a cadence that holds.
- A workspace built for your company, live from week one and yours to keep.
- Agent workflows shipping real work, each with a named job and a review gate.
- A weekly rhythm of briefs and reviews that keeps the system compounding.
The part that matters
The machine proposes. Your team keeps the last word.
Each proposal from a source arrives carrying the sentence it came from and the reason behind each value it filled in. A person approves, corrects or dismisses it in one click; the platform applies some itself and sets aside what the board already records, and says so each time. A wrong approval, whoever made it, is undone from a record rather than deleted.
It shows its source
The quoted line, the meeting it was said in, the day it was said. You can disagree with the evidence, not just the answer.
It shows its reasoning
Why this owner and why this date, stated on the card. Where nothing in the source supports a value, it says so instead of inventing one.
It stays reversible
Undo from the ledger. The record keeps who decided what and when — including the handful of measured, additive updates the platform applies itself, which mark themselves as its work and reverse like anything else.
And that includes your own agents
Your agents get a door — a narrow one.
Connect your own AI to the workspace. It reads the board, drafts the update, proposes the task — and its proposals wait for a person. What it can change outright is only what you granted it — a status, a session, your drafts, the team’s system links — always under your name, and never beyond what you can already see.
- A personal token is scoped to what you grant it, and revoked in a click.
- Whatever it proposes joins the same review queue your team already works.
- A change it proposes is marked as its work on the review card.
Connected agent
Read the board · propose work
Proposed: “Summarise the pricing workshop”
Awaiting a person · filed 2 min ago
Its proposal waits for a person.
The platform
One place where work becomes clear.
Every engagement runs on a live workspace built for you — meetings and notes in; decisions, tasks, deliverables, and briefs out. Agents and people report to the same board.
This week’s priorities
AI brief · Mon 08:00- Lock the vendor shortlist for the data platform — decision owner is confirmed.
- The Q3 partner update draft is due Thursday; last week’s metrics are attached.
- The leadership offsite agenda needs a final read before Friday.
Team board
12 open- Draft the Q3 partner updateThu
- Confirm the vendor shortlistWed
- Circulate the offsite agendaFri
Leadership sync — meeting record
Recorded Tue · analyzed in 4 min
Capture becomes action
Meetings, notes, and email threads become tracked, owned work — proposed for you to confirm, with receipts.
Briefs that write themselves
Monday priorities and Friday recaps, grounded in the live board — no status meetings needed to feed them.
Judgment built in
AI proposals wait for you, and the few exceptions mark themselves and undo in a click — so the board stays trustworthy enough to run the company on.
How we work
Not a report. A running system.
We work inside your company, not beside it: sessions with your leadership, the workspace live from week one, the first agents shipping real work within the month.
Map and design
Working sessions with your leadership: how decisions get made, where work stalls, what agents should own first. The operating model takes shape on the board — not in a deck.
Install
Your workspace goes live and the first agent workflows ship — real jobs with owners, deadlines, and definitions of done. Your team runs it with us.
Run and raise the bar
A weekly rhythm of briefs, reviews, and working sessions. Every cycle, more work moves onto the system — and the system gets sharper.
From the field
What installing philter actually looks like.
A two-hour leadership session becomes a structured record by the next morning — decisions, owners, and open questions, with the uncertain spots flagged instead of glossed.
A forwarded email thread becomes three tracked actions with receipts. Approving them is one click each — a verification pass already checked every proposal against the live board.
Monday, 08:00: a priorities brief grounded in the board lands in every inbox. Nobody wrote it. Friday, the recap closes the loop with what shipped.
A dozen working sessions in, the team runs an operating model that holds without us in the room — named owners, visible work, decisions that stay decided.
Who we are
Built by managers, for managers.
Fifteen combined years supervising work — including a decade at the world’s leading management consultancy.
Every workspace tool is shipping agents, every team is hiring them, and every leader is being asked to manage something they’ve never managed before. We’ve spent our careers learning what makes management work — the discipline is universal: who should do what, what they need to succeed, and how to know they did. The medium just changed.
philter is what we wished existed: the operating system for that job — built, installed, and run by people who’ve managed the work themselves.
- McKinsey & Company
- Sony
- A.P. Moller-Maersk
- Harvard Business School
- HEC Paris
FAQ
Questions, answered.
Is philter a product or a consultancy?
Both, on purpose. The platform is the management operating system your company runs on; the agency is how it takes hold — we design, install, and run it with you. You don’t buy licenses and hope; the system arrives working.
Who is philter for?
Founders and executives scaling AI-native organizations — and the chiefs of staff, heads of operations, and project leaders who orchestrate work across agents, people, and tools.
Do we have to replace the tools we already use?
No. philter sits on top of your existing stack and orchestrates work across it — there’s nothing to migrate.
What does an engagement look like?
Working sessions with your leadership, the workspace live from week one, agents shipping real jobs within the month — then a weekly rhythm of briefs and reviews that keeps it compounding.
What will it cost?
Engagements are scoped individually — the operating-model work, the workspace, and the agent workflows priced as one. The software ships as part of the engagement, not as a separate license.
Is my data secure?
Your data is isolated per organization with strict access controls, and we don’t use it to train shared models.