What one person told their AI about a client never reaches a colleague's AI.
In agencies, studios and consultancies of two to ten people, several people work on the same client, each with their own assistant. The brief goes round on Slack and nobody reads it. The work comes back asking the same questions again, or contradicting what was agreed.
"The current method of sharing these files via Slack is ineffective because team members do not consistently read them."A founder, from the notes of his first call. The cost he named: his colleague stops using her own judgement and copies his examples.
- Every session starts from zero. People spend 5 to 10 minutes re-explaining the client each session, for each client.
- Output ignores the brand or the client's template.
- One method gets copied into every client project, and the copies drift: "Four Projects meant four copies. Fix a phrase in one and the other three were quietly wrong until I remembered."
- Months in, the brief goes stale. The client changes the target on a call, one copy gets updated, and the next piece of work comes back confident and wrong.
One workspace per company: the files, the tools, and each person's own agent.
Everything the team works from
Client briefs, proposals, decisions, brand rules, decks, sheets and data, kept as files. Every change is versioned, and every machine has the same copy.
Tools made for the firm's work
Documents, sheets, decks, contact lists and progress widgets, shaped to the company and working directly on its files.
Their own agents
Claude or ChatGPT, on the subscription each person already pays for, working in the Syns app. Nobody needs a terminal.
- Sign in with Google
- Import from Drive
- Accept the proposed structure
- Connect the agent
- Ask for the work in plain words
- The agent reads the client's briefing first
- The deliverable builds in the client's folder
- Teammates join, each with their own agent
- Comment on a file and tag an agent
- Every agent reads the same current brief
Forever, for one person.
From the second team member ($10 billed yearly). The AI is paid to its vendor. Syns never marks up tokens.
See it work at syns.app.
Agents already do client work. What they work from hasn't caught up.
The assistants are already paid for
Small client-service firms already pay for Claude and ChatGPT and use them on client work: 30% of Dutch business-services firms with 2 to 9 people use AI, and 52% in information and communication (CBS, 2025).
Agents are becoming the reader
Tools built to share knowledge between people don't reach the agent. The agent can't read the call notes, the inbox or a colleague's head. What it works from has to be written down where it reads.
Memory is built for one person
The assistants' memory and project features are built around one person and one vendor. The pain starts with the second person on the same client, and that is where the paid plan starts.
The window is months, not years
Every assistant vendor is building shared context inside its own product. Nobody holds the neutral workspace that crosses vendors and companies, and that gap won't stay open for long.
Hard to copy for the people best placed to try.
Their agents, their subscription
The team pays its AI vendor for the agents, not us, and that holds even for a team on a single vendor. An AI vendor gains nothing by making a rival's model better, so its own shared context stays tied to that vendor.
A workspace made for the company's task
Custom tools and interfaces for exactly the work the firm does, with all its data in one place for the agent. The work gets done there; it isn't just a memory the agent reads from.
Local files
For the agent, the workspace is files on its own filesystem: no tool calls, no round trips, and always in view. That is more efficient than anything reached through MCP or a CLI.
It follows the work across companies
Client material sits in the client's accounts, and the method in the firm's own. Products built around one company account can't follow the work across that line.
It knows what every agent read
It records which agent read which version of which line before it wrote. That record is what lets Syns catch a brief drifting from what was agreed. Holding it takes both the sync and the read record, and no one else has both.
Every workspace teaches the next
Structures, tools and interfaces that work for one kind of engagement become the starting point for the next client.
A beachhead that can carry $1M a year in the EU, then wider.
- Each paying team brings $24 to $35 a month, and gross margin is estimated at about 95%, so cost doesn't set the price.
- Growth comes from inside teams: the paid plan starts at the second person, so every colleague invited is a paid seat.
- Beyond the beachhead: IT consultancies (~146,000 EU firms of 2 to 9, already working with agents), firms of 10 to 49, and developer teams.
From official CBS, Eurostat and US Census data. Every figure is an upper limit.
Client work is the first vertical, not the category.
Syns is where agents keep what a person or a team has decided and knows: persistent, shareable and reusable. The written context behind all knowledge work is the category.
Over time, Syns becomes the workspace every assistant works in by default, with a growing library of proven structures and tools that each new workspace starts from.
The plan, and where we stand on it.
What we are proving next: that client-service teams use Syns every week, and then that they pay for it.
Teams work from Syns, by 31 Jan 2027.
- 9 Oct 2026Problem conversations with client-service team ownersactive
- 31 Oct 2026The Syns app works for a client-service teamactive
- 15 Dec 2026Three design partners, client-service teams using it every weekplanned
- 31 Jan 2027Teams active, and the first team payingplanned