Triage & chasing
Open threads get chased without you asking. What needs a decision reaches you ranked, not raw.
systems that work
Clockwork. Coordination that stops landing on you.
AI and automation wired into the tools you already use, then run on a month-to-month retainer. Your accounts, your data, your documentation.
Installed remotely in weeksRuns inside your accountsMonth-to-month retainer
the four phases ● standby
Full integration takes access. Real access takes trust, on both sides. Without both, this stays a surface fix instead of an actual system.
How Clockwork ticks.
Phase 01 · Audit Map the coordination load
A few calls with you, access to your tools, and time to review how work actually runs. The output is a written plan: what gets wired, what gets replaced, what stays human-led, what it costs, and what you keep if we stop. What I look at
audit · what I look at
For small teams this is often founder-only, the simplest version of the audit. Bigger teams mean more people in the room and more calls, since the coordination load sits in more places. Usually wraps inside two weeks, faster or slower depending on how quickly you can get on calls and hand over access. That's not just a scheduling question: handing an outsider access to your systems is a real ask, and it's fair to take the time that takes. What gets looked at: your PM, people ops, data, and comms tools, where AI is already in use, where the risk sits, what's worth automating, and what should stay human-led.
Phase 02 · Wire Install the layer
What gets replaced and what stays gets decided in the audit, not assumed going in. Some tools survive because switching costs more than it's worth. Others get replaced because the lock-in was already costing you, you just hadn't paid the bill yet. Whatever survives that call gets automations, AI steps, and reporting wired around it, with review gates wherever judgment or sensitive data is involved. How it goes in
wire · how it goes in
The plan from the audit decides this phase, so there are no surprises mid-wire. Connection work first: automations between the tools that survived. Then the AI steps, each behind a review gate until it has earned trust. Reporting comes last, once there is something real to report on.
What the layer does Phase 03 · Run Operate the rhythm
Clockwork isn't a setup project that ends at handover. The retainer is where the tools meet real work: triage rules, prompts, and review gates get adjusted and trained on what actually comes through, not just what the audit predicted. AI integrations need time and iteration to become useful. Skip that, and they either get in the way or perform productivity while contributing little. You see a decision queue, not a message flood. A sample week
run · a sample week
Monday opens with the queue already sorted: what landed over the weekend, what is blocked, what actually needs you. Some of it is sorted wrong. A thread got chased that should have been left alone, a summary pulled the wrong numbers. That is the actual work of the retainer: each miss gets documented, the rule or prompt behind it gets fixed, and that mistake stops repeating. The team is training too, learning when to trust the queue and when to override it. Friday, the report assembles itself: what moved, what stalled, what got flagged for review. Early weeks are heavier on ironing than automating. This is the shape, not the contract.
Phase 04 · Handover Yours either way
No lock-in is the point, not a caveat. The system runs in your accounts, documented from day one: workflow map, runbook, automation inventory, review rules, access notes. The retainer can keep running as long as it earns its keep. It just doesn't have to, because you and the team can run this without me. What you keep
handover · what you keep
Handover is not an event at the end, it is how the system is built from day one.
- The system itself. Runs in your accounts and keeps running. Nothing switches off when I step back.
- A team that can run it. They learn the rhythm during the retainer, not from a binder at the end.
- Documentation that answers back. Where it fits, the AI layer explains how something works and helps troubleshoot when it breaks.
- The paper trail. Workflow map, runbook, automation inventory, review rules, access notes. My access is scoped and removable.
Some clients take it in-house after a few handover sessions. Others keep the retainer because the iteration keeps paying. Both are fine outcomes.
What the layer does.
Not a new toolset. AI and automation wired into the one you already have, doing the coordination work that used to land on you.
Decision queue
One place where the flood becomes a shortlist. You see decisions, not messages.
Reporting rhythm
Reports assemble themselves on a schedule: what moved, what stalled, what got flagged.
Review gates
AI stops for human sign-off where judgment, client data, or people data is involved. Some workflows only get summarized or routed, never processed. Those boundaries are explicit.
Documentation that explains itself
Where it fits, the layer can answer how something works and help troubleshoot when it breaks.
The exact stack depends on your tools. A typical install includes a workflow map, automation inventory, decision queue, reporting rhythm, AI review rules, and handover documentation.
Clockwork works with the stack you already have.
Project & delivery
ClickUpLinearJiraAsanaGitHub IssuesOpenProject
Docs & knowledge
NotionConfluenceGoogle DriveMicrosoft 365
Comms
SlackMicrosoft Teamsemail
People operations
HR systemstime trackingbilling/invoicingcontractor coordinationonboarding
People workflows are handled with explicit boundaries. Not every people-related workflow gets automated, and some only get summarized or routed, not processed by AI.
Data & reporting
AirtableGoogle SheetsExcelPostgreSQLdashboards
Automation & AI
native integrationsAPIsMakeZapiern8n
Model-agnostic by design: Claude, GPT models, open models. Whatever fits the job and the budget. More on the AI layer
Open models run via OpenRouter, OpenCode, Hermes, Odysseus, and similar alternatives. Agent and memory setups are built to avoid vendor lock-in. Token spend is something I actively watch and trim, not a solved problem, but not ignored either.
Data, access, and review gates.
access
Clockwork runs inside your accounts. Access is scoped, documented, and removable. AI steps get review gates where judgment, client data, or people data is involved. Some workflows shouldn't be automated at all. That decision belongs in the plan, before anything is wired.
governance
For work involving personal data, processing roles, retention, subprocessors, and handover get agreed before implementation. The point isn't to make AI invisible. It's to make the workflow legible enough to govern.
Pricing.
01 · Audit
From €500.
02 · Install
For small teams, typically €1,500 to €5,000, depending on how many tools get wired in. Bigger teams are a different conversation with different numbers. Either way, you get the exact figure in the written plan, before any commitment.
03 · Retainer
From €500/month. Scope and rhythm are set in the plan. Cancel monthly.
Every figure lands in the written plan before any commitment.
Start with a systems call.
Forty-five minutes on where the coordination load actually sits. If there's a fit, the next step is a scoped audit, and the audit produces the written plan: scope, price, and handover, in writing, before anything gets installed.
Book a systems call or email: