A profitability agent reads the week's Harvest entries, joins them to the Asana plan, and drafts a budget brief. The brief lands in Dock. The PM approves, adjusts, or rejects each line. Harvest keeps the timesheet. Asana keeps the task. Dock keeps the interpretation and the signed decision. This is the consulting and agencies pattern applied to weekly margin review.
Harvest and Asana stay the system of record for the raw data. Dock is the system of record for what the AGENT INTERPRETS. Each Dock row carries a pointer back to the platform record, agent identity, decision, reviewer, and timestamp. The agent re-fetches platform data via fresh API reads when it needs current state.
The surface: Weekly budget briefs
| Project | Hours logged (Harvest) | Budget burn | Agent brief | PM decision | Reviewer | Timestamp |
|---|---|---|---|---|---|---|
| Acme rebrand (H-4412) | 47.5 of 80 | 59% at week 4 of 8 | On track. Two design rounds remaining within budget. | Approved | govind@ | 2026-05-30 09:14 |
| Northwind site (H-4418) | 112 of 120 | 93% at week 6 of 10 | Over-pace. Recommend scope conversation before week 7. | Escalated to client | govind@ | 2026-05-30 09:18 |
| Mercer audit (H-4421) | 18 of 40 | 45% at week 2 of 4 | Discovery underran. Reallocate 8 hours to recommendations phase. | Approved with note | priya@ | 2026-05-30 09:22 |
Each row links to the Harvest project, the Asana plan, the agent that drafted the brief, and the human who signed off. None of the hours live in Dock. The interpretation does.
One workflow: Monday margin review
Monday 08:00. The Harvest agent pulls the previous week's entries via fresh API read, joins each to the Asana project, computes burn, and writes a one-line brief per active project. Briefs land in the surface as draft rows.
08:45. The PM opens Dock. Each draft row shows the agent's reasoning, the Harvest link, and related Asana milestones. The PM approves on-pace projects in bulk, escalates Northwind to the client, and adds a reallocation note to Mercer.
09:30. Approved decisions trigger downstream actions. The finance agent updates the forecast model. The design lead's surface reflects the reallocated Mercer hours. The research pipeline sees the Northwind flag.
Three decisions in 45 minutes. The agent did the joining. The platforms kept their data.
Why this matters
Time tracking answers "where did the hours go." It does not answer "what should we do about it." That second question is where margin is won or lost, and it needs a name attached. When the agent drafts, the agent identity is on the row. When the PM signs, their name is on the row. Six months later, when a client disputes a scope change, the trail shows who decided what.
Harvest reports "70,000+ customers" who have tracked "4.5 billion hours" and invoiced "$64+ billion" through the platform (Harvest customers). Harvest notes its budget tools let teams "see which projects earn and get alerted before it's too late" (Harvest). The signal exists. What agencies lack is a place to record interpretation with attribution.
Agent keys rotate. PMs rotate. Clients audit. The identity lifecycle keeps every past decision traceable to the principal that made it.
Start with one project
Pick one active project. Connect Harvest and Asana to Dock. Let the agent draft this Friday's brief. Approve, edit, or reject.
FAQ
Do hours move out of Harvest? No. Harvest stays the timesheet of record. Dock reads hours via API and writes interpretation rows. The agent re-fetches when the PM opens the brief.
What if the PM disagrees with the brief? The PM edits or rejects the row. Both actions are logged with name and timestamp. The original draft stays in row history.
How does this connect to invoicing? Approved briefs flag projects that need a scope conversation before the next invoice. Invoicing stays in Harvest. The escalate-or-absorb decision lives in Dock.
Can the agent approve briefs autonomously? No. The agent drafts. A named human signs. This is the same pattern across every Dock surface, and it is why the audit trail holds up.
