What an Agency Operations System Is (and Isn't)
An agency operations system is the set of processes, records, and rules that carry client work from first contact to final payment. It answers, at any moment and without a meeting: which deals are live, what's being delivered this week, where every hour went, what's been invoiced, and whether the agency made money doing it.
It is not a tool stack. Most agencies already own more software than they need—a CRM here, a project tracker there, a timer app, an invoicing tool, a reporting spreadsheet. What they don't own is the connective tissue: agreed rules for how a record moves between those stages, who owns each handoff, and what "done" means at every step. Tools without rules produce the familiar symptoms—deals that close in the CRM but never become projects, hours tracked against nothing, invoices reconstructed from memory at month-end.
The one-sentence test
You have an operations system when a new hire can answer "what happens after we win a deal?" by reading a document, not by asking three people. Everything in this guide works backwards from that test.
The framework below splits agency operations into five layers, ordered the way money flows: pipeline → delivery → time → billing → reporting. Each layer has a job, a small set of records it must maintain, and a handoff to the next layer. Weakness in any one layer caps the value of the others—perfect time tracking is worthless if invoices don't use it, and a beautiful pipeline is worthless if won deals land in an inbox instead of a project plan.
This guide covers building the system. For what it looks like to run a single deal through all of it end to end, see the companion guide on the lead to invoice workflow.
The Five Layers of Agency Operations
Layer 1: Pipeline
Everything before the work: outreach, leads, qualification, proposals, and closing. The pipeline layer's job is to make future revenue visible and comparable—every opportunity has a stage, a value, a next action, and an outcome when it closes.
Core records: contacts, outreach touches, deals with stages, proposals, win/loss outcomes with reasons.
Handoff it must produce: a won deal with enough context (scope, budget, contacts, promises made) that delivery can start without re-interviewing the salesperson. Structured outreach tracking is what keeps this layer honest at the top of the funnel.
Layer 2: Delivery
Projects, tasks, milestones, and client communication. The delivery layer's job is to turn the promise sold in Layer 1 into shipped work—visibly, so both the team and the client can see status without asking.
Core records: projects linked to clients and deals, tasks with owners and estimates, milestones, a scope document, client-visible status.
Handoff it must produce: tasks specific enough that time tracked against them means something. "Website" as a task makes Layer 3 useless; "homepage design, est. 8h" makes it powerful.
Layer 3: Time
Where the agency's largest cost—labor—gets measured. The time layer's job is to attribute hours to tasks and projects, split billable from non-billable, and surface budget burn while there's still time to act on it.
Core records: time entries linked to tasks, billable flags, estimates vs actuals, per-person utilization.
Handoff it must produce: approved, client-ready hours that billing can invoice without a reconciliation meeting. If someone spends a day each month "cleaning up the timesheets," this handoff is broken.
Layer 4: Billing
Invoices, payment terms, reminders, and collections. The billing layer's job is to convert delivered work into cash quickly and without disputes—which mostly means invoicing from records, not from memory.
Core records: invoices linked to projects and time entries, payment status, reminder schedule, per-client terms.
Handoff it must produce: payment data granular enough that reporting can compute what each project actually earned—not just what was billed.
Layer 5: Reporting
The feedback loop. The reporting layer's job is to answer the management questions—which clients are profitable, whose utilization is slipping, how much capacity is left this month—from data the other four layers already produced as a side effect of doing the work.
Core records: delivery margin per project, utilization per person, pipeline conversion rates, revenue vs capacity.
Handoff it must produce: decisions. A report nobody changes behavior over is decoration. Reporting closes the loop back into Layer 1 (what to sell more of) and Layer 2 (what to estimate differently).
Notice the dependency direction: each layer consumes the previous layer's output. That's why "we bought a reporting dashboard" so rarely fixes anything—reporting is the last layer, and it can only be as good as the pipeline, delivery, time, and billing data feeding it.
Maturity Self-Assessment (Copy This)
Score each layer honestly against the three levels below. Level 1 is ad hoc (it lives in someone's head or inbox), Level 2 is tracked (it's written down somewhere, updated inconsistently), Level 3 is systematized (one place, updated as a side effect of doing the work, with rules everyone follows). Copy the table into a doc and score yourselves as a team—disagreement about a score is itself a finding.
| Layer | Level 1 — Ad hoc | Level 2 — Tracked | Level 3 — Systematized |
|---|---|---|---|
| Pipeline | Leads live in inboxes and DMs; "pipeline" is what the founder remembers | Deals listed in a spreadsheet or CRM, stages updated when someone remembers | Every opportunity has a stage, value, next action, and a recorded outcome with reason; reviewed weekly |
| Delivery | Work assigned in chat; status known by asking; scope lives in an email thread | Projects have task boards, but estimates, scope docs, and client updates are hit-or-miss | Every project created from the won deal, tasks estimated, milestones dated, client sees status without asking |
| Time | Hours reconstructed at invoice time, or not tracked at all | Timers or timesheets exist, but entries land on projects, not tasks, and billable split is fuzzy | Time logged to tasks daily, billable flagged, actuals vs estimates visible mid-project |
| Billing | Invoices written manually at month-end from memory; follow-up when cash gets tight | Invoicing tool in place, but line items are retyped from other tools and reminders are manual | Invoices generated from tracked time and milestones within days of completion; reminders automatic |
| Reporting | Profitability judged by bank balance and gut feel | Quarterly spreadsheet exercise merging exports from several tools; usually stale | Margin per project, utilization per person, and pipeline conversion available live, from operational data |
Two readings matter. First, your lowest-scoring layer—that's your bottleneck, and it's where the next month of process work should go. Second, the gaps between adjacent layers: a Level 3 time layer feeding a Level 1 billing layer means you're doing the hard part (tracking) and forfeiting the payoff (fast, defensible invoices).
Most agencies that self-assess land at Level 2 almost everywhere: tools exist, rules don't. That's actually good news—moving from Level 2 to Level 3 is mostly deciding on handoff rules, not buying anything.
Decision Rules Per Layer
A layer reaches Level 3 when its handoffs are governed by rules, not judgment calls. These are the default rules to adopt—change the numbers to fit your business, but write them down and enforce them.
Pipeline rules
- • No opportunity exists outside the pipeline. If it's in a DM, it gets logged the same day or it doesn't count.
- • Every open deal has a next action with a date. A deal without one is stalled by definition.
- • Every closed deal gets an outcome and a reason. "Went quiet" after your full follow-up sequence plus 14 days is an outcome—record it and move on.
Delivery rules
- • A won deal becomes a project within 48 hours, with the deal's scope, budget, and contacts attached—never a fresh blank project.
- • No task without an estimate; no project without a milestone date. Estimates can be wrong, but they must exist, or Layer 3 has nothing to burn against.
- • The client gets status on a fixed cadence (weekly, in writing) rather than on request.
Time rules
- • Time is logged to tasks, daily. Weekly reconstruction is how 20% of hours vanish.
- • Every entry is flagged billable or not at entry time, not at invoice time.
- • When a task hits 80% of its estimate unfinished, it gets flagged—that's a scope conversation this week, not a surprise at delivery.
Billing rules
- • Invoice within 48 hours of a milestone or period end. Month-end batching is an interest-free loan to your clients.
- • Invoice line items come from tracked time and completed milestones—never retyped.
- • Reminders fire automatically at day 7, 14, and 21 past due. Chasing money should not require remembering to chase money.
Reporting rules
- • Review margin per project monthly and utilization per person weekly—against targets, not vibes.
- • Every report review ends with at most three decisions. A dashboard that produces no decisions gets deleted.
- • Post-mortem every project against its estimate; feed the delta back into how the next similar project is priced.
The utilization and capacity targets these rules reference shouldn't be guesses. Use the billable utilization calculator to set a realistic per-person target, and the agency capacity calculator to know how many delivery hours your team actually has to sell before the pipeline over-commits them.
Worked Example: A 10-Person Agency Leaves Tool Sprawl
Take a 10-person digital agency—two founders (one selling, one running delivery), six designers and developers, a PM, and a part-time bookkeeper. Roughly $1.2M in annual revenue across a dozen active clients. Their stack: Pipedrive for deals, Trello for projects, Toggl for time, QuickBooks for invoices, and a "Profitability v7 FINAL" spreadsheet nobody fully trusts.
On the maturity table they score: Pipeline 2, Delivery 2, Time 2, Billing 2, Reporting 1. Every layer has a tool; no layer has rules; nothing is connected. The visible symptoms:
- •Won deals reach the PM as a Slack message: "closed Meridian, kickoff next week?" Scope and budget stay in the selling founder's email. Two projects last quarter started without anyone confirming the revision count that was promised.
- •Toggl entries say "Meridian — dev work." Nobody can say which tasks blew the budget, only that the project did.
- •The bookkeeper spends the first three days of each month cross-referencing Toggl exports against QuickBooks. Invoices go out on the 8th, on average.
- •The founders believe their retainers are the profitable half of the business. Nobody can prove it.
They fix it over one quarter, one handoff at a time—consolidating the five layers into one platform, in this order:
Month 1: time → billing
The fastest cash win. Tasks get estimates, time gets logged to tasks with billable flags, and invoices generate from tracked time. Invoices now go out within 2 days of period end instead of 8+. The bookkeeper's three-day reconciliation ritual disappears; average payment arrives 11 days sooner.
Month 2: pipeline → delivery
Deals move into the same system as projects. The 48-hour rule takes effect: winning a deal converts it—scope, budget, contacts attached—into a project the PM owns. The kickoff scramble ends, and the selling founder stops being a single point of failure for project context.
Month 3: reporting
With four layers feeding one database, reporting stops being a spreadsheet project. First real margin report produces two surprises: the "flagship" $180k/yr client runs at an 18% delivery margin (unbilled revision rounds), while the boring $60k/yr maintenance retainer runs at 68%. They renegotiate the flagship's revision policy at renewal and raise its rate 15%—a conversation they could never have opened without the numbers.
End state after one quarter: Pipeline 3, Delivery 3, Time 3, Billing 3, Reporting 2 (they still owe themselves post-mortems). Same ten people, no new hires—roughly a day per week of combined admin time returned, invoices out five times faster, and pricing decisions made on data.
The margin math this example turns on—delivery margin, effective rate, estimate vs actual—is covered in depth in the complete guide to agency profitability.
Build Order: Which Layer First
Don't fix all five layers at once—you'll change too much process simultaneously and the team will revert. The sequencing rule from the worked example generalizes:
- 1.Start where cash leaks fastest: the time → billing handoff. It has the shortest payback (days-faster invoices, hours recovered) and the least behavior change—people already track time, they just track it uselessly.
- 2.Then fix pipeline → delivery. This is the handoff that protects margins before a project starts. It mostly requires discipline from whoever sells, which is easier once the team has seen the first fix pay off.
- 3.Reporting last, always. It's tempting first—dashboards are fun—but reporting on ad hoc data just automates fiction. Earn it with the other four layers.
One quarter is a realistic horizon for a team under 20 people. Signs it's working, in order of appearance: invoices go out days after work completes, kickoffs stop needing the salesperson in the room, and—the endgame—a margin report someone actually acts on.
Running All Five Layers in One System
Everything above is tool-agnostic—you can reach Level 3 with five disconnected tools and enough discipline. But notice where the maturity table's Level 3 column keeps landing: "created from the won deal," "generated from tracked time," "from operational data." Every hard part is a handoff between tools, and handoffs between tools are exactly what integrations, exports, and copy-paste rituals keep breaking.
That's the case for running the five layers on one platform with one client record. Corcava was built as exactly that: deals convert to projects with context intact, time logs against tasks, invoices generate from tracked time, and margin reports come from the same database the work happened in. The rules in this guide stop being discipline and become defaults. See how agencies set this up on the Corcava for agencies page.
Build your operations system in Corcava
Pipeline, delivery, time, billing, and reporting on one client record—set up the five layers from this guide in an afternoon.
No credit card required