ASC 606 Rev Rec: Software vs Excel vs AI Agents

By Jude Lee · · Comparison

Two accountants reviewing a customer contract and a deferred revenue schedule on screen in a firm office

Three realistic ways firms handle revenue recognition today

If you serve SaaS, subscription, agency, construction, or multi-element product clients, ASC 606 shows up as recurring monthly work: abstract the contract, build or update the schedule, post the journal entry, reconcile deferred revenue, and support the disclosure at year end.

The three operating models in the wild:

  1. Spreadsheets. A waterfall tab per client, monthly manual updates, formulas that only one person understands. Cheap, flexible, fragile.
  2. Revenue recognition software. Either a module inside the ERP (NetSuite’s advanced revenue management, for example) or a standalone subscription-billing/rev-rec engine. Deterministic, auditable, opinionated — and priced for mid-market clients, not a two-person e-commerce shop.
  3. AI assistants and agents layered on top of either. Not a replacement for the schedule engine. A worker that handles the reading, checking, and drafting around it.

The interesting question for an operations lead isn’t “which one.” It’s which half of the workflow each tool should own.

What the five-step model tells you about where automation belongs

Per the FASB, ASC 606 (Revenue from Contracts with Customers) applies a five-step model: identify the contract, identify the performance obligations, determine the transaction price, allocate that price to the performance obligations, and recognize revenue as each obligation is satisfied. The AICPA publishes revenue recognition guidance and industry-specific interpretations; confirm application questions against those primary sources.

Read that model as an automation map:

What AI agents actually do well here

An “agent” in this context means an AI assistant that can take multi-step actions against your systems — open a document, extract fields, write to a schedule, query the GL, post a draft for review — not a chatbot you paste text into. Concretely, for a rev-rec engagement:

Let the agent read the contract and check the math. Never let it be the math.

Where agents break

They break in predictable places, and you should design for them:

Choosing between a schedule engine and an agent-assisted spreadsheet

Dedicated rev rec software
Best when: high contract volume, recurring billing systems to integrate, multi-element arrangements, audited financials, or a client planning to raise capital. You get deterministic schedules, versioning, and an audit trail out of the box. Where it fails: implementations run long and need clean billing data you may not have; rigid configuration tends to break on non-standard contracts, pushing odd deals back into side spreadsheets anyway; and migrating off — or onto — the platform when a client outgrows it carries real cost and re-testing.
Spreadsheet model + AI agent
Best when: a handful of contracts per client, mostly ratable subscriptions, no audit, and a firm that already runs a clean workpaper template. The agent does the reading and the tie-out; the spreadsheet does the math. Where it fails: you own the controls, the version history, and the review discipline entirely, and the model degrades quietly as volume grows.

There’s a third honest answer: for a single-product SaaS client billing annually in advance, a deferred revenue entry in QuickBooks Online plus a one-tab schedule is sufficient. Not every workflow deserves AI — simple, stable, deterministic work is usually best left alone.

Wiring the agent to the ledger without handing over the keys

One practical mechanism as of 2026 is MCP — the Model Context Protocol, an open standard for giving an AI assistant governed access to specific tools and data. Instead of copying trial balances into a chat window, you connect the assistant to the systems you already run. We walk through the setup in connecting an AI assistant to QuickBooks or Xero via MCP. It is not the only option, and often not the first one: a vendor-native API, an iPaaS connector, or your rev-rec tool’s own built-in integration will beat a custom MCP server whenever the data you need is already exposed, the volume is low, or nobody on staff can maintain a bespoke server after the person who built it leaves.

For revenue recognition specifically, a sensible scope looks like:

  1. Read-only on the ledger

    Grant the agent read access to the GL accounts in scope — revenue, deferred revenue, contract assets, AR — and nothing else. It proposes journal entries as drafts; a person posts.

  2. Scoped document access

    Point it at one client folder of executed contracts, not the whole document management system.

  3. A custom MCP server for your schedule

    If your rev-rec model lives in a spreadsheet or internal database with no usable API, a small custom MCP server can expose two or three safe operations — get_schedule, propose_update, get_rollforward — with per-client permissions and full audit logging of every call.

  4. A review gate before anything lands

    Every output is a draft with its source cited back to a contract clause or a GL transaction. Build the gate in before you build the automation; see our take on AI review gates rather than autopilot.

Package it as a skill so it runs the same way every time

A skill is a reusable, packaged instruction set that teaches an assistant to do one job your firm’s way. For ASC 606, the skill is the contract abstract: a fixed field list, a required citation to the clause supporting each field, an explicit “UNCLEAR — escalate” value instead of a guess, and a standard output format that drops straight into your workpaper. Same idea as the skills approach to workpaper prep from a trial balance — consistency is the product, not cleverness.

Model the economics with your own numbers

Don’t trust anyone’s published savings figure, including ours. Here is a worked example with every assumption on the table — replace each number with your own.

40 × 12 min = 8 hrs
Assume 40 contracts/month at 12 minutes to abstract by hand
Illustrative assumption — time your own team
8 hrs → 3 hrs
If reviewing agent drafts takes roughly a third as long; your ratio will differ and starts worse
5 hrs × your blended rate
Gross monthly recovery, before build and maintenance time

Time your team on three real contracts: minutes to abstract, minutes to update the schedule, minutes to tie out deferred revenue. Multiply by monthly volume and your blended rate. Then subtract the build and review time honestly — reviewing an agent’s abstract is not free, and in month one it may take as long as doing it yourself. If the remaining gap is small, buy software or leave it in Excel. If the gap is large and your client contracts are reasonably uniform, the agent-assisted path is worth a pilot on one client.

Does any of this replace the accountant?

No, and the ASC 606 workflow is a clean illustration of why. The automatable parts are extraction, arithmetic, tie-out, and first-draft prose. The parts that carry professional responsibility — whether a promise is a distinct performance obligation, how to estimate SSP, whether variable consideration is constrained, whether a modification is prospective — are judgments a licensed professional signs. What changes is the ratio: less time typing contract terms into a tab, more time on the judgments and on explaining the result to the client.

That’s also the honest answer to what the largest firms do. They run enterprise revenue engines and proprietary platforms, and they staff the judgment layer heavily — a resourcing difference, not a magic one. A ten-person firm can close much of the gap by automating the reading and the tie-out while keeping the signature human.

Not sure where to start?

Get a free automation audit: we map your bookkeeping, month-end close, client onboarding, document collection, and AP/AR — and show you what's worth automating before you spend a dollar.

Get a free automation audit

Related guides