What can a CPA firm calculate from Cranston’s per-return pricing?

Cranston’s public pricing lets a CPA firm answer a more useful question than “what does the AI cost?”: how many books and returns would have to move through the workflow before the bill fits the firm’s operating model?

The two billing units

Cranston’s pricing page has two public rates as observed on September 19, 2026:

• $30 per client book, per month. One connected client entity counts as one book. A client with two entities counts as two books.
• $40 per prepared return, any form. The page lists 1040, 1120, 1120-S, and 1065, and says the return is billed once after preparation.

The same page says practice management, client status, document tracking, and workflow are included at $0. It also says there are no per-form add-ons, hourly billing, per-seat charges, platform fees, or onboarding charges. Those are published product and pricing claims, not an independent audit of a firm’s eventual invoice.

The unit definition matters. “Twenty clients” is not enough to forecast the bookkeeping line until you know how many legal entities are connected. “One hundred returns” is a better input because the public price is attached to the prepared return, not to the number of staff members using the account.

A worked price model

Use this formula first:

annual bill = (books × $30 × 12) + (prepared returns × $40)

These scenarios are arithmetic examples using Cranston’s posted rates, not customer bills:

They assume a constant book count for all twelve months and one annual return season.

Example firm	Books	Returns in a season	Bookkeeping	Tax prep	Illustrative annual total
Tax-focused practice	0	100	$0	$4,000	$4,000
Small mixed practice	10	100	$3,600	$4,000	$7,600
Growing mixed practice	20	150	$7,200	$6,000	$13,200
Two-entity client base	40	150	$14,400	$6,000	$20,400

The last row is the common counting trap. It assumes 20 client relationships, each with two connected entities. Change that assumption and the bookkeeping number changes even though the number of relationships does not.

For firms above 2,000 returns in a season, Cranston says it prices the volume together. That is a point to request a quote, not a reason to extrapolate the $40 rate indefinitely.

Compare the rate with the work you would otherwise pay for

The cleanest internal comparison is marginal labor, not a whole-firm salary budget. Suppose, purely for an example, that a firm uses a fully loaded preparer cost of $45 per hour and spends 1.5 hours of marginal staff time on a routine return. That is:

$45 × 1.5 hours = $67.50 of illustrative preparation labor per return

At 150 returns, that assumption produces $10,125 of preparation labor, compared with $6,000 at Cranston’s public tax rate. The $4,125 difference is not a promised saving. It is a sensitivity result that disappears or reverses when the firm changes the hourly cost, time per return, review burden, client communication, or the mix of complex work.

Run the same model for books:

internal monthly bookkeeping cost per entity
  = marginal hours per month × fully loaded hourly cost

compare with $30 per connected entity per month

If your internal number is below $30 before review and exception handling, the product may not be a cost play for that service line. If it is above $30, the next question is whether Cranston’s workflow fits the firm’s controls. The price table cannot answer that by itself.

What the buyer is actually evaluating

Cranston says it works inside software firms already use, including QuickBooks and tax systems such as Drake, CCH Axcess, Lacerte, and UltraTax. Its public workflow says the AI prepares the work, while a reviewer approves it before anything posts, files, or reaches a client. The company’s overview describes document checks, source-linked workpapers, bookkeeping exceptions, and a client-status board.

That creates three distinct buying cases:

Firm situation	First calculation	Fit question
Tax-only	Returns × $40	Is reviewer time still lower after checking the prepared work?
Books-only	Books × $30 × 12	Are entity counts and monthly exceptions stable enough for a recurring workflow?
Mixed practice	Both lines together	Does keeping books and returns in one client workflow reduce handoff work, or merely add another system to supervise?

The decision tree

1. Do you want to price a pilot? Count actual connected entities and expected prepared returns. Do not use client count as a proxy for books.
2. Is your bottleneck routine preparation or reviewer capacity? If reviewers are the constraint, a lower preparation rate may not change throughput without a clear exception workflow.
3. Does your return mix exceed 2,000 in a season? Ask for volume pricing instead of using the public $40 figure.
4. Are you buying a calculator or a workflow? Include document collection, review, status tracking, and client follow-up in the comparison because Cranston presents those as part of the operating system, not separate line items.

Founder implication

Cranston’s interesting pricing decision is that it makes the bill legible before a demo: count the units, multiply by the published rates, then challenge the assumptions. The founder lesson is equally concrete. If you sell AI into a professional workflow, price around the unit the buyer already manages, expose the exceptions, and make the human approval boundary explicit. A per-return number is useful only when the buyer can tell which work remains after the agent prepares the draft.

Sources — snapshots observed 2026-09-19

• Cranston AI on Y Combinator — observed 2026-09-19.
• Cranston homepage — observed 2026-09-19.
• Cranston pricing — observed 2026-09-19.