Plan
Reads the engagement, the client's prior-year return and the firm's checklist, then proposes the steps before touching anything.
AI Agents
Purpose-built agents pull the firm's own precedent into a draft memo, response or checklist. Every claim carries its source, and every draft waits for a person to approve.
Reads the engagement, the client's prior-year return and the firm's checklist, then proposes the steps before touching anything.
Works through source documents, prior-year files and workpapers step by step, splitting off subagents where the work runs in parallel.
Reads its own draft against the return and last year's review comments, rewrites what does not hold up, and marks it ready for a person.
Leaves workpapers, emails and briefs in the client's space as drafts. Nothing is filed or sent until a person approves it.
Tracks what has arrived against what was asked for, and drafts the chase when something is late. It never sends one.
Takes each of the firm's templates, fills in everything that carries from last year, and leaves blank anything that needs a decision.
Reads an uploaded return and runs six layers over it: complete, adds up, honours last year, would the IRS reject it, last year's comments, and what is new.
Builds a first draft of the return from the trial balance, the ledger and last year's filing. Every line shows where its number came from, and any line without a source is left for a person to fill.
Writes the five-minute summary a signer reads before signing. It waits until the return is far enough along to be worth reading.
From last year’s return and what has already arrived.
Line by line, with the differences named.
And show what changed since it was signed.
With the assumptions carried and flagged.
Against the positions this client actually takes.
From every comment this return has ever drawn.
A draft email per client, ready to send.
What is in, what is out, and who is waiting on whom.
No. Every agent produces a draft. Filing, sending and signing stay with a person, and Pernee never writes back into your tax software.
From the client's space: this year's documents, prior-year returns, workpapers and the review comments the team left. Every claim cites the source it came from.
Document chaser, Workpaper builder, Review checkpoint, Draft return and Signer brief. Each has one job and runs on the same review memory.
Yes. Agents follow the firm's own templates and checklists, so changing the checklist changes what the agent looks for.