To get an AI tool to explain a Schedule K-1 in plain English, paste this prompt: "You are a tax professional writing to a non-expert client. Explain the following Schedule K-1 in plain language. Cover what the form is, what each populated box means for the client, whether it increases or decreases their taxable income, and what (if anything) they need to do. Use short sentences and no jargon. Do NOT give filing advice or dollar-figure tax projections. Then paste the K-1 data below." Below is a longer version, why each part matters, and how to keep the answer safe to send.
The full prompt to copy
``` Role: You are an experienced U.S. tax preparer writing a short note to a client who has no accounting background.
Task: Explain the attached Schedule K-1 (Form [1065 / 1120-S / 1041]) in plain English. Keep it under 300 words.
Cover, in this order:
- What a K-1 is and why they received one.
- A line-by-line summary of ONLY the boxes that have amounts, in simple
terms (e.g., "Box 1 is your share of the business's ordinary income").
- Whether each item generally increases or decreases taxable income.
- What the client needs to do next (usually: nothing, we'll enter it).
Rules:
- No jargon. Define any term you must use.
- Do NOT calculate their tax owed or give a refund estimate.
- Do NOT give investment or legal advice.
- If a box is ambiguous, say "we'll confirm this on our end" instead
of guessing.
- Friendly, reassuring tone. Short paragraphs.
Here is the K-1 data: [paste box numbers and amounts] ```
Why the prompt is structured this way
Each instruction closes a common failure mode.
- "Role" line sets tone and reading level. Without it, models default to textbook definitions that clients skim past.
- "Only boxes with amounts" stops the model from explaining all 20-plus possible boxes when the client's K-1 has three.
- "Increases or decreases taxable income" is what clients actually want to know. It reframes abstract codes into impact.
- "Do not calculate tax owed" prevents the model from inventing a number the client will quote back to you.
- "We'll confirm on our end" gives the AI a safe fallback instead of hallucinating on codes it misreads.
Fill in the form type
The three K-1 flavors read differently, so tell the model which one it is:
| Form | Entity | Common client confusion |
|---|---|---|
| K-1 (1065) | Partnership / LLC | "Why do I owe tax on income I never received?" (undistributed pass-through) |
| K-1 (1120-S) | S corporation | Difference between distributions and W-2 wages |
| K-1 (1041) | Estate / trust | Whether the beneficiary or the trust pays the tax |
A generic prompt often mislabels boxes across these forms. Specifying the form cuts that error sharply.
A quality-control checklist before you send
AI-generated explanations are a first draft, not client-ready copy. Review for:
- Box mapping. Confirm the model matched each amount to the correct box description for that specific form and year.
- Direction of impact. Verify "increases" vs. "decreases" claims, especially for items like Section 179 deductions, self-employment earnings, and separately stated items.
- No tax figures. Delete any sentence that estimates tax, penalty, or refund.
- State issues. The prompt covers federal only. Add a line if the client has a multistate K-1.
- Passive vs. active. The model won't know the client's participation level. Strip any statement about passive loss limits unless you've confirmed it.
Handling data privacy
Never paste a client's name, SSN, or EIN into a public AI tool. Replace identifiers with placeholders or use only the box numbers and dollar amounts. If your firm uses a tool with a business-tier data agreement that excludes your inputs from training, document that in your written information security plan. When in doubt, strip everything but the numbers.
A shorter prompt for a quick client email
When you just need two sentences for a portal message:
`` In plain English, write 2-3 sentences a client can understand that summarize what this K-1 means for their taxes and what they should do. No jargon, no tax calculations. Data: [paste boxes with amounts] ``
Where AI helps and where it doesn't
Use the model to draft language and translate codes into readable prose. Keep the judgment calls, basis tracking, at-risk and passive limits, and the actual return entry with you. The prompt saves the ten minutes you'd spend rewriting the same explanation for every partner in a fund, not the analysis itself.
If staying current on practical AI workflows for tax and advisory work is worth five minutes a day, DayLift sends a short daily briefing built for CPAs and EAs.