Can AI help me pick my work benefits at open enrollment?
Partly. Paste your plan options and how you used care last year, and it lays out the real yearly cost of each plan side by side. It can't see your full plan documents or know next year's health needs, so HR and the plan papers have the final word.
01Why half-bot
- Comparing premiums, deductibles and out-of-pocket maximums is math AI does well.
- It turns a confusing benefits guide into a side-by-side yearly cost.
- It only knows what you paste and can't predict your health, so it can't make the choice for you.
02The prompt
Replace everything in [BRACKETS] with your own details, then paste it into your AI app.
You are a benefits counselor who explains open enrollment choices to employees in plain English. My plan options: [PASTE EACH PLAN'S PAYCHECK COST, DEDUCTIBLE, OUT-OF-POCKET MAX, COPAYS AND ANY HSA MONEY FROM MY EMPLOYER] Who I'm covering: [ME, ME + SPOUSE, FAMILY] Care last year: [DOCTOR VISITS, PRESCRIPTIONS, SURGERY OR BABY PLANNED, OR ROUGH TOTAL SPENT] Doctors I want to keep: [NAMES OR NONE] Your task: compare my plan options. Rules: - MUST use only the numbers I pasted. NEVER invent premiums, deductibles, prescription prices or rules for a plan. If a cost isn't in what I gave you, write "not given" and leave it out of the math. - MUST show a table with each plan's yearly premium, deductible, out-of-pocket max and estimated yearly total for a light year and a heavy year, with a dollar amount in every cell. - MUST work out each total this way and NO other way: light year = yearly premium + what I spent on care last year (capped at the out-of-pocket max) - any employer HSA money; heavy year = yearly premium + out-of-pocket max - any employer HSA money. - MUST show the math for each total on its own line, like this: "Plan A light year: $X + $Y - $Z = $T". - MUST tell me to check that my doctors are in network on the insurer's website, because you can't check that. - MUST start your answer with this exact line: "Not financial, medical or legal advice." - NEVER tell me which plan to pick; say which one costs least in each scenario. - If anything is missing, ask me up to 3 short questions first. Give me, in this order: the table, the math lines, then 3 questions to ask HR before the enrollment deadline.
03Steps
- Open your benefits guide and copy each plan's costs.
- Fill in the brackets and paste the prompt into your AI app.
- Check your doctors on the insurer's site and ask HR anything unclear before the deadline.
04What can go wrong
- This isn't financial or medical advice; the plan documents and HR have the final word.
- A plan that's cheapest in a light year can cost far more in a bad year, so look at both.
05What to use
06Our test run
Sample input we used: Covering me + spouse. Plan A (PPO): $210 per month from paycheck, $1,000 deductible, $5,000 out-of-pocket max, $30 copays. Plan B (HSA high-deductible): $95 per month, $3,200 deductible, $7,000 out-of-pocket max, employer puts $1,000 a year in the HSA. Last year: about 6 doctor visits and 2 monthly generic prescriptions, roughly $1,400 total spent. Want to keep Dr. Patel (family doctor).
- Uses only the plan numbers given and the formulas in the prompt, and invents no premiums, prices or plan rules
- Shows a table with yearly premium, deductible, out-of-pocket max and light-year and heavy-year totals for each plan
- Shows the math for each of the four yearly totals on its own line using the prompt's formulas
- Tells the user to check that their doctor is in network on the insurer's website
- Does not tell the user which plan to pick and says it is not professional advice
What came back (excerpt):
Plan B light year: $1,140 + $1,400 - $1,000 = $1,540
Majority of 3 runs. Revised twice before publishing: earlier versions invented prescription prices, left totals blank or skipped the disclaimer, so the prompt now spells out the formulas and the opening line.
One run with an invented sample, checked against the prompt's own rules by a second AI model (word limits counted by code). Your results can differ; how we test.