How your AI plan is worked out
An AI tool writes the words in your report. Costiva’s own code does every bit of ranking and math, using the published rules on this page. The AI never sees or writes a money figure.
1. From your answers to a list of ideas
Your answers go to the AI inside a clearly marked box with a rule that nothing inside it is an instruction. It suggests 15 to 30 realistic ideas across three areas of your business (customers, running the work, and the back office). For each idea it picks one level on each of five fixed scales — nothing else. We check every answer against a strict format and reject anything that doesn’t fit.
2. How ideas are scored
Each level is worth 1 to 3 points (3 is best for you). Points are weighted and scaled to a score out of 100:
| Factor | Weight | Levels (points) |
|---|---|---|
| How much it would help | 35% | A little (1) · Noticeably (2) · A lot (3) |
| How much work it takes to set up | 20% | An afternoon or less (3) · A few days (2) · Weeks, maybe outside help (1) |
| What could go wrong | 15% | Little harm if it slips (3) · Needs checking (2) · Mistakes could really hurt (1) |
| Is the information it needs ready? | 15% | Yes, already in one place (3) · Some tidying first (2) · Not yet (1) |
| How soon you would see a difference | 15% | Within days (3) · Within weeks (2) · After months (1) |
If an idea shrinks a job you rated more than 3 out of 5 for “how annoying”, it gets 2.5 extra points for each point above 3. Scores never go above 100.
3. Tiers
- Quick win: Low effort, and you would see a difference within days or weeks.
- Bigger bet: Takes weeks of work or months before it pays off.
- Project: Everything in between.
4. How your top 5 are picked
- Anything with a high risk is left out unless a person checks every result.
- The best-scoring quick wins go in first — at least 2 of them when there are enough.
- The rest are filled by score.
- No more than 2 come from the same area of the business, so your plan isn’t lopsided.
If these rules leave fewer than 5 ideas, your list is simply shorter — we never bend the rules to fill it.
5. How savings ranges are worked out
Only ideas linked to one of your time-sinks get a savings range. Every number is a range, from the low end of each band to the high end:
- Hours a week = the hours band you picked for that job × the share of it the idea could take off your plate.
- Hours a year = hours a week × 48 working weeks.
- Money a year = hours a year × the hourly cost band you picked. If you chose “Not sure”, we assume $25 to $40 an hour and say so.
- When two ideas shrink the same job, only the bigger one counts in your total — the same hours can’t be saved twice.
- Hours are rounded to the nearest half hour; money to the nearest 50, or 500 above 10,000, so the ranges don’t look more precise than they are.
Hours bands
- Less than 2 hours: 0.5–2 h
- 2 to 5 hours: 2–5 h
- 5 to 10 hours: 5–10 h
- 10 to 20 hours: 10–20 h
- More than 20 hours: 20–30 h
Share taken off your plate
- A little (5–15%)
- Some (15–30%)
- A lot (30–50%)
- Most of it (50%+)
Hourly cost bands
- Under $25 an hour (we use 15–25)
- $25 to $40 an hour (we use 25–40)
- $40 to $60 an hour (we use 40–60)
- $60 to $90 an hour (we use 60–90)
- More than $90 an hour (we use 90–130)
6. What these estimates are not
- They are not a promise. They come from your own rough answers, and real results depend on how you set things up and how your team uses them.
- They don’t include what new tools cost to buy or the time it takes to learn them.
- Time freed up only becomes money if it goes to other paid work, or saves overtime or a hire.
7. Tracking results
When you try an idea, you record a “before” number and check in now and then. We show what changed and every reason to be careful reading it — for example fewer than 3 check-ins, a “before” number recorded after you started, or a seasonal business. We show what changed, never claim why.
Scoring version 2026.10-1. Every report records the version it was scored with.
Back to New to AI