The AI Recommendation in Rotor Rate: How the Second Opinion Is Built

# Article title The AI Recommendation in Rotor Rate: How the Second Opinion Is Built

# Article body (markdown)

Rotor Rate's calculator gives you a number anchored to your rates, your drive, your overhead. That's the fair-bid rate. Right below it sits the AI Bid Recommendation card — a deliberately independent second opinion that asks a different question: what does the market actually pay for a job like this?

This post is a tour of how that recommendation is built — what the model sees, what it doesn't, what the audit-trail signals mean, and how to read the verdict.

Hi, I'm Bill Ferguson, Part 107 pilot and Rotor Rate's creator and I'm going to walk you through how AI, everyones favorite topic these days, can actually help arm you with some pretty valuable information regarding how much you should charge, or accept for a job. Think of this tool as an unbiased, second opinion.

Why a second opinion exists in the first place

Your calculator math is internally consistent — it knows your costs, so it'll never recommend a number that loses you money. That's a strength, but it's also a blind spot. If your hourly rate is set conservatively, your fair-bid will quietly track that conservatism into every quote, and you'll never know if you're leaving 15–25% on the table (note: the paid version of Rotor Rate remedies this by tracking your jobs and recommends price increases when you hit certain gates - it actually helps you stay profitable instead of blindly following along).

The AI Bid Recommendation exists to challenge that. It's explicitly told to price the market, not to mirror your fair-bid. It can — and often does — land above your number. Sometimes it lands below. Both are useful signals.

What the model actually sees

When you click "Generate recommendation," Rotor Rate packages up a structured snapshot and sends it to the Lovable AI Gateway (which routes to a Gemini reasoning model). The payload includes:

Job context

• Industry and service type (real estate, inspection, mapping, etc.)
• Location (city, region)
• Pricing situation — open-bid platform, flat-rate take-it-or-leave-it offer, or a direct-to-client simple-project quote (each behaves differently)

Job scope

• On-site work: either hours × your fly rate, or acres × your per-acre rate
• Drive: hours and miles (supporting both one-way and round-trip inputs)
• Out of Pocket Expenses (OoPE): (out-of-pocket pass-through expenses like tolls, parking, meals, etc.)
• Scope Add-ons: specialized equipment utilized and post-processing/deliverable time requirements

Industry benchmark (when present)

• The low / average / high from the Benchmark card
• The short justification + source notes
• Explicitly labeled as supporting context only

Your cost structure (clearly labeled "reference only, NOT a target")

• Your fly rate, travel rate, gas price, MPG
• Your payment processor fees (if you pass on the charge to your invoiced client)
• The three deterministic numbers your calculator already produced:

- Break-even floor - Cost-of-doing-business (CODB) - Your fair-bid/rate

What you're currently considering

• Whatever number is in the bid / offer field at that moment (or "not yet entered")

That's it. The model never sees your personal info, your client's name, your saved jobs, or your subscription details. Every number it sees is one you can already see on the screen.

What the model is told to do

The prompt is explicit about this — I'll paraphrase the relevant parts:

To the AI: You are an INDEPENDENT pricing analyst. You are NOT the operator's calculator. The operator's rates are useful context for understanding their cost structure, but they are NOT a target — your number can land above OR below their fair-bid, and should reflect what the market actually pays for this scope.

The only hard constraint: it can never recommend a price below your break-even floor. That's enforced both in the prompt and in the server code that hard-clamps the response. If the model returns something below break-even, the server overrides it before you ever see it.

What you get back

Five things:

1. A recommended price — a single dollar figure.
2. A negotiating range (low / high) around that price.
3. A verdict comparing your current bid to the recommendation:

- Below — your bid is under the low end of the recommended range. - Fair — your bid is inside the range. - Above — your bid is over the high end. - No-bid — you haven't entered a bid yet, so no comparison.

1. A rationale — two to three plain sentences explaining the recommendation and noting explicitly when/why it diverges from your fair-bid.
2. An audit trail of 2–5 market signals — the actual reasons the number is what it is.

How to read the signals (the part most pilots skip LOL)

Each signal has three pieces:

• Label — short, like "Regional benchmark," "Scope complexity," "Drive-time burden," "Pricing situation," "Deliverable expectations."
• Detail — one sentence explaining what the signal said.
• Impact — `raises`, `lowers`, or `neutral` relative to your fair-bid.

This is the audit trail. When the AI says "I'd quote $475 instead of your $400," the signals tell you why. Maybe regional benchmark + scope complexity both raised it, while drive-time was neutral. That's actionable — you can verify each signal against your knowledge of the market.

If a signal feels wrong (e.g. it cites a benchmark that you know doesn't apply to your niche), discount the recommendation accordingly. The model is a junior analyst, not an oracle. The signals are how you grade its work.

The divergence summary

Below the rationale you'll see a divergence summary — one to two sentences explaining specifically why the AI's number differs from your fair-bid (or why it agrees with it, when they line up).

This is where the value lives. "Your math is consistent for your cost structure, but regional benchmarks suggest the market typically pays 18% more for inspection scopes of this size — much of the gap is room you're leaving on the table" is a much more useful sentence than just a dollar figure.

How reliable is the recommendation?

Same caveats as the benchmark card, with one addition: the model is reasoning, not measuring. It's good at:

• Spotting when your rates are clearly out of line with regional norms.
• Adjusting for scope complexity that your calculator doesn't know about (a "1-hour mapping job" of 200 acres is not the same scope as a "1-hour real-estate shoot").
• Recognizing pricing-situation effects (you should generally bid more aggressively on a take-it-or-leave-it flat-rate offer than on an open-bid platform).

It's less good at:

• Hyper-local pricing in markets without much public data (sorry middle-of-nowhere Nebraska).
• Specialty payloads (Light Detection and Ranging (LiDAR), multispectral, etc.) where benchmarks are sparse.
• One-off jobs that don't fit any standard scope bucket.

When the AI's number is inside ±10% of your fair-bid, the market and your math agree — high confidence. When it's outside ±25%, treat it as a strong signal to investigate, not to blindly follow. Read the signals and decide.

What to do with the verdict in practice

• Verdict "fair": Send the bid. You and the market are aligned.
• Verdict "below": Don't just match the AI's higher number. Ask why — is your fly rate set too low? Is the scope heavier than typical for this industry? Adjust the root cause, then re-quote.
• Verdict "above": Check the signals. Sometimes you're a premium operator and that's the brand. Sometimes you're over-pricing a routine job and you'll lose it for no reason. The signals tell you which.
• Verdict "no-bid": Enter a bid first, then re-run.

The point

The AI Bid Recommendation isn't there to replace your calculator. It's there to ask the calculator a question it can't ask itself: is this number any good in the real world?

Treat it like a second pilot looking over your shoulder. Smart, well-informed, occasionally wrong, but always willing to show its work. That last part — the signals and the divergence summary — is the entire reason it exists. The dollar figure is just the headline.

For context on the benchmark numbers feeding this recommendation, see Industry Benchmarks in Rotor Rate .

Sources & further reading

The AI second-opinion is only as good as the pricing logic and benchmark data behind it. If you want to dig into either:

Pricing fundamentals

• McKinsey & Company — *The power of pricing
• Harvard Business Review — *Pricing strategy topic hub

Industry data

• Drone Industry Insights (DroneII) — *Drone Industry Insights — market reports
• Skylogic Research — *DroneAnalyst / Skylogic Research
• Droners.io — *Droners.io public job board

Rotor Rate companion reads

• How Industry Benchmarks work in Rotor Rate
• When to raise your drone rates: signals & benchmarks
• Why most drone quotes lose money
• Setting your walk-away floor