Verified against ChatGPT · 2026-08-11
Run a desk research pass that tells you which findings to trust and which to verify yourself
Produces a structured desk research brief that separates findings by how confident you should be in them, rather than presenting a scraped-together summary as uniformly reliable.
The prompt
Ready to copy — highlighted parts are example details you can swap.
I need a desk research brief on the topic below. Structure it so I can immediately tell which claims are solid enough to act on and which need independent verification before I use them — do not present everything with the same level of apparent confidence. RESEARCH TOPIC How mid-market SaaS companies typically structure usage-based pricing tiers. DECISION THIS WILL INFORM Deciding whether to move our product from seat-based to usage-based pricing next quarter. WHAT I ALREADY KNOW (so you don't re-explain it) We already know the basic difference between seat-based and usage-based models; skip that part. TIME BOX This needs to be useful for a decision meeting tomorrow, not an exhaustive report. STRUCTURE THE BRIEF AS THREE CONFIDENCE TIERS: TIER 1 — WELL-ESTABLISHED: points that are widely corroborated or definitional, where you're confident in the claim itself even if you can't pin an exact source. TIER 2 — LIKELY BUT UNVERIFIED: points that sound right based on general patterns in the space but that you cannot confirm are current, precise, or specific to the exact context in How mid-market SaaS companies typically structure usage-based pricing tiers. — flag these explicitly as needing a check against a live source before they're used in Deciding whether to move our product from seat-based to usage-based pricing next quarter.. TIER 3 — OPEN QUESTIONS: things relevant to the decision that desk research genuinely can't answer without either primary research or a source you don't have access to — name these directly rather than papering over them with a plausible-sounding guess. For every specific number, statistic, or named claim (a market size, a percentage, a named competitor's practice), do not present it as fact unless you're highly confident it's both accurate and current — if you're not sure, either omit the specific figure and describe the shape of the finding qualitatively, or place it in Tier 2 with an explicit note to verify. Never invent a number to make a point sound more concrete than it is. Given the time box, prioritize breadth of the Tier 1 and Tier 3 sections over exhaustive Tier 2 detail — a decision-maker on a tight clock needs to know what's solid and what's missing more than a long list of half-confirmed maybes. OUTPUT FORMAT Three labeled tiers as above, each with short bulleted points. End with a one-line summary of what would most change the decision in Deciding whether to move our product from seat-based to usage-based pricing next quarter. if it turned out to be wrong.
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
The three-tier confidence structure directly counters the single biggest risk of using a language model for desk research: GPT-5.1 states well-established background knowledge and half-remembered, possibly outdated specifics in the exact same fluent, confident register, so a reader has no built-in signal for which claims to trust — forcing an explicit tier split makes the model's own uncertainty visible instead of hidden behind uniformly confident prose. The instruction to never invent a specific number targets a well-documented failure mode where a model asked for concrete figures will produce a plausible-sounding statistic that fits the expected shape of an answer even when it has no reliable basis for that exact number — permitting a qualitative description instead of a fabricated figure gives the model an honest way out that doesn't sacrifice usefulness. Naming Tier 3 as questions desk research genuinely cannot answer matters because a model under pressure to be comprehensive will otherwise stretch a weak inference to cover a real gap rather than admitting the gap exists, which is far more dangerous for a decision-maker than an honest "we don't know" — an unacknowledged gap gets treated as covered ground. The time-box instruction to prioritize Tier 1 and Tier 3 breadth over Tier 2 depth reflects how these briefs actually get used: a decision-maker on a clock needs to know what's solid and what's missing far more urgently than an exhaustive list of maybes, so the prioritization keeps the brief matched to how much attention it will realistically get.
What you get back
Tier 1: Usage-based pricing is now standard for infrastructure and API-metered products; hybrid seat-plus-usage models are increasingly common as a transition step. Tier 2: exact adoption percentages across mid-market SaaS specifically vary by source and may be dated — verify against a recent survey before citing a number in the meeting. Tier 3: how usage-based pricing affects net revenue retention specifically for products with highly variable per-customer usage isn't something desk research alone can answer confidently; this needs your own cohort data.
Verified against
ChatGPT GPT-5.1 · 2026-08-11
Changelog
- 2026-08-11 — Initial publish, verified against ChatGPT GPT-5.1.
Need this built into your business?
If a prompt isn't enough — what Scult builds, built and maintained for you — that's Scult's day job.
EXPLORE WHAT SCULT BUILDS
