Methodology
How the blueprint estimates work
Most online estimators are either a contact form in disguise or a language model guessing. This one is neither: the numbers come from a deterministic estimation engine with assumptions we can show you, and AI is only allowed to touch the words.
Your answers select the feature list
Every feature carries an effort band and a rule for when it applies. Say there’s nothing to integrate with and the integration work disappears — from the list and from the price. The scope you see in your blueprint is literally what got priced: each item shows the hours it contributed.
Overheads and multipliers scale that effort
Discovery, design, project management, QA, and documentation are added as a share of the selected build effort, so they scale with your actual scope. Then five multipliers apply: complexity, rollout size, how many hours a week the work consumes today, how hard your named systems are to integrate with, and whether you need it rushed. Anything you answered "not sure" to widens the band rather than pretending precision exists.
Four delivery models price the same scope
The identical effort estimate is priced under a 2023 pre-AI market benchmark, a U.S. agency today, an offshore team (including the coordination overhead clients usually discover later), and Atlas AI-assisted delivery. Each carries its own rate, overhead, and AI-acceleration assumption, and each shows where its rate came from. Atlas is deliberately not the cheapest column — offshore is.
The business case computes, or it refuses
Payback comes from arithmetic you can check: the hours you told us this consumes each week, times the share a working system realistically removes, times 48 working weeks, times a loaded labor cost — then the estimate divided by that. If you didn’t tell us the hours, the blueprint says the business case can’t be computed instead of inventing an ROI figure.
AI writes words, never numbers
AI reads your problem description to pre-fill answers, and it writes the narrative sections of your blueprint. Every figure — cost, timeline, complexity — comes from the deterministic engine. A grounding check rejects any AI-written sentence that mentions a number the engine didn’t produce.
What this is not
A planning estimate, not a quote.
- This is benchmark-grounded, not fitted to a secret dataset. The Atlas column is back-solved from our own published package prices, so a blueprint can't quote you something we don't actually charge. The agency and offshore rates come from commonly published market ranges. Every model shows its rate basis next to its numbers.
- The offshore coordination contingency is our judgment about specification and rework overhead, not a measured figure — it's labeled as an assumption because that's what it is.
- The “2023 build” column is a pre-AI market benchmark for the same scope, not a quote from any specific vendor.
- Assumptions carry a version stamp — your blueprint records exactly which configuration produced it, so the same answers always produce the same numbers.
- Final scope, cost, and timeline always require human review — which is also the first thing we'd say on a call.