As of 13 August 2026, AI can only partly estimate the cost of your software development project.
Most people should hand this to a purpose-built tool.
Can you do it?
15 minutesto a draft.
1 hourto something you’d act on.
Cost, all in£0
Skill neededpower-user
Who has to check ita colleague
What the alternative costsThe supplied tool information does not state a price for a human software project estimation service.
If this goes wrong: you approve an under-scoped budget, then absorb delays, change requests or an unplanned reduction in features.
What to actually do
Use a tool built for this
The route this page recommends
Hand it to a person
Second choiceA person who owns the outcome does this end to end, worth it when the failure is dear.
Do it yourself
The distant thirdA chat interface, power-user skill, and roughly 1 hour until you can act on the result.
How to actually do it
- Open the project brief, user stories, acceptance criteria and any existing technical documentation, then combine the current requirements into one feature list.
- Gather the proposed team roles, hourly or daily rates, working hours, VAT treatment, target date, platforms, integrations and non-functional requirements.
- Export comparable past project estimates or actual effort and cost data, removing confidential information that is not needed for the estimate.
- Paste the gathered information into the prompt and label every missing item as 'NOT PROVIDED' rather than filling gaps yourself.
- Ask the model to produce the work breakdown, low, expected and high scenarios, arithmetic, exclusions, dependencies and validation questions specified in the prompt.
- Copy the estimate into a spreadsheet and compare every multiplication, total and currency assumption against the supplied rates and source data.
- Ask a developer or delivery manager to check the scope, technical risks, dependencies, testing effort and scenario ranges before using the estimate for approval or a client quote.
- Update the estimate with the reviewer’s corrections and record which assumptions, exclusions and risks remain unresolved.
Prompt
Estimate the cost of this software development project for a UK business. Use only the information I provide and do not invent requirements, rates, staffing, durations or external market prices. Project description: [PASTE DESCRIPTION] Features and acceptance criteria: [PASTE FEATURE LIST] Platforms, integrations and technical constraints: [PASTE DETAILS] Non-functional requirements, including security, performance, availability and compliance: [PASTE DETAILS OR WRITE 'NOT PROVIDED'] Team roles and hourly or daily rates in pounds sterling: [PASTE RATES, INCLUDING WHETHER VAT IS INCLUDED] Known dates, working hours and delivery constraints: [PASTE DETAILS OR WRITE 'NOT PROVIDED'] Past project data or comparable estimates: [PASTE DATA OR WRITE 'NOT PROVIDED'] Produce: 1. A work breakdown by phase and feature. 2. Estimated effort for each item, showing the unit used and the reason for the estimate. 3. Labour cost for each item and a total in pounds sterling. 4. Separate costs for discovery, design, development, testing, project management, deployment and post-launch support where the information permits. 5. A low, expected and high scenario, with the assumptions that create each scenario. Do not create a contingency percentage unless I provide one. 6. A list of missing information that could materially change the estimate. 7. Dependencies, risks, exclusions and items that need technical discovery. 8. A short list of questions for the client or project team. Show the arithmetic clearly. Distinguish supplied facts from assumptions. Do not present the result as a quote, fixed price or delivery promise. End with a section titled 'What a technical estimator must validate' listing the parts that require review by an experienced developer or delivery manager.
Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.
What it gets wrong
- AI cannot discover unstated requirements that the client, users or existing systems have not documented.
- AI cannot reliably judge whether an unfamiliar integration, migration or security requirement will be straightforward in your environment.
- AI cannot validate its effort ranges without relevant historical delivery data or an experienced technical reviewer.
- AI cannot take responsibility for a budget, fixed-price quote, missed deadline or scope dispute.
- AI can make an estimate look precise by turning weak assumptions into neat tables and totals.
What caps this at PARTLY: judgement under ambiguity, stakes of error and context depth.
How we scored this
Five axes, each scored nought to two by hand: ten means AI carries the task cleanly, and the thresholds that turn a total into YES, PARTLY or NO are published in the methodology. Each axis name links to its definition.
| Axis | Score (0–2) |
|---|---|
| Output | 2 |
| Inputs | 1 |
| Verification | 1 |
| Liability | 1 |
| Effort delta | 2 |
| Total | 7 / 10 |
The methodology and its thresholds are published in full.
FAQ
- Can ChatGPT estimate the cost of a software project?
- Yes, it can produce a useful first estimate from a detailed scope, team rates and project constraints. Treat it as planning material, then have a developer or delivery manager validate the effort, risks and missing requirements.
- How accurate is an AI software development estimate?
- There is no reliable accuracy level without knowing the quality of the scope and the availability of comparable project data. AI is better at structuring assumptions and arithmetic than predicting work hidden in ambiguous requirements.
- What information does AI need to estimate a software project?
- Give it the features and acceptance criteria, platforms, integrations, security and performance requirements, team roles, rates, delivery constraints and relevant past project data. Mark anything unknown clearly, because an estimate based on invented inputs is not useful.
- Can I use an AI estimate as a fixed-price quote?
- No. An AI estimate is not a fixed-price quote unless the scope, assumptions, exclusions, risks and commercial terms have been reviewed and accepted by the accountable people in your organisation. A technical estimator should validate it before you send it to a client.
Nearby answers
Assessed by gpt-5.6-luna (gpt-5.6-luna) on 2026-08-13, second-checked by an independent model. Wrong somewhere? Email [email protected] and it gets re-checked.
The newsletter
AI news, new answers and product picks, straight to your inbox.