What this is about
A customer enquiry arrives by email. Some details are in the message, some in an attachment, and no price has been discussed. Before you write a quote, you first have to work out what the customer actually wants.
An AI can do that groundwork: it sorts the details and shows what is still missing. Whether you commit to anything, and at what price, remains your decision.
An enquiry becomes a basis you can check.
Fictional enquiry. Not tested with an AI tool. The result shown was not produced by AI.
Maintenance of office lighting
“We would like annual maintenance of our office lighting. The work covers two floors. We would prefer a date next month. Please send us a quote.”
“Sort this enquiry: what is confirmed, what is open? Suggest an opening sentence for the reply. Do not invent a price, scope or date.”
- Annual maintenance
- Office lighting
- Two floors
- Number and type of lights
- The work required and access
- Exact period and pricing basis
“Thank you for asking about maintenance of the lighting on two floors. Before we can provide a firm quote, we need the number of lights and details of the work required.”
Why you still need to check
AI can produce convincing but false statements. The US institute NIST warns about this explicitly and recommends that people review the results [Q1].
So check every statement in the draft: is it in the enquiry or in your price and service documents? If the draft says "all lights including replacement materials" or "date guaranteed", that was made up and has to go.
Who this suits
It suits you if similar enquiries keep coming back, your quote process is settled and someone in the business is authorised to decide scope and price.
It suits you less if every project is entirely different, nobody can check the draft, or the enquiry contains confidential details and you have not settled which tool may be used. Personal data call for data minimisation and a lawful basis [Q2]. This is a general note, not a legal assessment of your case.
Your small next step
Take a fictional enquiry and your current quote template. Check whether the AI draft shows every open point and promises nothing new. Only when that works is a test with real, permitted data worthwhile. The AI Check can help you pick a single process for it.
The example is fictional and was not tested with an AI tool. It states no time saving and no accuracy rate.
- [Q1] NIST, Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile, section 2.2 and guidance on human review. On false AI output and human review.
https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.600-1.pdf(checked 11 November 2026) - [Q2] European Union, General Data Protection Regulation, Article 5(1)(c) and Article 6. On data minimisation and lawful bases. Whether a particular tool is permitted needs a separate check.
https://eur-lex.europa.eu/eli/reg/2016/679/oj(checked 11 November 2026)