How to Spot a Time-Wasting Enquiry With AI
Score every enquiry before you quote it. A simple AI scoring pass on the signals that predict non-conversion, and the rule that keeps it safe.
To spot a time-wasting enquiry with AI, have it score each new enquiry against signals that genuinely predict non-conversion, then sort your replies by that score. You still answer everyone. You just stop spending two hours on a detailed quote for someone who sent four words at 1am to six businesses at once.
The important design decision is up front: the AI ranks, it does not reject. Every rubric below produces a score and a reason, never a decision to ignore someone. Get that wrong and you will eventually auto-ignore your best customer of the year.
The signals that actually predict a time-waster
Generic lead scoring looks at company size and job title, which is useless for a trades business or a small studio. What correlates for small businesses taking enquiries by form, email or message is more specific:
Signal | Weight | Why it predicts |
|---|---|---|
No budget or price range mentioned, and none asked for | Medium | Serious buyers usually anchor on money early. Silence on it often means no budget exists yet. |
Asks only for a price, with no detail about the job | High | A price-only question with no specifics cannot be answered accurately and is normally being asked of several suppliers at once. |
No timeline, or a timeline of "ASAP" with no date | Medium | Real projects have a reason they need to happen by a date. ASAP with no date usually means no decision has been made. |
Text looks copy-pasted or addressed to nobody | High | Identical enquiries sent widely convert poorly. A message that does not mention what you do is a strong tell. |
Scope is vague but large | High | "I need a full website and an app and maybe AI" with no detail is a conversation, not a job, and often a long unpaid one. |
Mentions a specific thing you actually do | Negative | The strongest positive signal there is. Specificity costs effort, and effort means intent. |
Note that the last row subtracts. A scoring system with no negative signals will rank everything as medium and tell you nothing.
A prompt that produces something usable
The difference between a useful scoring prompt and a useless one is forcing the model to cite evidence from the enquiry. Without that, you get confident numbers with nothing behind them.
Score this enquiry from 1 to 10 for how likely it is to become paid work.
My business: [one line, what you do and your typical job size]
Use these signals:
- No budget mentioned or asked about: lower
- Price-only question with no job detail: much lower
- No timeline or a date-free "ASAP": lower
- Looks copy-pasted or does not mention what I do: much lower
- Vague but very large scope: lower
- Names a specific service I offer, or a specific constraint: higher
Reply with exactly:
SCORE: n/10
WHY: two sentences, quoting the words from the enquiry that drove the score
MISSING: the one question I should ask before quoting
Enquiry:
"""
[paste the enquiry]
"""The MISSING line is the part that earns its keep. Most low scores are not rejections, they are enquiries with one unanswered question, and asking that one question takes a minute instead of the hour a full quote takes.
How to use the score without being rude
Sort your effort, not your manners. A practical tiering:
Score 7 and above: reply properly today, with the quote or a call booking.
Score 4 to 6: send a short reply asking the MISSING question. Quote only once you have the answer.
Score 3 and below: still reply, but with your shortest template and a link to your pricing or process. Cost to you: two minutes.
Everyone gets a response. That matters commercially as well as ethically, because the scoring will be wrong sometimes and a two-minute reply keeps the door open when it is. If you do not have those short templates yet, the drafting approach in how to automate email replies with AI covers building them without sounding robotic.
Where this goes wrong
Writing style is not intent. Someone terse, dyslexic, typing on a phone, or working in a second language will score low on a rubric that rewards polish. Score on the presence of facts, never on the quality of the prose.
Repeat customers must bypass the whole thing. Check your customer list before scoring, or you will score a loyal client as a stranger.
Referrals arrive looking like time-wasters. "Hi, Dave said to call you" has no budget, no timeline and no detail, and is often the best enquiry of the month. Add an explicit rule that a named referral scores high.
Do not let it reply by itself. A scoring pass that also sends the response is one bad judgement away from insulting a customer in writing.
That third point is worth a manual override. Keep a short list of referral sources in the prompt. On data handling, remember an enquiry contains someone's personal details, so the same care applies as anywhere else you paste customer information into a model.
What to do with the scores over time
Log the score next to what happened. After thirty enquiries you will know whether your rubric predicts anything, and you will almost certainly find one signal that matters far more than the others for your particular trade. Reweight toward it and drop the signals that never moved.
This is also the cheapest route to understanding your own enquiry mix, which tends to be more revealing than any marketing report. If the data lives in a messy spreadsheet of past enquiries, cleaning it is its own small job, covered in how to clean up a customer list with AI. For the stage after qualification, pricing the work itself, see how to use AI to write customer quotes, and for the wider set of jobs AI can take off a small team, AI for small business.
Frequently asked questions
Can AI tell me which leads will definitely convert?
No, and a tool claiming to is selling confidence rather than accuracy. Scoring sorts a pile by likelihood so your first hour goes to the top of it. It does not predict individual outcomes.
Should I tell customers their enquiry is being scored?
You are prioritising your replies, which every business does informally. You do not need to announce it. You would need to be transparent if a score triggered an automatic rejection, which is one more reason not to build that.
What if the AI scores a good customer badly?
It will, which is why the lowest tier still gets a reply rather than silence. Review your misses monthly and add a rule for the pattern you missed, such as referrals or repeat clients.
Is it worth doing with only a few enquiries a week?
Probably not as a system. At five enquiries a week you can read them. The scoring pass pays off when the volume is high enough that you are quoting in the order things arrived rather than the order that makes sense.
How did this land?
About the author
Developer Advocate
Steve builds something with Swarmz every week and writes up what worked, what broke, and what he'd do differently. Tutorials and hands-on guides are his lane.
