Answer
How do you use AI with forms and enquiry data?
To turn what people wrote into fields you can count, and to tell you what your form should have asked.
Use it to convert what people wrote into structured fields you can count and route on, and to find what your form failed to ask. The second output is usually worth more than the first.
Enquiries arrive as prose and businesses need structure. What they want, where, when, roughly what budget, how urgent, whether they are an existing customer. Extracting these into fields is the highest-value routine application available to most small businesses, because it makes the enquiry countable, routable and answerable without reading the message again, and because the extraction is checkable — the fields sit beside the text they came from.
The structure is what makes measurement possible at all. Response time, conversion by source, enquiry volume by type, which categories are growing: none of these can be computed from a mailbox of prose. Businesses that add extraction usually discover within a month that they had been wrong about where their enquiries came from and what they were asking for, which is a finding that changes decisions rather than saving time.
The second output is the more interesting one. Every field the system had to infer, and every enquiry where it could not, is evidence about your form. If the budget is never stated, the form should ask. If half of enquiries need a follow-up question about location, the form is missing a field. Using the extraction failures as a form specification closes the loop, and each field added upstream removes an inference downstream, which is strictly better than improving the inference.
Routing follows from the fields and should be rules rather than judgement. Once the category, urgency and location are extracted, deciding who handles it is a lookup. This is worth stating because routing is often built as a second model step, which introduces variance into a decision that has a right answer and makes it impossible to explain why an enquiry went where it did.
Two cautions. Extraction failures should leave a field empty rather than guessing, because an empty field prompts a person and a wrong one does not. And an enquiry containing something outside the ordinary — a complaint, a legal matter, an unusual request — should be routed to a person rather than categorised, since the categories were designed for the ordinary case and the value of the unusual ones is concentrated in handling them well.
The last application is drafting a first reply from the extracted fields, and it works because the content is assembled rather than invented. Acknowledging what they asked for, confirming what happens next, asking the one question that is genuinely missing. That is a message drawn from your records and your process rather than a generated statement, which is what makes it safe to send quickly.
Every enquiry you had to read twice is a field your form should have had.
Siddharth Sharma, Context Theory
Related questions
Should the form just be longer instead?
Up to a point, and the trade is real: each additional required field reduces the number of people who complete it. The productive resolution is to ask for the fields that change what you do and let the rest be extracted from whatever they wrote, which gives you the structure without the abandonment.
What about enquiries by phone?
The same structure applies and the capture is the hard part. A short form that whoever answers fills in during or after the call gets you the same fields, and it is more reliable than transcription for a small business. The value is in having every enquiry in one structured place, regardless of how it arrived.
METHOD
Every figure below carries its source and the date it was verified. Nothing on this page is asserted.
The numbers on this page.
| What | Value | Specific to |
|---|---|---|
| Firms that never responded to a web enquiry at all | 23% | Category-wide |
| Average B2B first-response time | 42 | Category-wide |
Oldroyd, McElheran & Elkington, "The Short Life of Online Sales Leads", Harvard Business Review (March 2011) · hours · 1.25M inbound leads across 2,241 US firms · verified
What is specific to this page.
| Kind | Claim | Check it against |
|---|---|---|
| Workflow | Response time, source conversion, volume by type and category growth cannot be computed from unstructured messages, so extraction into fields is the prerequisite for any enquiry measurement rather than a convenience. | Attempting to compute median first-response time by enquiry type from an unstructured mailbox. |
| Response | Fields the system had to infer and enquiries where inference failed constitute a specification for the form, and each field added upstream removes an inference downstream, which is more reliable than improving the inference. | Listing the fields most frequently absent from enquiry text and comparing against the current form. |
| Software | Routing from extracted fields is a lookup with a right answer, so implementing it as a second model step introduces variance into a deterministic decision and removes the ability to explain why an enquiry was routed as it was. | Checking whether the routing decision for a given enquiry can be reproduced from its extracted fields. |
| Constraint | Enquiries containing complaints, legal matters or unusual requests should route to a person rather than be categorised, because the category set was designed for ordinary cases and the value of the unusual ones is concentrated in handling them well. | Reviewing how the current categorisation handles a complaint arriving through the enquiry form. |
Each row would be wrong on another industry's page. Where a sourced figure exists it is in the table above instead; these are the constraints that shape the work and do not happen to be numbers.
Start with the measurement.
Reading about a benchmark is not the same as knowing your own number. The audit produces yours, measured rather than estimated.
$497 · delivered in 5 business days · credited against month one