Sub-grade spoke

Inquiry Flow — can an agent complete your contact form on behalf of a user?

Inquiry flow is the contact-and-lead-capture surface that catches everything direct booking can't — B2B sales conversations, custom-quote requests, professional services intake, partnerships, anything that needs a human in the loop after structured submission. An inquiry form an agent can complete is the difference between participating in agent-driven B2B lead routing and being skipped. Most inquiry forms are designed for cold human visitors with infinite patience, not for agents acting on a user's behalf in a single turn.

By Chris Mühlnickel · 2026-05-16

What is Inquiry Flow?

Inquiry Flow is whether your contact, lead-capture, or quote-request form is completable by an AI agent on behalf of a user — minimal required fields, no bot-killing CAPTCHA, predictable validation, and a machine-parseable success/error response.

By the numbers

Why it matters

Inquiry flows are how non-self-serve services convert agent-driven interest into qualified leads. For most B2B services, custom-quote work, professional services, and complex SaaS sales, the "buy" action isn't a cart checkout — it's a contact submission that triggers a human sales conversation. If the inquiry form is agent-completable, the agent routes high-intent users into the sales funnel at a steady rate; if the form is a 12-field wall with a reCAPTCHA, agents abandon and the user goes to a competitor that submitted faster. Per Unbounce, the median B2B form conversion rate is 6.6% across 57 million submissions — the headline metric for the entire category. The agent-driven population concentrates above that median when the form is completable and well below it when it isn't.

The HubSpot 4.1% per-field drop is the linear cost of complexity. Every additional field a form asks for drops conversion ~4.1% on human users — and the drop is steeper for agents, because agents can't make qualitative judgment calls about which optional fields to skip and which to attempt. A 4-field form (name, email, company, intent) clears the bar; a 12-field form with company-size, annual-revenue, role-title, source-attribution, marketing-consent, and three open-ended descriptions blocks most agent-completable scenarios. The fix is not abandoning qualification — it's moving qualification after the submission rather than gating it before.

Multi-step forms work for agents too, when structured correctly. HubSpot's research puts multi-step forms at ~86% higher conversion than single-page equivalents — and the agent-completion rate tracks the human pattern when each step has a stable URL the agent can return to. The failure mode is multi-step UIs that hold state in JavaScript memory with no URL anchor for each step — agents that lose context between steps can't resume. The pattern that works: each step is a route (/contact/step-1, /contact/step-2), and the agent submits the cumulative payload at the final step's endpoint. Same UX for humans, fully completable for agents.

B2B agent-driven discovery is the under-built channel. Today, agent-routed B2B lead generation barely shows up in the marketing dashboards because the volume is small and the attribution is opaque. By 2027 — as Claude.ai, ChatGPT, and Perplexity become the entry-point for B2B vendor research at scale — the cohort of inquiry forms designed for agent completion captures a disproportionate share of the agent-routed pipeline. The cost of being above-median on inquiry-flow design is low (drop fields, drop CAPTCHA, return structured responses); the leverage is high in a channel that's still uncrowded.

Where it's heading

Inquiry forms gain structured-intent schema. Schema.org and parallel industry standards are converging on machine-readable intent declarations for inquiry forms — Action-style schema that lets agents pre-classify what the form is for and what the response will be. Sites that ship the schema let agents route the right user intent to the right form without ambiguity; sites that don't get included as second-choice candidates.

Multi-turn agent-driven qualification becomes the post-submission norm. Today the inquiry-to-conversation flow is: user submits → human SDR responds → multi-turn email exchange. Tomorrow: user submits via agent → an SDR-agent (or a tightly-supervised SDR-assist tool) handles qualification in 2-3 structured turns → human sales is involved only at the decision moment. The inquiry form is the entry point; everything downstream gets agent-augmented over the next 24 months.

Agent-platform identity replaces form bot-defense. The strategic shift: stop treating every submission as potentially fraudulent, start verifying the agent-platform identity that signed the submission. ChatGPT signed-user tokens, Claude signed-organization tokens, platform-level reputation. Sites that wire this up keep their CAPTCHA gates off entirely for verified agent-platform traffic and gate only the unsigned submissions — the inverse of today's pattern.

Common mistakes

  • 12-field forms gating cold contact. Per HubSpot's per-field 4.1% drop, every field past the essentials costs conversion. Agents inherit the cost without the human's ability to skip an optional field with judgment.
  • CAPTCHA on every submit. The single biggest agent-killer at the inquiry layer. Use rate-limit-by-IP at the gateway and invisible challenges, not full-page reCAPTCHA.
  • 'Call us' as the only path. A phone-only inquiry route is a hard fail for the agent-completable check, even when voice agents technically exist.
  • Validation that lies about required fields. Marking a field as optional in the HTML but rejecting submissions that omit it. Agents follow the spec; humans figure it out by trial and error. Be honest in the markup.
  • Success states that aren't machine-readable. Redirecting to a thank-you page with no structured response forces agents to parse HTML to confirm submission. Return a JSON success body alongside the rendered redirect.

Frequently asked

What does an agent-completable inquiry form look like?

Four things: (1) every field is explicitly labeled with <label for=...> and a matching name attribute — agents read the semantic markup, not the placeholder text, (2) required vs. optional is unambiguous — required attribute on required fields, no JavaScript-only validation that lies about which fields are needed, (3) submission target is a stable URL accepting standard form-encoded or JSON POST — no signed CSRF tokens that expire before the agent can submit, (4) the response is machine-parseable — HTTP status + structured success/error body, not a redirect to a thank-you page that requires HTML parsing to interpret.

How many fields are too many for an agent-completable form?

Per HubSpot, conversion drops ~4.1% per additional field for human users — and the agent-completion drop is steeper because agents don't have a 'I'll skip this and come back' option. The pragmatic answer: required fields = the absolute minimum your sales process needs (name, email, company, intent). Everything else moves to the post-submission qualification flow. Multi-step forms (per HubSpot, 86% higher conversion than single-page equivalents) work for agents too — but only if each step has a stable URL and the agent can resume from any state.

What about CAPTCHAs on the inquiry form?

Most break agents. The 'just whitelist OpenAI's UA' answer is brittle — the safer move is to use friction-low alternatives: Cloudflare Turnstile in invisible mode, honeypot fields the agent can detect and skip, rate-limit-by-IP at the gateway rather than challenge-on-every-submit. CAPTCHA-on-every-form is one of the top three Usability blockers across verticals in our calibration corpus. Sites that need bot defense get it from upstream (CDN, gateway) without gating the form itself.

Does an agent-friendly inquiry form mean lower lead quality?

No — and the assumption is backward. The lead-quality lever is qualification after submission, not friction before. A clean low-field form with strong post-submission scoring (CRM enrichment, intent-based routing, BANT qualification in a follow-up email) produces higher net-quality leads than a 12-field gate that 60% of legitimate prospects abandon. Agent-completable forms route higher-intent prospects in (a user wouldn't ask an agent to fill the form if they weren't intent on outcome) and the cost is structural simplification you'd want anyway.

Should the form integrate with my CRM directly?

Yes, and the right pattern is the same one that works for human submissions: form POSTs to your endpoint → endpoint validates and creates the CRM record → endpoint returns structured confirmation. Agents don't need a separate path — they consume the same submission flow. The work is making sure the endpoint is well-documented (see Agent API Spec) and the CRM downstream handles the agent-flagged submission correctly (don't auto-send a 'thanks for reaching out' email that confuses the user when it was an agent-routed lead).

Do agents need to render the success page, or just parse the response?

Response parsing is the right surface. The success page is a human-rendered view of the same state. The structured pattern: POST /inquiries returns { "status": "received", "inquiry_id": "inq_123", "next_step": "sales_will_contact_within_24h" }, and the rendered success page reads from the same model. Agents check the JSON and confirm to the user; humans see the page and feel reassured. Sites that ship only the rendered page force agents to scrape HTML to confirm submission, which is brittle.

What about file attachments — can agents include them?

Yes, with caveats. Agents can attach files the user provided (a PDF brief, a screenshot, a portfolio sample). The constraints are size limits documented upfront, accepted MIME types enumerated in the form's accept attribute, and clear error responses when an attachment fails validation. Files added by the agent itself (synthesized reports, mocked data) are a downstream policy question for the agent platform; from the inquiry-form side, the contract is the same as for human submissions.