AI for WordPress Form Submissions
Score, summarize, categorize, and analyze every form submission - without sending data to OpenAI, Anthropic, or any third-party LLM by default. Self-hosted inference. Opt-in per feature. Monthly caps make cost predictable.
Four AI capabilities, two add-ons
AI Lead Scoring is one add-on. AI Analytics groups the remaining three capabilities. Run one, the other, or both.
AI Lead Scoring
Every submission scored 0-100 with hot, warm, or cold category and a one-sentence reasoning. Columns land directly in your Google Sheet. Ideal for sales prioritization.
AI Analytics - Aggregated
Weekly natural-language brief covering volume trends, source attribution shifts, duplicate detection, and churn scoring. The model sees aggregates, never raw PII.
Per-Submission Sentiment
Classify each submission as positive, neutral, or negative with a 0-100 confidence score. Flag angry customer replies before they escalate.
Per-Submission Summarize
One-to-two-sentence plain-text summary of what the submitter wants. Read an inbox of 200 submissions in 5 minutes instead of 45.
Per-Submission Categorize
Pick exactly one category from your allowed list (sales, support, spam, partnership, etc.) with a confidence score. Route or filter without writing regex.
BYOK for AI Analytics
Optional: route AI Analytics calls through your own OpenAI or Gemini API key. Agency Plus plan includes this capability. Useful if your tenant contracts forbid shared inference.
Score every submission 0-100
Each WordPress form submission runs through a self-hosted Nemotron model and returns a purchase-intent score, a hot/warm/cold category, and a one-sentence reasoning. The three columns write directly to your Google Sheet alongside the submission data, so your sales team filters, sorts, and follows up inside the tool they already use.
- Scoring bands: 70+ hot (ready to buy), 40-69 warm (interested), below 40 cold (unlikely)
- Signals evaluated: field content, email domain (business vs. free), message intent, company name presence
- Concurrency: 1 concurrent job on the add-on, 2 on Agency Bundle, 4 on Agency Plus
- Security: prompt-injection preflight + output sanitization; malicious submissions auto-scored cold
- Latency: typically under 3 seconds per submission; non-blocking (never slows down form delivery)
| Message | Score | Category | Reasoning | |
|---|---|---|---|---|
| j.smith@acme-corp.com | Need pricing for 50-seat rollout by end of Q2 | 87 | hot | Specific seat count and deadline signal active evaluation. |
| alex@gmail.com | Just browsing your product, looks interesting | 48 | warm | Generic interest language with consumer email domain. |
| noreply@spam-bot.xyz | Click here for cheap SEO services | 12 | cold | Outbound solicitation, not an inquiry. |
Patterns and per-submission intelligence
AI Analytics covers two jobs. The aggregated layer watches submission volume, UTM source mix, and duplicate rate over rolling windows and writes a weekly plain-English brief to your admin dashboard (and optionally to email). The per-submission layer is three independent opt-in toggles that run on every new delivery.
Aggregated features
- Trend detection: volume, UTM source, and category shifts week-over-week
- Duplicate detection: fuzzy match on email, content similarity, IP, timing
- Churn scoring: identify lapsed or disengaging submitters (Agency Plus)
- Weekly email digest: natural-language summary for the business owner
Per-submission extras (opt-in)
- Sentiment: positive / neutral / negative + confidence + rationale
- Summarize: one-to-two-sentence plain-text abstract of the submission
- Categorize: pick exactly one of your allowed categories, with confidence
Submissions this week: 284 (+18% vs. last week)
Notable shifts:
- Google Ads UTM volume up 34%, driven by the new /demo landing page
- Facebook Ads UTM volume down 40% - pausing ad budget may explain this
- Duplicate rate dropped from 7.1% to 2.4% after you enabled email-window dedupe
Sentiment mix: 62% positive, 28% neutral, 10% negative (up from 6% last week - 3 refund requests flagged)
Category mix: 58% sales, 22% support, 12% partnership, 5% spam, 3% other
Example output only. Actual briefs describe your real data.
Self-hosted by default, BYOK when you need it
We built the AI features so that the default answer to 'where does my form data go?' is a boring one: nowhere new.
Self-hosted Nemotron
Default inference runs on a Nemotron Nano model we host on GPU infrastructure we operate. No third-party LLM provider sees your data. Traffic never leaves our network unless you explicitly configure a BYOK key.
BYOK (Agency Plus or AI Analytics BYOK)
Prefer to route analytics inference through your own OpenAI or Gemini API key? Enable BYOK in the add-on settings. Your API key is encrypted at rest. Model calls are billed directly by the provider, not by us.
Retention
AI Lead Scoring: 7-day inference logs. AI Analytics aggregated features: 0 days raw, aggregates only. AI Analytics per-submission extras: 7 days when enabled. Opt-in gates prevent accidental data movement.
Which plan unlocks which AI features?
Pick individual add-ons, or grab a bundle that includes both.
| Capability | AI Lead Scoring add-on $29/mo | AI Analytics add-on $39/mo | Agency Bundle $69/mo | Agency Plus $99/mo |
|---|---|---|---|---|
| AI lead scoring (per submission) | 300/mo | — | 1,500/mo | 3,000/mo |
| Weekly trend brief | — | ✓ | ✓ | ✓ |
| Duplicate detection | — | 50/mo | 25/mo | 100/mo |
| Churn scoring | — | 500/mo | — | 1,000/mo |
| Sentiment / Summarize / Categorize (shared monthly cap) | — | 100/mo | 50/mo | 200/mo |
| BYOK (OpenAI / Gemini) | — | ✓ | — | ✓ |
| Scoring priority queue | Standard | — | Priority 1 | Priority 0 (highest) |
Need more volume? Contact sales for a custom enterprise allocation. All allowances refresh monthly on your billing date.
AI Features FAQ
Where does the AI inference run?
By default, all AI runs on a self-hosted Nemotron model on GPU infrastructure we operate. Your form submission data never touches OpenAI, Anthropic, or any other third-party LLM. Optional BYOK mode (AI Analytics add-on only) lets you route per-submission extras to your own OpenAI or Gemini API key if you prefer that pathway.
Are these features included with the core plugin?
No. The core plugin ships with Smart Mapping (a local field-name suggester) and the analytics table - no hosted AI. To unlock AI Lead Scoring you need the AI Lead Scoring add-on ($29/mo) or a bundle that includes it. To unlock AI Analytics and the per-submission extras you need the AI Analytics add-on ($39/mo) or a bundle.
What is the difference between AI Lead Scoring and AI Analytics?
AI Lead Scoring evaluates each submission and returns a 0-100 quality score plus a hot/warm/cold category. It is a single number per submission, designed for sales prioritization. AI Analytics is broader: weekly trend briefs, duplicate detection, churn scoring, plus three per-submission extras (sentiment, summarize, categorize). Many customers run both - one for prioritization, one for pattern detection.
How do the per-submission extras (sentiment, summarize, categorize) work?
Each extra is an opt-in toggle in SheetLink > Analytics. When enabled, every successful submission fires a non-blocking request to our inference endpoint. Sentiment returns positive/neutral/negative with a confidence score. Summarize returns a one-to-two-sentence plain-text summary. Categorize picks exactly one category from your allowed list. Results post into your analytics table and can be mirrored to Google Sheet columns via mapping.
Is there a monthly cap on AI usage?
Yes. Each plan has a monthly AI Analytics allowance (300 calls on the AI Analytics add-on, higher on agency bundles). Trend, duplicate, churn, and per-submission extras all count against the same cap. AI Lead Scoring has its own separate monthly cap (300 on the add-on, 1500 on Agency Bundle, 3000 on Agency Plus). Caps are soft in the sense that exceeding them returns a 429 - submissions keep flowing to your sheet, the AI just stops running until the next month.
Can I see what the model returned?
Yes. For lead scoring, each submission lands in your Google Sheet with Score, Category, and Reasoning columns. For per-submission extras, results are stored on the analytics job record and can be mirrored to sheet columns via field mapping. Reasoning is sanitized server-side to strip any quoted field content from the submission back out of the model output.
What happens to submissions with prompt-injection attempts?
The scoring model treats every field value as untrusted user data. Prompts are structured with explicit system instructions, delimited user content, and a preflight injection detector. If the submission looks like an injection attempt, it is auto-scored as cold with score <= 30 and the reasoning is set to a neutral fallback message. Per-submission extras apply the same delimited-content pattern.
How long are AI inference logs retained?
AI Lead Scoring: 7 days. AI Analytics aggregated features (trend, duplicate, weekly digest): 0 days raw, aggregated metrics only. AI Analytics per-submission extras (sentiment, summarize, categorize): 7 days when those toggles are enabled. See the Privacy page for the complete retention schedule.
Turn on AI in about 3 minutes
Install the core plugin, subscribe to an AI add-on from your account portal, paste the activation token, flip the toggles you want. No deploys, no model configuration, no separate vendor onboarding.