Skip to content

DigElite CRM · Function 2 of 4

AI search — the agent radar.

You tell the CRM what type of company you are looking for — industry, region,
Language. The CRM calls the selected provider and checks the responses.
against a strict scheme and places the results in a review queue.
You manually select what fits.

← Back to CRM overview

Three search methods under one hood

Three separate backend pages, one inspection queue.

The CRM backend has three independent search tools — one per provider,
Each with its own backend page and settings block:

  • Contact Radar (Google Places) — classic industry-in-region searches via the Places API (New). Synchronous, low cost, default 40 hits per run.
  • AI search (OpenAI) — prompt-based, synchronous, strict JSON output. Default 20 hits per run (intentionally lower due to token costs and PHP timeout).
  • AI search (n8n) — prompt-based, asynchronous via webhook. A custom settings block plus a ready-made workflow template (JSON file) is included in the plugin.

All three tools write to the same checklist — you do it manually,
which fits. The source (Places / OpenAI / n8n) is assigned as a chip for each candidate.
It's shown so you know where the lead came from.

Version history: Three search providers since v1.31.0; in v1.32.0
Separated into their own backend pages; v1.33.0 has OpenAI and n8n each on a separate page.
It has its own page and provides the n8n workflow as a JSON template.

Three providers, three backend pages

For each search, you work on the page that matches the provider.

The CRM is not tied to a single provider. Classic, generative.
or via its own workflow — each variant has its own backend page and
Their own settings block. There is no longer a provider selection per job.


Classic

Google Places (New)

Fast, accurate for local searches, predictable costs. Delivers publicly published business listings.

  • Synchronous — the answer comes back immediately.
  • Standard for "company + location" searches.
  • Billing via the Google Cloud project (own API key).

Generative AI

OpenAI directly

Synchronous, with strict JSON schema output. Good for more freely formulated search queries or niche industries.

  • Synchronous — the answer comes back immediately.
  • Strict JSON output — no format surprises.
  • Three-level prompt architecture (system / template / free text override).

Workflow

n8n-Workflow

Asynchronous, infinitely scalable. You build the search workflow in n8n, the CRM starts it and receives the response as a webhook.

  • Asynchronous — Webhook trigger starts, REST callback delivers result.
  • You can integrate your own data sources that we are unaware of.
  • Recommended if you need more complex enrichment (industry database, custom datasets).
How to control the radar

Four default templates + free text override.

For the generative providers (OpenAI directly, n8n with OpenAI nodes)
The plugin uses these four default prompt templates. verbatim with.
Variables in {curly brackets} You enter it in the job application form.

  1. Industry in the region.

    „"Find {number} companies in the industry "{category}" in {region} (within a radius of {radius_km} km around {city}). For each match, provide company name, website, business email (if public), phone number, city, and source."„
  2. Specialized providers.

    „"Find {number} specialized {category} providers with a focus on {specialization} in {region}. Prefer small to medium-sized businesses with their own website.""
  3. Associations & multipliers.

    „"Find {number} associations, chambers, guilds or industry networks in the field of "{industry}" in {region} that could act as multipliers."„
  4. Free research.

    „{free text}" — you formulate everything yourself, the system only sets the answer structure and the GDPR limits.

A fixed layer is placed on top of each template. System prompt with the rules
(only public B2B data, no PII scraping, strict response scheme).
You can create your own templates in the template management — save them
Name, body and variables as JSON.

The process of a search job

From inquiry to well-maintained contact.

  1. Open backend page: For each search, you work on the backend page that matches the provider (Contact Radar, AI Search OpenAI or AI Search n8n) — there is no longer a provider selection per job.
  2. Create a job: Default template or free text, fill in variables, language, category, max. hits (default 40 for Places, 20 for OpenAI/n8n), optionally set the "Email enrichment" switch.
  3. Call provider: CRM calls Places or OpenAI synchronously using a strict JSON schema or triggers n8n via webhook.
  4. Normalization: Answers are sent to the unified system. hits[]-Schema (Company, Website, Email, Telephone, City, Region, Country, Source).
  5. Inspection queue: Results are displayed unprocessed — you can discard or accept them with a click.
  6. Contact system: Imported entries receive source=radar, a unique tracking code and status not_contacted. Ready for campaigns and personalized landing pages.

No automatic transfer. The CRM doesn't present anything without verification.
Contact us. The review queue is a mandatory step — also relevant under GDPR, because you
You decide which publicly available data you include in the CRM.

Expected latency per run: Google Places ~0.5-2 s, OpenAI direct
~10–30 s (scale depending on model and number of hits, limited by hard PHP timeout —
(Therefore, the default number of hits is 20). n8n is running fire-and-forget:
The CRM sends the webhook, the WP-PHP timeout does not apply; match.
They end up in the same checklist view minutes to hours later.
Default OpenAI model: gpt-4o-mini, can be changed in the settings.

What AI search is NOT

No full-text search across CRM content.

Important distinction: AI search is no embeddings-based
Full-text search index of your own CRM data ("Which contacts have in the
"We've talked about topic X in the last few months?"). That's the roadmap, currently not in...
Scope of delivery.

The AI search is Discovery — it finds B2B operational data that
These are publicly published and not yet in your CRM. For internal use only.
For CRM search, do you use the normal WordPress search or the list filters in the
CRM backend.

n8n-Workflow — Setup

Workflow template is included in the plugin — five steps to the first search.

Since plugin version v1.33.0, the CRM delivers the n8n workflow JSON as
finished template with — file admin/n8n-radar-workflow.json,
It can be downloaded directly from the "AI Search (n8n)" backend page. This eliminates the need for a separate download.
The DIY workflow; you only need one n8n instance and a
OpenAI API key.

  1. Load JSON template: On the CRM backend page „AI Search (n8n)", click on „Download workflow template", and import the JSON into your n8n instance.
  2. OpenAI credentials: In n8n, create an OpenAI credential and assign it to the OpenAI node in the imported workflow.
  3. Set ENV: The environment variable on the n8n host NZ_CRM_SECRET Store it with the same secret that you enter directly into the CRM.
  4. Activate workflow: In n8n, set the workflow to "Active" and copy the webhook URL.
  5. Enter the webhook URL into the CRM: In the CRM, under „AI Search (n8n) → Settings", enter the webhook URL and the secret — done.

If you want to customize the workflow (your own data sources,
(Additional enrichment): the hit scheme in the callback must be preserved,
Everything in between can be expanded.

n8n workflow — the interface

The contract between CRM and n8n.

For those who want to recreate or adapt the workflow themselves, here is the full version.
Interface:

  • CRM → n8n (webhook trigger):
    POST to your webhook URL, header
    X-NZ-CRM-Secret:, Body
    { job_id, prompt, lang, max, category, do_scrape, callback_url, system_prompt, output_schema }.
    Since v1.33.0 system_prompt and output_schema
    Included — the n8n workflow does not need to maintain these rules itself.
  • n8n → CRM (Callback):
    POST /wp-json/nz_crm/v1/radar/ingest with the same secret header.
    body
    { job_id, hits: […], error? } — Scheme identical
    with the OpenAI direct output.

Auth runs via hash_equals() on the Secret — no OAuth, no
Signatures. This keeps the n8n workflow simple and allows you to use any signatures.
Integrate data sources without the CRM needing to know about them.

Watch the research live.

In Early Access, we'll show you the radar in a short demo on a real CRM —
Including job history and exam queue. Send us a short message about it.

Request early access

Back to CRM overview ·
Google integration ·
Personal approach ·
Individual landing pages