tools.digitalbridge.ie — a 50-tool applied AI platform

What problem was this solving?

Most firms claiming AI capability have slide decks. We wanted an estate we could point at, where the inference runs server-side, the costs are metered, and the scoring logic is ours rather than a wrapper around someone else's answer.

What did we build?

Fifty-plus shipped tools on one platform, including two proprietary scoring algorithms and production generative AI running server-side.

• AI Product Angle Generator — production generative AI turning a product into positioning angles.
• AI Batch Image Generator — high-volume generative imagery with queueing and cost control.
• Website Worth Analyzer, carrying the proprietary Website Worth Index scoring algorithm.
• Business Profile Authority Ranker, carrying the proprietary DigitalBridge Authority Score.
• A shared server-side inference layer so no model key is ever exposed to a browser.
• Crawl and enrichment pipeline combining Firecrawl retrieval with Gemini reasoning.

What does it integrate with?

The integration surface technical buyers ask about.

• Server-side model invocation with keys held in edge-function environment, never client-side.
• Firecrawl for structured retrieval, Gemini for reasoning over the retrieved corpus.
• Shared scoring library so both indices stay reproducible between runs.
• Rate and cost controls at the platform layer rather than per tool.

How does the system actually work?

The decision path in order. Client-confidential detail is deliberately excluded.

• 1. Input — A URL, a business profile or a product description, submitted by the user.
• 2. Retrieve — Firecrawl gathers the structured source material the score depends on.
• 3. Reason — Gemini interprets the corpus server-side; no key ever reaches the browser.
• 4. Score — A deterministic scoring library produces the Website Worth Index or Authority Score.
• 5. Return — A reproducible figure plus the reasoning behind it, not an opaque number.

What made it hard?

The real engineering constraints on this build.

• A score is only useful if it is reproducible — both indices are deterministic given the same inputs, which rules out free-form model output as the final figure.
• Generative image work is expensive at volume, so batching and queueing were design requirements, not optimisations.
• Public tools attract abuse; inference had to be gated without making the tools useless to genuine users.

What was the outcome?

The platform is live and public. It is also the single clearest answer to the question 'have you actually shipped AI in production' — a prospective client can use the tools themselves before speaking to us.

What capability does this demonstrate?

Engagement band: Applied AI engineering band.

• Two proprietary, named scoring algorithms
• Production generative AI (text and image) at volume
• Retrieval-augmented reasoning over crawled corpora
• Server-side inference with zero client key exposure

Are the scoring algorithms just a model prompt?

No. The model does the retrieval-side reasoning; the score itself is computed by a deterministic library so the same inputs always produce the same figure. A score that changes between runs is not a score.

Can we have a tool like this for our own business?

Yes — a customer-facing AI tool of this shape is one of the more common briefs we take. It sits in our applied AI engineering band, and you own the source code and the prompts at the end of it.

Where does the inference run?

Server-side, in EU-hosted edge functions, with provider training disabled. No model key is ever present in browser code.

A note on what is published here

Case studies on this page describe work Digital Bridge delivered. Client-confidential detail, credentials and internal data structures are deliberately excluded. Where a client has not authorised publication of measured figures, outcomes are described qualitatively rather than quantified — we do not publish numbers we cannot evidence. Last reviewed 2026-08-18.