Shopify Magic configuration, Sidekick + Inbox AI setup, and custom large-language-model integrations. OpenAI, Anthropic Claude, Google Gemini — wired into Shopify catalogs, checkout, search, and customer service. AI-driven product recommendations, conversational commerce, and generative content. UK-registered, remote-first, fixed-price.
We build on Shopify Magic, Sidekick, and Inbox AI as a foundation — then layer custom OpenAI, Anthropic, and Gemini integrations on top via Functions, custom apps, or Hydrogen. The native tools first, the custom layer second. Not the other way round.
Every AI integration we ship is observable, retry-aware, GDPR-compliant, and cost-bounded. We track token spend per request, redact PII before any model call, and fall back gracefully when the LLM is slow or down. AI in production is a different problem from AI in a notebook.
We don't add AI because it's the trend. We add AI when an A/B test proves the AI variant lifts conversion or AOV vs the baseline. Shade finders, ingredient discovery, AI search, generative PDP copy — only ship the ones that earn their token budget.
Magic covers product descriptions, blog posts, theme content, and basic image editing — and for stores under ~500 SKUs, that is often enough. Above that, or where you need PDP personalisation, AI search, or conversational commerce, you graduate to custom OpenAI / Anthropic integrations layered on top of Magic. We audit which side of that line you sit on before quoting.
All three, depending on the job. OpenAI for generative content and structured output. Anthropic Claude for long-context tasks (catalog-wide reasoning, large product feeds, complex CS triage) and where safety / refusal behaviour matters. Gemini for multimodal (image + text) and where Google's pricing wins. We pick by fit, not by hype, and we benchmark on your data before committing.
We redact personally identifiable information before any external model call by default. Customer names, emails, addresses, payment info — never sent to a third-party LLM. Where the use case requires personal context (e.g. customer service triage), we run a local redaction layer or use the model provider's zero-data-retention enterprise tier. Every AI integration ships with a DPA, a data-flow diagram, and a token-cost cap.
Different jobs. LimeSpot / Rebuy / Nosto are excellent at collaborative-filtering recommendations from purchase history. They cap at the patterns in your existing data. AI personalisation (custom LLM-driven) is better at cold-start, semantic matching, and explaining why a product was recommended — which converts better in some categories (beauty, fashion, B2B). We often run both: rules-based recs as the default, LLM recs as the high-value upsell layer.
Yes — and Plus is where most of the meaningful AI integrations live. Plus B2B benefits from AI-driven account-specific catalogs and pricing rules; Hydrogen unlocks custom AI-powered UX (shade finders, AR-augmented try-on, AI configurators) that OS 2.0 cannot host. We have shipped AI integrations on all three architectures.
Depends on the use case. AI product descriptions on a 4,000-SKU store: usually pays back in 2-4 weeks via lifted PDP CR. Conversational commerce: 3-6 months depending on AOV. AI search: 1-3 months for stores doing >£100k/month. We build a token-cost model into every quote so you can see the run-rate cost vs the projected uplift before you commit.
Yes. Public AI apps for Shopify (using Polaris admin UI, App Bridge, and OpenAI / Anthropic on the backend) are a defined service. We have shipped private AI apps for individual brands and one public app currently in App Store review. Custom AI app development starts at £8,999 with a 4–10 week build.
Shopify Magic configuration: same-week start. Sidekick / Inbox AI setup: same-week. Custom OpenAI / Anthropic builds: 1–2 weeks lead time for spec and architecture, then 4–6 weeks build. We do not run multi-month discovery phases — the spec is part of week 1.
Send your store URL and the AI use case you have in mind — generative content, AI search, conversational commerce, custom personalisation. We come back inside two business days with a fixed scope, a model recommendation (OpenAI / Anthropic / Gemini), and a token-cost projection.
