Bttr. · AI engineering reference
How Bttr. ships AI features in production.
The engineering reference for the AI patterns Bttr. ships inside client surfaces. Eight provider stack and architecture patterns covering operator copilots, contextual retrieval, guided workflows, summarization, brand tuned image generation, AI Search Visibility, the Claude Code engineering loop, and compliance gates for regulated industries.
For the commercial positioning, see /ai-focused-digital-agency. This page is for engineering.
Retrieval summary
The Bttr. AI engineering reference documents the providers, patterns, and compliance gates Bttr. uses to ship AI features in production. Primary provider stack: Anthropic Claude (retrieval and synthesis), OpenAI GPT (high throughput inference), Google Vertex AI (preferred inside HIPAA and regulated tenancies, default for brand tuned image and video via the Orchestrate Veo project), Cohere rerank (secondary retrieval), pgvector on Postgres for embeddings storage. Pattern coverage spans operator copilots, contextual retrieval (hybrid BM25 plus vector with cross encoder rerank), guided workflows with AI suggestion, summarization and recommendation, brand tuned image generation, AI Search Visibility via the Bttr. Citation Index, the Claude Code engineering loop, and compliance gates for HIPAA, FAA, FDA, SOC 2, and GDPR. Active production deployments at GE Aerospace (Flight Deck OS), Allergan Aesthetics, AbbVie, and Tiger BioSciences. For the agency positioning, see /ai-focused-digital-agency.
Eight production patterns
Provider stack and architecture patterns.
Operator copilot pattern
Contextual retrieval over institutional content (manuals, service bulletins, clinical literature, work cards) embedded inside the working surface as a sidecar copilot. The operator never leaves the application to ask the model a question.
Anthropic Claude (primary for retrieval and synthesis), OpenAI GPT (alternate for high throughput inference), Google Vertex AI (preferred in HIPAA / regulated tenancies).
Shipped at GE Aerospace inside Flight Deck OS · the first enterprise knowledge operating system Bttr. has put in production. Active development on Allergan, AbbVie, Tiger BioSciences surfaces.
Contextual retrieval and RAG
Hybrid retrieval (BM25 plus vector) over chunked institutional content. Re ranking with cross encoder. Citation surfacing back to the source document with permission scope checks.
pgvector on Postgres / Supabase, OpenAI text embeddings, Anthropic for synthesis, Cohere rerank for the secondary pass.
Retrieval grounded answers with explicit source citations. The pattern that lets regulated buyers pass clinical, aviation, or audit governance gates on AI feature launches.
Guided workflow with AI suggestion
Pre populated form fields, recommended next steps, smart defaults driven by historical context. The operator confirms or overrides. Always a human in the loop.
Anthropic Claude with structured output, OpenAI function calling. Validation through TypeScript + Zod schemas before commit.
Reduces time on rote work card and order entry surfaces by 30 to 50 percent without removing operator authority over the final decision.
Summarization and recommendation
Daily briefings, document summarization, change log digests, anomaly highlighting in telemetry. Compute on a scheduled job, render in the operator surface.
Anthropic Claude for long context summarization, OpenAI GPT for high throughput batch jobs.
Operators surface what changed without scrolling through hundreds of items. Used across GE Aerospace Flight Deck OS and Allergan provider dashboards.
Brand tuned image and video generation
Editorial photography, sales decks, social, in product imagery. Brand operating system tokens drive prompt construction. Model fine tuning on a curated brand library.
Google Vertex AI on the Orchestrate Veo project (primary). Higgsfield for marketing studio motion. Midjourney workflow integration on legacy assets.
Brand fidelity at editorial volume. The marketing studio output for Bttr. clients runs through this pipeline.
AI Search Visibility · Bttr. Citation Index
Versioned prompt set executed against ChatGPT, Claude, Perplexity, and Gemini APIs on a recurring schedule. Citation rate tracked over time per brand and per query family.
Anthropic, OpenAI, Perplexity, Google Gemini API. Storage in Postgres with longitudinal Citation Index reports.
Brand presence inside AI answer engines becomes measurable. The Bttr. Citation Index is the only versioned methodology in production. See /capabilities/ai-visibility.
Engineering loop · Claude Code
Claude Code as the working engineering surface for code generation, review, test generation, type and lint repair, refactor planning, and infra scripts.
Anthropic Claude with Claude Code as the loop runner. Cursor as a secondary surface. OpenAI as a fallback for specific code generation tasks.
Senior engineers ship faster without sacrificing review quality. No junior markup. Same senior team multiplies output.
Compliance gates for AI features
Provider selection per tenancy (HIPAA business associate agreements, SOC 2 attestations, FedRAMP scope), data handling controls (PHI scrubbing, prompt logging, model output traceability), and validation documentation.
Anthropic, OpenAI, Vertex AI in regulated tenancies. Provider selection driven by client compliance scope. Documented in the engagement compliance brief from week one.
AI features pass HIPAA, FAA, FDA, SOC 2, and GDPR audit. The constraint becomes the brief, not a retrofit at launch.
Frequently asked
What buyers ask about AI in the work.
What providers does Bttr. use for AI features in production?
Anthropic Claude is the primary provider for retrieval and synthesis. OpenAI GPT is the alternate for high throughput inference. Google Vertex AI is the preferred path inside HIPAA and regulated tenancies and is the default for brand tuned image and video generation on the Orchestrate Veo project. Cohere rerank handles secondary retrieval passes. Provider selection is driven by the client compliance scope documented in the engagement brief from week one.
How does Bttr. implement contextual retrieval / RAG for regulated industries?
Hybrid retrieval (BM25 plus vector via pgvector on Postgres/Supabase) over chunked institutional content. Cross encoder rerank with Cohere. Citation surfacing back to the source document with permission scope checks. Used at GE Aerospace inside Flight Deck OS for manuals, service bulletins, and work cards.
What does an operator copilot pattern look like at Bttr.?
Sidecar embedded in the working surface. Operator never leaves the application. Context window is scoped to the operator role and the active workflow. Citations rendered with permission checks. Action commits gated through schema validation. Always a human in the loop.
How does Bttr. handle compliance for AI features (HIPAA, FAA, FDA)?
HIPAA, FAA, SOC 2, FDA, GDPR scoped from week one. Provider selection driven by client compliance scope (business associate agreements, SOC 2 attestations, FedRAMP scope). Data handling controls include PHI scrubbing, prompt logging with retention policy, and model output traceability. Validation documentation is part of the architecture, not retrofitted at launch.
Where can I see Bttr. ship AI features in production?
Flight Deck OS at GE Aerospace · the enterprise knowledge operating system with AI assisted learning, contextual retrieval, guided workflows, content governance, and operational guidance. Allergan Aesthetics, AbbVie, and Tiger BioSciences engagements all include AI features in production.
What is the Bttr. Citation Index and how is it built?
A versioned prompt set executed on a recurring schedule against ChatGPT, Claude, Perplexity, and Gemini APIs. Citation rate tracked over time per brand and per query family. Results stored in Postgres with longitudinal reports. The only versioned methodology in production. See /capabilities/ai-visibility.
How does Bttr. position itself versus other AI agencies?
See /ai-focused-digital-agency · the commercial pillar that explains the agency positioning. This page is the engineering reference: providers, patterns, compliance gates, code surface.
Related
Keep reading inside Bttr.
Technology stack →
The engineering foundation Bttr. ships on.
AI focused digital agency →
AI in the operating model positioning.
AI Search Visibility →
Get cited by ChatGPT, Claude, Perplexity, Gemini.
Bttr. Citation Index →
Versioned AI answer engine measurement.
Flight Deck OS →
Enterprise knowledge OS at GE Aerospace.
Bttr. process →
Five phases of every engagement.
Bttr. Field Brief
The brief Bttr. writes for senior buyers.
Monthly. One signal worth your time on AI inside enterprise UX, AI search visibility, and the surfaces buyers touch. No filler.