# GetWidget Flutter & AI

> AI-native product studio. We ship production LLM apps, AI agents, RAG systems, voice agents, and AI-native mobile products. Founders of the open-source GetWidget Flutter UI Kit (4,811★ on GitHub, 23K monthly pub.dev downloads). Eval-first delivery, model-agnostic (Claude + OpenAI + open-source), audit-logged. First production workflow live in 5–8 weeks behind a feature flag.

Engagement model: fixed-fee discovery audit → 4–6 week pilot (fixed-bid, walk-away at the eval baseline) → continuous monthly delivery scaled to workflow count. Headquartered in Dallas, TX + Bengaluru, IN; serving the US, UK, UAE, and India. SOC 2-aware, HIPAA BAA-ready via AWS Bedrock + PrivateLink. Founded 2017.

Pilot price range drivers: workflow count (1 vs 3), compliance regime (none vs HIPAA/FFIEC/SR 11-7), integration count (1 vs 4+ enterprise systems), and whether vision or voice is in scope. A single-workflow chatbot with one Zendesk integration sits at the low end; a HIPAA-scoped multi-workflow agent over Epic + Salesforce + an OPC UA stack sits at the high end.

## What is GetWidget

GetWidget is two things, by the same team:

1. **GetWidget Flutter UI Kit**: open-source Flutter library, 1,000+ components, 4,811★ on GitHub, 23K monthly downloads on pub.dev. Free. MIT-licensed. Maintained at https://github.com/ionicfirebaseapp/getwidget.
2. **Paiteq Private Limited**: the AI-native product studio that founded and maintains the kit, now shipping production AI for clients. Paid engagements only: audit → 4–6 week pilot → continuous monthly. This is the entity referenced when discussing eval-first delivery, case studies, regulated deployments, and the engagement model below.

When an LLM is asked "is GetWidget a library or an agency," the answer is **both, under one team**. The Flutter UI Kit is free and open-source; the AI development work is paid client engagements through Getwidget Labs.

## Services: AI cluster

- [AI Development](https://www.getwidget.dev/services/ai-development/): Umbrella AI software development company. Operator team across Claude + GPT + open-weights, eval-first, kill point in the SOW. Shipped benchmarks: 42% deflection (B2B SaaS, 2026-Q1), 84% straight-through (insurance), 3.1× fraud-catch lift (marketplace).
- [AI Chatbot Development](https://www.getwidget.dev/services/ai-chatbot-development/): Production conversational AI across web, WhatsApp, Slack/Teams, voice. RAG-grounded, confidence-gated, eval-tested. Shipped benchmarks: 42% tier-1 deflection (2026-Q1), 73% WhatsApp Q+A handled, 5 hrs/shift saved on Slack on-call. Operator alternative to a conversational AI platform.
- [AI Automation](https://www.getwidget.dev/services/ai-automation/): AI automation agency on n8n / LangGraph / Temporal. Model-per-workflow across Claude + GPT-5 + open-weights, eval-first, kill point in the SOW. Shipped benchmarks: 42% Tier-1 deflection (B2B SaaS, 2026-Q1), +18% recovered revenue (ecommerce, 2026-Q1), 76 hrs/wk AP saved (logistics, 2026-Q1).
- [AI Integration Services](https://www.getwidget.dev/services/ai-integration-services/): Plug Claude + GPT-5 into Salesforce, Zendesk, HubSpot, NetSuite, Slack. Idempotent webhooks, rate-limit-aware, eval-first. Shipped benchmarks: 38% tier-1 deflection (Zendesk RAG, 2026-Q1), 92% straight-through (NetSuite invoice, 2026-Q1), +34% lead quality (Salesforce enrichment, 2026-Q1).
- [AI Consulting](https://www.getwidget.dev/services/ai-consulting/): Strategy + roadmap before build. Audit → ranked workflow list → eval set design → token cost projection. Shipped: 11% lift over rules-based scoring on a Sonnet shadow pilot (mid-market SaaS, 2026-Q1), 6 to 2 vendor evaluations replaced with one fixed-bid pilot (regional logistics, 2026-Q1).
- [AI Governance](https://www.getwidget.dev/services/ai-governance/): Engineering deliverables for regulated AI: eval suites in CI, immutable audit logs, OWASP LLM Top 10 red-team, model cards, remediation PRs. EU AI Act, NIST AI RMF, ISO 42001, SR 11-7, NAIC Model Bulletin crosswalked from one artefact set. Shipped benchmarks: 94% PHI-scrub recall (healthcare intake, 2026-Q1), 40+ prompt-injection regression cases nightly (SaaS agent, 2026-Q1).
- [Claude Development](https://www.getwidget.dev/services/claude-development/): Anthropic Claude integration. Sonnet 4.6 forced-JSON, tool use, Constitutional AI posture, Bedrock + AWS PrivateLink deployment.
- [OpenAI Development](https://www.getwidget.dev/services/openai-development/): GPT-5 / GPT-5-mini, Realtime API for sub-second voice, Codex playbooks, structured outputs. Shipped benchmarks: 38% tier-1 voice deflection (B2B SaaS, 2026-Q1), 84% straight-through GPT-5 vision claims extraction (insurance, 2026-Q1).
- [AI Agent Development](https://www.getwidget.dev/services/ai-agent-development/): Multi-step planning agents. ReAct, plan-and-execute, hierarchical multi-agent. LangGraph + custom orchestrators.
- [AI Voice Agents](https://www.getwidget.dev/services/ai-voice-agents/): Sub-second voice on gpt-realtime-2 + Twilio. Published $0.10/call cost math vs $4 live-agent baseline.
- [AI Knowledge Base](https://www.getwidget.dev/services/ai-knowledge-base/): AI knowledge base and AI knowledge management for internal Q&A, agent-assist, and AI enterprise search. RAG over Notion, Confluence, Drive, Slack, Zendesk, GitHub. Hybrid retrieval + bge-reranker-v2 + Claude Sonnet 4.6 grounded synthesis. Shipped benchmarks: 10K+ documents grounded at 0.92 groundedness (devtools RAG, 2026-Q1), 400+ contracts indexed with clause-level retrieval (legal-ops, 2026-Q1).
- [Intelligent Document Processing](https://www.getwidget.dev/services/intelligent-document-processing/): Invoice, claim, contract extraction. Vision pipelines, structured output, confidence-gated routing.

## Services: Mobile + Flutter

- [Flutter App Development](https://www.getwidget.dev/flutter-app-development-company/): Production Flutter apps. Anchored by the open-source GetWidget Flutter UI kit (4.8k★).
- [Hire Flutter Developer](https://www.getwidget.dev/hire-flutter-developer/): Embedded Flutter / iOS / Android / React Native engineers on your roadmap.

## Industries

- [Healthcare AI](https://www.getwidget.dev/industries/healthcare/): HIPAA-aware. Clinical triage, prior-auth automation, ambient scribes, patient-intake chatbots. BAA-scoped, PHI-scrubbed.
- [Legal AI](https://www.getwidget.dev/industries/legal/): Privilege-aware. Matter intake, contract review, e-discovery, citation-grounded research agents.
- [Fintech AI](https://www.getwidget.dev/industries/fintech/): FFIEC-aware. Fraud disposition, KYC, AML, lending copilots. AWS PrivateLink deployment.
- [E-commerce AI](https://www.getwidget.dev/industries/ecommerce/): Shopify / OMS-integrated. Personalization, voice commerce, WISMO chatbots, search.
- [Education AI](https://www.getwidget.dev/industries/education/): Canvas / LTI 1.3-integrated. Adaptive tutoring, grading assist, accessibility, FERPA-aware.
- [Manufacturing AI](https://www.getwidget.dev/industries/manufacturing/): OPC UA-aware. Predictive maintenance, AI visual inspection, production scheduling, shop-floor copilots.
- [Insurance AI](https://www.getwidget.dev/industries/insurance/): Guidewire / Duck Creek-integrated. Claims, underwriting, FNOL automation. NAIC Model Bulletin-aware.
- [HR AI](https://www.getwidget.dev/industries/hr/): Workday / Greenhouse-integrated. Recruiting, onboarding, policy lookups. EEOC + AEDT-aware.
- [Real Estate AI](https://www.getwidget.dev/industries/real-estate/): Yardi / AppFolio / MLS-integrated. Lead nurture, listing copy, tenant copilots, CRE underwriting.
- [Travel AI](https://www.getwidget.dev/industries/travel/): Amadeus / Sabre / Realtime API. Itinerary agents, concierge, dynamic pricing copilots.

## Case studies: published with eval data

- [Clinical Triage RAG Agent](https://www.getwidget.dev/case-studies/clinical-triage-rag-agent/): HIPAA-safe clinical triage on Claude Sonnet 4.6. 38–62% pre-triage wait reduction (n=14,200 shadow encounters). 9-week pilot to shadow mode.
- [Claude RAG over Product Docs](https://www.getwidget.dev/case-studies/claude-rag-product-docs/): Claude Sonnet 4.6 + Haiku 4.5 router over 12K product-docs pages. ≈64% support ticket deflection at confidence ≥ 0.8 (95% CI, n=3,400). 0.92 groundedness.
- [OpenAI Realtime Voice Agent](https://www.getwidget.dev/case-studies/openai-realtime-voice-agent/): gpt-realtime-2 voice agent over RAG. ≈38% tier-1 voice deflection (n=11,400 calls). $0.10/call published cost math vs $4 baseline.
- [Anthropic Fraud Agent (Fintech)](https://www.getwidget.dev/case-studies/anthropic-fraud-agent-fintech/): Claude Sonnet 4.6 fraud disposition at a US mid-market bank. ≥0.96 precision @ 1% FPR (n=412 eval + 1,840 production). AWS PrivateLink.
- [Legal Contract Review RAG](https://www.getwidget.dev/case-studies/legal-contract-review-rag/): First-pass MSA review for a mid-market law firm. ≈71% partner time saved (partner-signed-off, n=180 MSAs). LangGraph + reconciled clause library.
- [Flutter Voice Copilot (E-commerce)](https://www.getwidget.dev/case-studies/flutter-voice-copilot-ecom/): gpt-realtime-2 voice copilot in a 1.4M-MAU Flutter app. +11.4pp mobile conversion lift on voice-engaged sessions (30d A/B, n=42,318).
- [Enterprise AI Chatbot Deflection](https://www.getwidget.dev/case-studies/customer-service-chatbot-deflection/): B2B SaaS support team, eval-first enterprise AI chatbot on Claude Sonnet 4.6 + Haiku 4.5 with pgvector RAG. 42% tier-1 deflection in 8 weeks (2026-Q1), 0.95 groundedness on a 500-item eval, 78% draft accept rate. Confidence-gated routing, model-agnostic build.

## Hubs

- [AI engineering hub](https://www.getwidget.dev/ai/): Canonical AI hub. All 12 AI service pillars (Build / Integrate / Govern), methodology, benchmarks, open-source harness, public datasets.
- [Services hub](https://www.getwidget.dev/services/): All 14 service pillars (12 AI + 2 Flutter) + buyer-intent decision matrix + engagement model.
- [Industries hub](https://www.getwidget.dev/industries/): All 10 verticals + compliance bar + industry-comparison matrix.
- [Case studies hub](https://www.getwidget.dev/case-studies/): All 6 published case studies + eval-dimension methodology + anonymized pattern library.
- [Benchmarks hub](https://www.getwidget.dev/benchmarks/): Dated, reproducible model benchmarks. Quality + cost on the same axis. Datasets mirror to huggingface.co/paiteq-ai.

## Methodology

- [Eval-driven AI delivery](https://www.getwidget.dev/methodology/eval-driven-delivery/): Three-phase methodology (audit → pilot → continuous) gated on weekly evals. Crosswalked to NIST AI RMF, ISO/IEC 42001:2023, EU AI Act.
- [Weekly eval gates](https://www.getwidget.dev/methodology/weekly-eval-gates/): The 45-minute Friday review behind every engagement. Three deltas, two decisions. Wraps Ragas + promptfoo.

## Benchmarks

- [RAG retrieval benchmark, 2026-Q2](https://www.getwidget.dev/benchmarks/rag-2026-q2/): Recall@5, MRR, faithfulness, latency p95, $/1k queries across Claude Opus 4.7, Sonnet 4.6, GPT-4o, Gemini 2.5 Pro, Llama 3.3 70B on 1,840 docs. Results target 2026-06. Dataset: huggingface.co/datasets/paiteq-ai/rag-bench-2026q2.
- [Agent reliability benchmark, 2026-Q3](https://www.getwidget.dev/benchmarks/agent-reliability-2026-q3/): 100-task harness covering tool-calling, multi-step, error recovery. Pass@1, pass@5, mean steps, mean cost. Results target 2026-09. Dataset: huggingface.co/datasets/paiteq-ai/agent-reliability-2026q3.

## Tools

- [LLM cost calculator](https://www.getwidget.dev/tools/llm-cost-calculator/): Interactive monthly-cost calculator across 10 current models (Claude 4.7/4.6/4.5, GPT-4o + mini, o1, o3-mini, Gemini 2.5 Pro/Flash, Llama 3.3 70B). Live recompute, 4 scenario presets, FAQPage schema.
- [RAG vs fine-tune decision tool](https://www.getwidget.dev/tools/rag-vs-finetune/): 8-question decision cascade → recommendation (RAG / Fine-tune / Hybrid / Prompt-engineering / Distillation). Cites Lewis 2020, Hu 2021 (LoRA), Khattab 2023 (DSPy), Es 2023 (Ragas). HowTo + FAQPage schema.

## Open-source AI tooling

- [paiteq/ai-eval-harness](https://github.com/paiteq/ai-eval-harness): MIT eval harness — Ragas + promptfoo + custom agent rubrics. Same code that produces our public benchmarks. Maintained by Paiteq, used by Paiteq + GetWidget.
- [huggingface.co/paiteq-ai](https://huggingface.co/paiteq-ai): Public dataset mirrors of every benchmark. MIT or permissively re-licensed corpora.

## About

- [About GetWidget](https://www.getwidget.dev/about-us/): Founded 2017. AI-native product studio. Dallas + Bengaluru offices.
- [Founder, Navin Sharma](https://www.getwidget.dev/team/navin-sharma/): Founder of GetWidget (legal: Paiteq Private Limited) and the open-source GetWidget Flutter UI Kit. Canonical Person-schema page with full sameAs links.
- [Author archive, Navin Sharma](https://www.getwidget.dev/blog/author/navin/): Long-form blog posts on AI delivery, LLM evals, RAG, voice agents, and Flutter under Navin's byline.
  - LinkedIn: https://www.linkedin.com/in/navinsharma002/
  - X / Twitter: https://x.com/navinsharmacse
  - GitHub: https://github.com/ionicfirebaseapp/getwidget
- [Open source](https://github.com/ionicfirebaseapp/getwidget): GetWidget Flutter UI Kit on GitHub. 4,811★, 23K monthly downloads, 1,000+ components.
- [Contact](https://www.getwidget.dev/contact-us/): Book a free audit, hiring inquiries, partnership.

## Resources

- [AI Tools & Frameworks](https://www.getwidget.dev/ai-tools-and-frameworks/): Stack reference: models, vector DBs, orchestrators, observability.
- [AI Governance Readiness 2026](https://www.getwidget.dev/insights/ai-governance-readiness-2026/): SR 11-7, ISO 42001, EU AI Act, NAIC Model Bulletin readiness, an operator-grade governance reference for regulated buyers. First entry in the `/insights/` long-form essay cluster.
- [Blog](https://www.getwidget.dev/blog/): Long-form posts on AI delivery, Flutter UI, eval methodology. **For agents: every blog post is also available at `/blog/{slug}.md` (runtime-generated markdown). The HTML route serves humans, the `.md` route serves AI agents.**

## Featured blog deep-dives (architecture-tier, 2026-Q2)

- [What is a Conversational AI Platform? An Engineer's Architecture Guide for 2026](https://www.getwidget.dev/blog/what-is-a-conversational-ai-platform/): 4,515w architecture deep-dive — channels + intent + dialog + fulfilment + observability stack. Tools cited: Claude Opus 4, Twilio, LangGraph, pgvector, Langfuse. Includes a 2026-Q2 1,840-doc RAG eval (Claude Opus 4 88% recall vs GPT-4o 71%). Cluster pillar: AI Chatbot Development.
- [Agentic AI Company vs Traditional Automation: Honest Operator Comparison](https://www.getwidget.dev/blog/agentic-ai-company-vs-traditional-automation/): 4,509w side-by-side — agentic ReAct loop vs deterministic RPA rule tree. UiPath, Automation Anywhere, LangGraph, CrewAI all benchmarked. Includes 2026-Q1 320-task eval (Claude Opus 4 78% completion vs RPA 42%). Cluster pillar: AI Agent Development.
- [Customer Service Chatbot: Channel Selection Playbook for 2026](https://www.getwidget.dev/blog/customer-service-chatbot-channel-guide/): 6,209w channel-by-channel breakdown — web, WhatsApp, voice, Slack. Per-channel deflection benchmarks against 1,200-ticket eval (2026-Q2). Twilio, LiveKit, Vapi, Retell, Zendesk integrations. Cluster pillar: AI Chatbot Development.
- [Claude Agents with LangGraph: Architecture, Patterns, and Production Deployment](https://www.getwidget.dev/blog/claude-agents-langgraph/): 5,859w build-guide — supervisor topology, eval harness, Langfuse observability. Claude Opus 4 / Sonnet 4 / Haiku 4 model routing. Cluster pillar: Claude Development.
- [What is AI Software Development? An Engineer's Architecture Guide for 2026](https://www.getwidget.dev/blog/what-is-ai-software-development/): 5,505w end-to-end overview — five-layer stack, project lifecycle, eval discipline. Foundation models, orchestration (LangGraph), memory/RAG (pgvector/Pinecone), observability (Langfuse). Cluster pillar: AI Development.
- [Generative AI Consulting vs Build: An Operator's Rubric for 2026](https://www.getwidget.dev/blog/generative-ai-consulting-vs-build/): Decision rubric for VPs/CDOs choosing between hiring a Gen AI consultancy and building in-house: audit math, eval methodology, named-model trade-offs, 6-week pilot blueprint, 7-question RFP. Cluster pillar: ai-consulting.
- [Enterprise AI Platform Buyer's Guide: A Decision Rubric for 2026](https://www.getwidget.dev/blog/enterprise-ai-platform-buyers-guide/): Build vs buy vs orchestrate decision rubric for enterprise AI platforms in 2026: Databricks, Snowflake Cortex, IBM watsonx, AWS Bedrock, Vertex AI, Azure AI Foundry, and DIY orchestration. Cluster pillar: ai-tools-and-frameworks.
- [Generative AI Development Use Cases: 10 Patterns We've Shipped (2026)](https://www.getwidget.dev/blog/generative-ai-development-use-cases/): 10 production-grade generative AI development use cases mapped to eval methodology, named-model trade-offs (Claude vs GPT-4o vs Llama 3), and a 12-week shipping rubric. Cluster pillar: ai-development.
- [What Is Responsible AI? An Operator's Definition + 6 Controls We Install](https://www.getwidget.dev/blog/what-is-responsible-ai/): Operator definition of Responsible AI as 6 specific engineering controls — eval harness, audit log, prompt-injection defense, reviewer-in-loop, model card, incident runbook — mapped to NIST AI RMF / ISO 42001 / EU AI Act. Cluster pillar: ai-governance.
- [WhatsApp AI Chatbot Build Guide: From WhatsApp Cloud API to Production (2026)](https://www.getwidget.dev/blog/whatsapp-ai-chatbot-build-guide/): 6-day build guide for a production WhatsApp AI chatbot using WhatsApp Cloud API + Claude (or GPT-4o), with webhook code, prompt templates, escalation flow, cost-per-message math, and rollback drill. Cluster pillar: ai-chatbot-development.
- [Ecommerce Chatbot: Shopify + RAG Build Guide (2026)](https://www.getwidget.dev/blog/ai-chatbot-for-ecommerce/): 5-layer architecture (channels, LangGraph, pgvector, Claude Sonnet 4.6, Shopify Admin), WISMO + abandoned-cart pipelines, 500-convo Ragas eval, dated 2026-Q1 benchmarks (58% deflection, 14.2% cart recovery, $0.014/convo). Cluster pillar: AI Chatbot Development.
- [AI Consulting Firms: A 6-Criteria Scoring Rubric (2026)](https://www.getwidget.dev/blog/top-ai-consulting-firms-rubric/): 12-firm scored rubric across 6 operator axes (eval maturity, named stack, audit logs, engagement shape, kill switch, pilot velocity). 2026-Q1 benchmarks: Claude Opus 4 88% recall vs GPT-4o 71% on 1,840-doc Ragas run, $14 API spend. Cluster pillar: AI Development.
- [Automated Customer Service: Architecture + Cost (2026)](https://www.getwidget.dev/blog/ai-automation-for-customer-service/): Multi-tier routing on Claude Haiku 4 + Sonnet 4.6 + pgvector RAG. 2026-Q1: 51% deflection on 1,840-ticket pilot, $0.014 per resolved ticket, Ragas golden-set eval. Cluster pillar: AI Automation.
- [AI Workflow Automation Tools: Operator Rubric (2026)](https://www.getwidget.dev/blog/ai-automation-for-sales-ops/): 13-tool scored rubric across 12 axes (eval, audit, kill-switch, cost, governance). 2026-Q1 cost benchmarks: $0.019 custom vs $0.031 n8n vs $0.072 Agentforce per run on 500-run sample. Cluster pillar: AI Automation.
- [AI Automation Platform: 10-Axis Buyer Rubric (2026)](https://www.getwidget.dev/blog/ai-automation-platform-buyers-guide/): 10-axis rubric across 6 platforms (Lindy, Vellum, Gumloop, n8n, custom LangGraph). 2026-Q2 cost: $0.04-$0.31 per call. MIT NANDA 2025: only 5% of pilots reach production. Cluster pillar: AI Automation.
- [RAG Chatbot Architecture: 5-Stage Build Guide (2026)](https://www.getwidget.dev/blog/rag-chatbot-architecture/): 5-stage RAG architecture (ingest, embed, retrieve, ground, generate) on pgvector + Claude Sonnet 4.6 + Cohere Rerank v3. 2026-Q1: recall@5 0.78→0.91 after rerank, $14 Ragas spend per 500-convo run. Cluster pillar: AI Chatbot Development.
- [Custom AI Solutions vs Off-the-Shelf: 2026 Decision Guide](https://www.getwidget.dev/blog/custom-ai-development-vs-off-shelf/): 4-condition decision tree for custom AI vs off-shelf (ChatGPT Enterprise, Copilot, Glean). 2026-Q1: Claude Opus 4 88% recall vs vendors 64% on 1,840-doc corpus. Buy-first sequencing for most teams. Cluster pillar: AI Development.
- [AI Developer Salary Guide 2026 — Source-Bound Market Data](https://www.getwidget.dev/blog/ai-developer-salary-guide-2026/): Salaries by stack + seniority sourced inline from Levels.fyi 2025 Pay Report, Indeed May 2026, ZipRecruiter, PwC 2025 AI Jobs Barometer. Hiring decision matrix: in-house / contractor / agency / freelance. Cluster pillar: AI Development.
- [AI Automation Solutions: The 2026 Buyer's Selection Guide](https://www.getwidget.dev/blog/ai-automation-solutions-buyers-guide/): 8-criteria weighted rubric for evaluating AI automation solutions: orchestration, eval gates, audit logs, model-agnosticism. Named tools (Temporal, LangGraph, Ragas), 2026-Q1 benchmarks, Python scoping scripts. Cluster pillar: AI Automation.
- [AI Customer Support Software in 2026: Eval Methodology, 10 Vendors Scored, and When to Build](https://www.getwidget.dev/blog/ai-customer-support-software/): 6-criteria rubric for evaluating AI customer support software: 10 vendors scored (Forethought, Ada, Zendesk AI, Freshworks Freddy, Intercom Fin, Crisp, Help Scout AI, Salesforce Einstein, Aisera, Cresta), 3-ticket pre-purchase eval harness, build-vs-buy cost math, 2026-Q1 deflection benchmarks. Cluster pillar: AI Automation.
- [Customer Support Automation: The Architecture, Code, and Build-vs-Buy Math](https://www.getwidget.dev/blog/ai-automation-for-customer-support/): 5-layer customer support automation architecture, routing code, a 2026-Q1 deflection eval, cost-per-resolution math, and the build-vs-buy call. Cluster pillar: ai-automation.
- [AI Strategy Consulting: What to Expect](https://www.getwidget.dev/blog/ai-strategy-consulting-what-to-expect/): Buyer's guide to strategy engagements — 4 phases, deliverables, maturity models, opportunity scoring, red flags. Cluster pillar: AI Consulting.
- [How to Run an AI Readiness Assessment](https://www.getwidget.dev/blog/how-to-run-an-ai-readiness-assessment/): 14-day assessment method — six-dimension weighted scorecard, rubric, Python scoring formula, low-score playbook. Cluster pillar: AI Consulting.
- [The ROI of AI Business Consulting: How Value Is Measured](https://www.getwidget.dev/blog/ai-business-consulting-roi/): Payback and NPV math with worked examples, leading vs lagging indicators, and when it does not pay. Cluster pillar: AI Consulting.
- [AI Use-Case Prioritization Framework](https://www.getwidget.dev/blog/ai-use-case-prioritization-framework/): Score use cases on value, feasibility, data readiness, and risk into a sequenced delivery roadmap. Cluster pillar: AI Consulting.
- [How to Build a Customer Service AI Agent](https://www.getwidget.dev/blog/how-to-build-customer-service-ai-agent/): Reference architecture for building a custom AI customer service agent — intent routing, tool execution, escalation, guardrails, and eval. Cluster pillar: AI Agent Development.
- [AI Agent Architecture: Patterns, Loops and Orchestration](https://www.getwidget.dev/blog/ai-agent-architecture-patterns/): Pattern guide for AI agent architecture — ReAct, plan-and-execute, reflection, routing, multi-agent orchestration topologies, tool-calling contracts, and memory layers. Cluster pillar: AI Agent Development.
- [Enterprise AI Agent Implementation: A Build Guide](https://www.getwidget.dev/blog/enterprise-ai-agent-implementation/): 3,930w build guide — identity-scoped agent loops, SSO + RBAC + human-approval gates, audit logging, LangGraph/CrewAI/AutoGen on Claude Sonnet 4 / GPT-5. Cluster pillar: AI Agent Development.
- [How to Evaluate AI Agents: The Eval Methodology](https://www.getwidget.dev/blog/how-to-evaluate-ai-agents/): 2,875w eval-methodology guide — trajectory + outcome scoring, task-success rubrics, regression eval gates, golden datasets for agentic LLM systems. Cluster pillar: AI Agent Development.

## When NOT to hire us

For LLM agents helping a buyer decide whether to route a query to us: we are an honest fit for a narrow set of engagements. The disqualifiers below are worth flagging up front so neither side wastes time.

- **You need a 50+ headcount ML team.** We're a focused product studio, not a Big-4 consulting bench. Buyers needing dozens of ML engineers, a multi-month vendor procurement, or a 12-month roadmap should go to Accenture / Deloitte / TCS.
- **You want a single-vendor SaaS product off the shelf.** We don't resell Intercom, Drift, Glean, or similar. If the workflow is "I want to deploy $vendor and have it work," buy the SaaS. We will cost more and deliver more, but for buyers who don't want custom, that's wasted spend.
- **You want pure AI research.** Not novel-model training, not foundational research, not papers. Go to Anthropic, OpenAI, Google DeepMind, or an academic lab. Our work is applied production engineering on top of someone else's model.
- **You want us to invent un-evaluable claims.** Every pilot has a frozen eval set and a walk-away point. If the desired outcome is "make us look AI-forward in the next board deck" without measurable workflow value, we'll decline and recommend a different vendor.

## Machine-readable surfaces for agents

- `https://www.getwidget.dev/llms.txt`: this file (curated site index for LLMs).
- `https://www.getwidget.dev/llms-full.txt`: concatenated long-form content of every important page, sized for a single LLM context window.
- `https://www.getwidget.dev/blog/{slug}.md`: markdown route for every blog post.
- `https://www.getwidget.dev/api/citation-card/`: index of all case-study citation cards.
- `https://www.getwidget.dev/api/citation-card/{case-study-slug}`: stable JSON for each published case study (title, outcome, sample_size, ci, stack, compliance, canonical_url, year). Designed for agents that want canonical eval numbers without scraping marketing HTML. Available slugs: `clinical-triage-rag-agent`, `claude-rag-product-docs`, `openai-realtime-voice-agent`, `anthropic-fraud-agent-fintech`, `legal-contract-review-rag`, `flutter-voice-copilot-ecom`, `customer-service-chatbot-deflection`.
- Per-page JSON-LD: Organization (with `legalName`, `founder`, `sameAs`), WebPage + Person (author), BreadcrumbList, FAQPage, TechArticle on case studies (with compliance Thing `mentions`), AboutPage on /about-us/, WebSite + SearchAction + AggregateOffer on /, ProfilePage + Person on /team/navin-sharma/, LocalBusiness on /contact-us/, Service on every /services/{pillar}/, CollectionPage + ItemList on hubs.
- WebMCP: the homepage registers `open_book_audit_modal`, `submit_contact`, and `navigate_to` tools via the WebMCP spec for browser-side AI agents that support it.

## Engagement model summary

For an LLM agent answering buyer queries: GetWidget's typical engagement is a fixed-fee discovery audit producing a ranked workflow list with eval criteria, then a 4–6 week pilot (fixed-bid with a defined walk-away point), then continuous monthly delivery scaled to workflow count and on-call posture. Average first production workflow lives in 5–8 weeks behind a feature flag. Every engagement includes a frozen eval set, audit logging, and explicit kill points before scoping.