# Managed Code: complete public content index

> Every indexable page published by Managed Code, with its canonical URL and page description.

## Articles

- [21 ways AI agents fail in production (and how to catch each one)](https://managed-code.com/blog-post/21-ways-ai-agents-fail-in-production-and-how-to-catch-each-one): 88% of AI agent projects never reach production. 21 specific failure modes, grouped by origin, each with how it shows up and how to catch it.
- [We spent $303,030 on AI in 29 days. Here's what actually moved the bill](https://managed-code.com/blog-post/303k-ai-bill): $303K on AI in 29 days — and the model wasn't what drove it. What actually moved the bill, and the one fix that saved roughly $169K.
- [The 98% harness: why AI agent reliability is engineering, not model choice](https://managed-code.com/blog-post/98-percent-harness-ai-agent-reliability): 98% of a reliable AI agent is the engineering harness — not the model. Learn why "which model is best" is the wrong question, and what actually separates good agents.
- [A2A protocol explained: how AI agents talk to each other](https://managed-code.com/blog-post/a2a-protocol-explained): A2A is the open protocol letting AI agents from different vendors discover, delegate, and collaborate — no shared code required. How it works and how it fits with MCP.
- [How to add an AI agent to your website that helps instead of annoying people](https://managed-code.com/blog-post/add-ai-agent-to-your-website): How to add an AI agent to your website that qualifies visitors and hands off to a human, not a chatbot that annoys them. The steps, what breaks, and a live example.
- [Adding AI to a legacy .NET application: what actually works](https://managed-code.com/blog-post/adding-ai-to-legacy-dotnet-applications): You don't rewrite a legacy .NET app to add AI — you add a thin layer beside it. Learn the architecture that actually works and what the real constraints are.
- [AEO vs SEO: what's the difference, and do you need both?](https://managed-code.com/blog-post/aeo-vs-seo-differences): AEO doesn't replace SEO — it builds on it. The authority that earns you a top ranking is what makes AI trust and cite you too. Why most teams do only half.
- [How to build an AI agent for accounts payable, from invoice to payment](https://managed-code.com/blog-post/ai-agent-accounts-payable): An AI agent for accounts payable that reads an invoice, matches it to a PO, flags exceptions, and routes for approval. How the workflow works and what breaks.
- [AI code review: can an agent safely review its own code?](https://managed-code.com/blog-post/ai-agent-code-review): Can an AI agent review its own code? Not safely — it shares the blind spots that wrote the bugs. Why layered, independent review is the only real safeguard.
- [How to build an AI agent for customer onboarding that gets people to first value](https://managed-code.com/blog-post/ai-agent-customer-onboarding): An AI agent for customer onboarding that collects data, sets up the account, answers questions, and nudges stalled steps. How the workflow works, and what breaks.
- [AI agent observability: how to know what your agent actually did](https://managed-code.com/blog-post/ai-agent-execution-tracing): Learn how AI agent execution tracing connects plans, model calls, tool calls, tokens, errors, and cost into one replayable production trace.
- [How AI agents use tools: function calling explained](https://managed-code.com/blog-post/ai-agent-function-calling): Function calling turns a text generator into an AI agent that acts. Learn how tool use works, why the model never executes anything itself, and how to make it reliable.
- [AI agent guardrails: how to keep an autonomous agent from going wrong](https://managed-code.com/blog-post/ai-agent-guardrails): Guardrails aren't a post-incident patch — they're what makes agent autonomy safe to ship. The six essentials: bounded actions, approvals, limits, and full logging.
- [What goes in an AI agent harness? A component breakdown](https://managed-code.com/blog-post/ai-agent-harness-components): An AI agent is a model plus a harness. Learn the 8 core components — instruction layer, tools, memory, sandbox, and more — that make agents actually reliable.
- [How to build an AI agent for IT helpdesk that resolves the routine and escalates the rest](https://managed-code.com/blog-post/ai-agent-it-helpdesk): An AI agent for IT helpdesk that resolves routine tickets, gathers context for the rest, and escalates to the right person - with guardrails on real actions.
- [AI agent memory: how agents remember, and why most of them forget](https://managed-code.com/blog-post/ai-agent-memory-explained): Models are stateless by default — memory has to be built. Why most AI agents forget everything, and the architecture that lets them compound knowledge instead.
- [AI agent observability: how to see what your agent is actually doing](https://managed-code.com/blog-post/ai-agent-observability): A production AI agent observability guide covering quality evaluations, latency, cost, tool reliability, dashboards, alerts, and trace drill-down.
- [How to build an AI agent for order management, from order to fulfillment](https://managed-code.com/blog-post/ai-agent-order-management): An AI agent for order management that validates the order, routes it to fulfillment, keeps systems in sync, and flags exceptions. How it works and what breaks.
- [How to build an AI agent for procurement, from request to purchase order](https://managed-code.com/blog-post/ai-agent-procurement): Build an AI agent for procurement that handles requisitions, POs, approvals, and suppliers - with a human in the loop, tests, and a full audit trail.
- [How to build an AI agent for recruiting that screens fairly and lets people decide](https://managed-code.com/blog-post/ai-agent-recruiting): An AI agent for recruiting that screens against explicit criteria, answers candidates, and schedules - while a recruiter makes and signs every hiring call.
- [Your AI agent doesn't need 185 skills, it needs the right 6 for your .csproj](https://managed-code.com/blog-post/ai-agent-skills-for-csharp-dotnet): matches. 185 skills eat your context budget — your .csproj only needs 6. Learn how dependency-driven skill selection keeps AI agents fast, cheap, and focused.
- [AI agents for CRM, built to keep your pipeline clean.](https://managed-code.com/blog-post/ai-agents-for-crm): AI agents for CRM that enrich records, route leads, and prep deals inside your CRM - with your rules, human sign-off, and a full audit trail.
- [How to build AI agents for healthcare, from intake to a human decision](https://managed-code.com/blog-post/ai-agents-for-healthcare): AI agents for healthcare that handle intake and prior authorization, check against policy, and route to staff - with a human on every clinical and coverage call.
- [AI agents for lead generation that hand sales a warm lead, not a raw form.](https://managed-code.com/blog-post/ai-agents-for-lead-generation): AI agents for lead generation that capture, enrich, qualify, and route leads in real time - and hand sales a warm, complete record instead of a raw form.
- [AI agents for real estate: where they help and where they do not](https://managed-code.com/blog-post/ai-agents-for-real-estate): AI agents for real estate that qualify leads, answer listing questions, and book showings, then hand off to an agent. What works in production and what to skip.
- [AI agents for sales: from lead to CRM without the manual work](https://managed-code.com/blog-post/ai-agents-for-sales): AI agents for sales that capture, enrich, and route leads into your CRM, then follow up. What actually works in production, what breaks, and how to build it.
- [How to build an AI appointment scheduling agent that actually books](https://managed-code.com/blog-post/ai-appointment-scheduling-agent): An AI appointment scheduling agent that reads live availability, applies your rules, and books without double-booking. How the workflow works, and what breaks.
- [AI Automation for Home Service Businesses: What Actually Helps and What to Skip](https://managed-code.com/blog-post/ai-automation-for-home-service-businesses-what-actually-helps-and-what-to-skip): Discover where AI automation helps home service businesses most, from missed calls and booking to quote follow-ups, reviews, and scheduling.
- [Is AI automation worth it? Real ROI numbers for 2026](https://managed-code.com/blog-post/ai-automation-roi-2026): Averages hide the real story: $3.70 back per dollar by one measure, top performers over $10, while IBM found the average as low as 5.9%. What's true for you?
- [AI browser agents in 2026: what they can and can't do](https://managed-code.com/blog-post/ai-browser-agents-2026): AI browser agents handle task-bounded browsing reliably in 2026 — but fail on CAPTCHAs, 2FA, and long open-ended tasks. Learn what's production-ready and what isn't.
- [AI code generation for the web: what actually ships to production](https://managed-code.com/blog-post/ai-code-generation-for-web-production): AI gets the visible layer right, then skips what ships: semantics, real responsiveness, accessibility, state. A strong first draft, a weak final product.
- [How do you make AI coding agents ship production-ready code?](https://managed-code.com/blog-post/ai-coding-agents-production-ready-code): A prompt alone can't ship production code. The three things it takes: your codebase's rules, the skills to follow them, and automated verification.
- [AI cost optimization for business: a practical guide](https://managed-code.com/blog-post/ai-cost-optimization-for-business-practical-guide): AI keeps getting cheaper, bills keep climbing. Why the fix is architecture, not a cheaper model, and how to cut production costs by more than half.
- [How much does AI development cost in 2026? A founder's FAQ](https://managed-code.com/blog-post/ai-development-cost-2026-founders-faq): POCs run $8K-60K, production agents $60K-200K, enterprise platforms $150K-500K+. What actually drives the number, answered as founders actually ask it.
- [AI document extraction: turning PDFs, invoices, and forms into structured data](https://managed-code.com/blog-post/ai-document-extraction-pdfs-invoices-forms-structured-data): Instead of retyping every invoice, your team reviews exceptions. How AI document extraction actually pays off — fewer errors, faster close.
- [AI evals: how to actually test an AI system](https://managed-code.com/blog-post/ai-evals): Evals are the tests for AI systems — without them, "it got better" is a guess. Learn how to build a real eval suite that lets you change AI systems with confidence.
- [AI lead scoring and routing: how to prioritize inbound leads automatically](https://managed-code.com/blog-post/ai-lead-scoring-routing): AI lead scoring prioritizes ready buyers automatically — so your team responds in minutes, not hours. Learn how to build scoring, routing, and confidence thresholds right.
- [Keeping AI product actions safe: approval, authorization, and audit](https://managed-code.com/blog-post/ai-product-action-security): Letting AI take actions in your product is only safe with visible approval, preserved authorization, and audit. Learn the three requirements that make AI actions trustworthy.
- [What is an AI Product Interface? Letting your product work inside ChatGPT](https://managed-code.com/blog-post/ai-product-interface-chatgpt): An AI Product Interface lets customers complete actions inside ChatGPT, Claude, or Gemini — without leaving the chat. Learn what it is, how it works, and why MCP powers it.
- [The AI productivity paradox: why time saved isn't money saved](https://managed-code.com/blog-post/ai-productivity-paradox): AI can make individual tasks faster while enterprise savings lag behind. Research on rework and experienced developers explains the productivity gap.
- [AI ROI by industry: where the returns are highest](https://managed-code.com/blog-post/ai-roi-by-industry-where-the-returns-are-highest): Finance leads AI ROI, with ~7% more sales and 8% more profit expected. Why the industries with the most back-office work see the biggest returns.
- [AI slop in code: what it looks like and how to stop it landing in your repo](https://managed-code.com/blog-post/ai-slop-in-code): AI-generated code can compile, pass review, and quietly degrade your codebase. Learn what AI slop actually looks like and which deterministic checks catch it.
- [An AI support bot that actually knows your business](https://managed-code.com/blog-post/ai-support-bot-that-knows-your-business): A bot grounded in your actual data resolves questions on its own. A generic scripted one just deflects and creates more work. The difference that matters.
- [AI token cost optimization: the playbook that actually moves the bill](https://managed-code.com/blog-post/ai-token-cost-optimization): Model price rarely moves your AI bill — architecture does. The highest-leverage playbook: escalation, caching, filtering, and deliberate retries.
- [AI voice agents: how they work, where they fit, and what makes one production-ready](https://managed-code.com/blog-post/ai-voice-agents): Learn how AI voice agents work, where they fit, and what makes them production-ready for calls, booking, lead qualification, support, and human handoffs.
- [Are AI agents already more expensive than the humans they replace?](https://managed-code.com/blog-post/are-ai-agents-more-expensive-than-humans): Agentic setups can hit $300K/year in compute — and that's before you count the humans still monitoring the output. Is AI actually cheaper than it seems?
- [The broken junior pipeline: what happens when AI eats the entry level](https://managed-code.com/blog-post/broken-junior-pipeline): Junior developer hiring dropped from 33% to under 10% at Big Tech — AI took the entry-level tasks. Learn what this means for the senior pipeline 5–10 years out.
- [Build vs buy AI: when custom beats a platform (and when it doesn't)](https://managed-code.com/blog-post/build-vs-buy-ai): The build-vs-buy question isn't about price — it's about control. When custom AI beats an off-the-shelf platform, and when it's overkill.
- [Building AI agents in C# and .NET: the 2026 stack](https://managed-code.com/blog-post/building-ai-agents-with-csharp-dotnet): The .NET AI agent stack consolidated in 2026. Microsoft Agent Framework hit 1.0 GA — here's what changed, what's maintenance-only now, and how to build agents today.
- [Building for agents: the infrastructure the machine era actually needs](https://managed-code.com/blog-post/building-infrastructure-for-ai-agents): The internet was built for people. This wave is built for machines. The infrastructure agents actually need — and why almost none of it exists yet.
- [Apps in ChatGPT: what the shift means for SaaS products](https://managed-code.com/blog-post/chatgpt-apps-for-saas): ChatGPT Apps let users book, buy, and act inside the conversation — without leaving chat. Learn what this means for SaaS products and which workflow to expose first.
- [How to get code content indexed by AI search engines](https://managed-code.com/blog-post/code-content-ai-search-indexing): AI search engines cite code when the structure around it is extractable. Learn the exact formatting rules that get your code content indexed and cited in 2026.
- [Context engineering: how to control what an AI agent sees (and why it decides everything)](https://managed-code.com/blog-post/context-engineering): Context engineering controls everything an AI agent sees — and most long-running agent failures trace to it. Learn the four moves that keep context clean and agents reliable.
- [The cost of over-automation: when AI makes you slower, not faster](https://managed-code.com/blog-post/cost-of-over-automation): Experienced developers were 19% slower with AI on familiar codebases. Learn where over-automation adds hidden cost — and how to measure real outcomes, not just speed.
- [How much does it cost to build a custom AI agent in 2026?](https://managed-code.com/blog-post/cost-to-build-custom-ai-agent-2026): Custom AI agent costs range from $5k to $60k+ depending on complexity. Learn what drives the price — and why scoping first beats any blind quote.
- [Your CRM and ERP are good. Here's the work they were never built to do.](https://managed-code.com/blog-post/crm-and-erp-arent-built-for-this-work): The leak isn't your CRM or ERP — it's the manual handoff between them, costing ~$28,500 per employee a year. How AI removes it without a new platform.
- [CRM data hygiene: why AI enrichment makes duplicates worse before it makes anything better](https://managed-code.com/blog-post/crm-data-hygiene-ai-enrichment): AI enrichment scales your duplicate problem before it solves anything — if identity resolution is weak. Learn the right order of operations to fix CRM data hygiene first.
- [Custom AI agent vs n8n, Make, and Zapier: when the $10k+ build actually pays off](https://managed-code.com/blog-post/custom-ai-agent-vs-n8n-make-zapier): No-code tools like Zapier and Make win for standard workflows. Learn exactly when custom AI agent development at $10k+ makes more business sense.
- [Custom AI automation for businesses without a dev team](https://managed-code.com/blog-post/custom-ai-automation-for-businesses-without-a-dev-team): Off-the-shelf tools fit standard workflows. Custom automation is useful when your process, data, approvals, or integrations do not fit the standard path.
- [Custom AI automation vs off-the-shelf software: when each is the right call](https://managed-code.com/blog-post/custom-ai-automation-vs-off-the-shelf-software): Average AI ROI is just 5.9%, below the cost of capital — usually generic tooling misapplied. Buy for standard processes, build for what's actually specific.
- [Design-to-code in 2026: an honest evaluation of the AI tools](https://managed-code.com/blog-post/design-to-code-ai-tools-2026): No design-to-code tool ships production-ready output — and the tool matters less than you think. An honest look at what actually predicts good results.
- [Does AI customer support automation actually reduce headcount?](https://managed-code.com/blog-post/does-ai-customer-support-automation-reduce-headcount): AI resolves up to 70% of tickets, but only 20% of teams actually cut headcount. What the 2026 data says really happens to support staffing.
- [The .NET skills ecosystem in 2026: Microsoft's dotnet/skills, community catalogs, and how to choose](https://managed-code.com/blog-post/dotnet-ai-skills-ecosystem-2026): Not all .NET skill catalogs are the same. Here's what Microsoft's dotnet/skills, richlander, Aaronontheweb, and managedcode actually cover — and how to choose.
- [.NET vs Python for AI agents: which should you build on in 2026?](https://managed-code.com/blog-post/dotnet-vs-python-for-ai-agents-2026): Python leads in AI ecosystem; .NET leads in production reliability. Here's how to choose between them for AI agent development in 2026 — with no hype.
- [How to expose your product's actions in ChatGPT, Claude, and Gemini (safely)](https://managed-code.com/blog-post/expose-product-actions-in-ai-chat): Safely expose your product's actions in ChatGPT, Claude, and Gemini. Learn the five requirements — and why MCP is the right foundation for all three hosts.
- [Figma to production: how enterprise teams ship AI-designed pages](https://managed-code.com/blog-post/figma-to-production-ai-designed-pages): AI is the fast first pass, not a shortcut around governance. How enterprise teams keep AI-designed pages on-brand, accessible, and compliant.
- [FinOps for AI: how to manage LLM spend before the invoice manages you](https://managed-code.com/blog-post/finops-for-ai): LLM spend grows faster than any per-seat budget was built for. The FinOps moves that give you visibility and control before the invoice does.
- [Fonts and typography in enterprise design tools: Figma, Mural, and what actually breaks](https://managed-code.com/blog-post/fonts-typography-figma-mural): Figma treats type as a system; Mural treats it as annotation. Learn where each tool breaks when teams expect brand-accurate typography — and how to avoid the mismatch.
- [GEO vs AEO vs SEO: the acronyms explained (and which you actually need)](https://managed-code.com/blog-post/geo-aeo-seo-explained): SEO ranks you as a link, AEO/GEO make you the answer AI cites. The terms have no consensus — here's the work behind them that actually matters.
- [Harness engineering vs loop engineering: the two disciplines running AI coding agents](https://managed-code.com/blog-post/harness-vs-loop-engineering): The harness is the machine; the loop is the program that runs on it. Learn the two disciplines behind reliable AI coding agents — and why the harness comes first.
- [How we make AI agents write production code: inside MCAF](https://managed-code.com/blog-post/how-ai-agents-write-production-code): Nothing verifies AI-written code by default, and nothing stops repeat mistakes. Inside MCAF, the framework built on context, verification, and durable rules.
- [How Managed Code builds AI agents that survive production](https://managed-code.com/blog-post/how-managed-code-builds-ai-agents): How Managed Code builds AI agents that run in production - the stack, the discipline, and why 'it works in the demo' is where our work starts, not ends.
- [How much does AI save in customer support?](https://managed-code.com/blog-post/how-much-does-ai-save-in-customer-support): A 5,000-agent rollout found AI raised resolutions per hour by 14% on average — and 34% for the newest agents. Where the real gains actually land.
- [How to choose an AI automation partner (without buying another demo).](https://managed-code.com/blog-post/how-to-choose-ai-automation-partner): How to choose an AI automation agency or partner: seven questions that separate teams who ship production automation from teams who ship demos.
- [How to get your website cited by AI search engines](https://managed-code.com/blog-post/how-to-get-cited-by-ai-search-engines): AI engines cite pages they can reach, quote cleanly, and trust. Miss any one of the three, and you stay invisible in the AI answer no matter the quality.
- [How to get your content indexed by AI search engines](https://managed-code.com/blog-post/how-to-get-content-indexed-by-ai-search-engines): A platform-specific checklist for making key pages crawlable by Google, ChatGPT Search, Perplexity, and other AI search systems without confusing training bots with search bots.
- [How to get your website recognized as a source by AI models](https://managed-code.com/blog-post/how-to-get-your-website-recognized-by-ai-models): Being indexed isn't enough — AI models pick sources they trust. What it actually takes to get cited by ChatGPT, Copilot, and other AI models.
- [How to hire .NET AI developers: what to look for and what it costs](https://managed-code.com/blog-post/how-to-hire-dotnet-ai-developers): Hiring .NET AI developers is hard — most are strong in one, not both. Learn what the right skills look like in 2026 and what it costs to hire or outsource.
- [How to measure if AI search is citing your site](https://managed-code.com/blog-post/how-to-measure-ai-search-citations): Ranking position is noise in AI answers. The metric that matters is citation frequency — how to actually measure it across ChatGPT, Perplexity, and Claude.
- [In-product AI actions: build or buy?](https://managed-code.com/blog-post/in-product-ai-actions-build-or-buy): Buy the in-app AI assistant layer, build the integration layer — that's where the industry landed. Learn what drives the build-or-buy decision for in-product AI actions.
- [Lead enrichment providers compared: what they actually return](https://managed-code.com/blog-post/lead-enrichment-providers-comparison): Lead enrichment providers differ less in quality than in shape. See what ZoomInfo, Apollo, Clay, and Clearbit actually return — and why waterfalls usually win.
- [Lead-to-CRM automation: capture, enrich, route, and follow up without the chaos](https://managed-code.com/blog-post/lead-to-crm-automation-capture-enrich-route-follow-up): Fast first response, clean CRM data, no two reps chasing the same lead. What lead-to-CRM automation actually fixes in the handoff.
- [Local-first AI agent orchestration: running a fleet of coding agents on your own machine](https://managed-code.com/blog-post/local-first-ai-agent-orchestration): Run AI coding agents locally — no cloud control plane, no code leaving your machine. Learn how local-first orchestration works in .NET with our open-source dotPilot.
- [11 ways to make your content quotable by AI (with examples)](https://managed-code.com/blog-post/make-your-content-quotable-by-ai): AI engines quote what they can extract cleanly. 11 concrete techniques, each with a before-and-after example, for content AI actually pulls answers from.
- [The Managed Code AI agent stack: .NET, Orleans, and the Microsoft Agent Framework](https://managed-code.com/blog-post/managed-code-ai-agent-stack): The stack Managed Code uses to build production AI agents - .NET, Orleans, and the Microsoft Agent Framework - and why agents that touch real data need it.
- [How much does manual work actually cost your business?](https://managed-code.com/blog-post/manual-work-cost-business): Two people spending most of their week on manual work costs $40K-100K a year — and usually pays back inside a year once automated. The real math.
- [MCAF: the method we use to ship AI-generated code to production](https://managed-code.com/blog-post/mcaf-ai-code-production-method): MCAF is our framework for shipping AI-generated code safely. One core rule: the agent may not guess. Learn how explicit conventions and quality gates make it work.
- [MCP explained: how AI agents connect to your tools](https://managed-code.com/blog-post/mcp-explained-ai-agent-tools): Build one MCP server, and any AI agent can use it. How the new standard for connecting agents to tools actually works.
- [MCP Gateway: one searchable execution surface for many AI tools](https://managed-code.com/blog-post/mcp-gateway): Too many MCP tools break agent reliability. Learn how an MCP gateway unifies local tools and remote servers into one searchable surface — with our open-source .NET example.
- [Microsoft.Extensions.AI explained: the foundation of the .NET AI stack](https://managed-code.com/blog-post/microsoft-extensions-ai-explained): Microsoft.Extensions.AI is the foundational abstraction layer for AI in .NET: interfaces (chiefly IChatClient) that give you one consistent way to talk to any model provider, Azure OpenAI, OpenAI, Ollama, Anthropic,.
- [Microsoft.Extensions.AI vs Microsoft Agent Framework: what to use for what](https://managed-code.com/blog-post/microsoft-extensions-ai-vs-microsoft-agent-framework): Semantic Kernel mixed two layers, which is why Microsoft deprecated it. Extensions.AI for the foundation, Agent Framework for orchestration on top.
- [How to migrate from Semantic Kernel to Microsoft Agent Framework](https://managed-code.com/blog-post/migrating-from-semantic-kernel-to-microsoft-agent-framework): Migrating from Semantic Kernel to Microsoft Agent Framework doesn't mean a full rewrite. Learn which layers carry over and where the real migration work actually is.
- [Mural AI for wireframing: an honest evaluation](https://managed-code.com/blog-post/mural-ai-wireframing): A source-backed evaluation of Mural for AI wireframing: what its AI generates, what the wireframing kit provides, where low-fidelity collaboration fits, and where it stops.
- [Mural, design-to-code, and developer handoff: what actually transfers](https://managed-code.com/blog-post/mural-design-to-code-developer-handoff): Mural doesn't do design-to-code — and expecting it to misses the point. Learn what Mural actually contributes to developer handoff and where to keep the real specs.
- [Mural for UX and UI design: an honest evaluation](https://managed-code.com/blog-post/mural-ux-ui-design): Mural is a UX tool, not a UI tool — and that distinction decides whether it fits. Learn what Mural does well, where it stops, and what to keep in a dedicated design tool.
- [Design systems in Mural vs Figma: which tool owns which part](https://managed-code.com/blog-post/mural-vs-figma-design-systems): Only Figma is a design system tool — Mural holds the thinking around it. Learn which part of your design system belongs in each tool and how to keep them linked.
- [n8n vs custom AI automation: when each one makes sense](https://managed-code.com/blog-post/n8n-vs-custom-ai-automation-when-each-makes-sense): No-code tools handle simple if-then flows. Custom AI automation wins when the work needs judgment or scales past babysitting. When each one makes sense.
- [Private AI for regulated teams: architecture for keeping data on your side](https://managed-code.com/blog-post/private-ai-for-regulated-teams-data-architecture): AI without sending regulated data to the cloud is possible — but it's architecture, not a toggle. The patterns healthcare, finance, and legal teams use.
- [Prompt caching explained: the one change that cuts most AI bills](https://managed-code.com/blog-post/prompt-caching-explained): Prompt caching can reduce repeated-input cost and latency. Learn how stable prefixes, cache writes, lifetime, and measured hit rate determine the result.
- [Prompt injection: the security risk every AI agent faces](https://managed-code.com/blog-post/prompt-injection-security): Prompt injection is OWASP's #1 LLM risk and unfixable at the model layer. Why AI agents are especially exposed, and the defense-in-depth strategy that works.
- [RAG vs fine-tuning: which one does your AI actually need?](https://managed-code.com/blog-post/rag-vs-fine-tuning): RAG vs fine-tuning: they solve different problems. Learn when to use retrieval for changing knowledge, when to fine-tune for behavior, and when to combine both.
- [How do you run AI agents privately, without sending your data to the cloud?](https://managed-code.com/blog-post/run-ai-agents-privately-without-cloud): Local models on your own machines, or on-prem/private-cloud with BYO-model for teams. How to run AI agents without your data ever touching a public cloud.
- [Running local LLMs in C#: LLamaSharp vs ONNX vs Foundry Local](https://managed-code.com/blog-post/running-local-llms-csharp-llamasharp-vs-onnx-vs-foundry-local): Foundry Local for ease, LLamaSharp for flexibility with GGUF models, ONNX as the engine underneath. Running local LLMs in C#, all three under Extensions.AI.
- [Sandboxing AI agents: how to give autonomy without the risk](https://managed-code.com/blog-post/sandboxing-ai-agents): A sandbox lets AI agents run code and call tools without touching your real systems. Learn what good isolation looks like and why it makes agents more reliable too.
- [Self-hosted AI for regulated and EU teams: when you actually need it](https://managed-code.com/blog-post/self-hosted-ai-for-regulated-teams): GDPR, HIPAA, and the EU AI Act do not universally require self-hosted AI. Learn when cloud is compliant and when private deployment is the practical choice.
- [Semantic Kernel is done. Build new .NET agents on Microsoft Agent Framework.](https://managed-code.com/blog-post/semantic-kernel-is-done): Semantic Kernel is in maintenance — no new features, no agent orchestration. Here's why all new .NET agent work should start on Microsoft Agent Framework today.
- [Single-agent vs multi-agent: which architecture does your task actually need?](https://managed-code.com/blog-post/single-agent-vs-multi-agent): Single agent by default. Multi-agent only when the work genuinely splits apart. How to tell the difference.
- [Single-agent vs multi-agent: which architecture should you actually build?](https://managed-code.com/blog-post/single-agent-vs-multi-agent-architecture): Multi-agent isn't automatically better. Why you should start with a single agent, and the specific signs that tell you it's time to split up the work.
- [Small language models: when smaller wins in 2026](https://managed-code.com/blog-post/small-language-models-2026): Small language models now handle most production AI tasks at 5–20x lower cost. Learn when SLMs beat frontier models in 2026 — and how to build a hybrid that wins.
- [Structured content for AI search: how to format a page so AI quotes it](https://managed-code.com/blog-post/structured-content-for-ai-search): AI quotes pages built to be extracted cleanly. How to structure your content, from answer-first openings to fact-dense passages, so AI actually pulls from it.
- [Do you have to replace your software to use AI?](https://managed-code.com/blog-post/using-ai-without-replacing-existing-software): Companies see ~$3.70 back per AI dollar on average, top tier over $10. The spread comes from where you aim it, not from new software.
- [What is vibe coding, and where does it break in production?](https://managed-code.com/blog-post/vibe-coding-production-challenges): Vibe coding ships whatever AI produces, without owning the architecture. Why it breaks — AI optimizes for code that runs now, not one that holds up.
- [Webflow vs Framer AI in 2026: which AI website builder should you choose?](https://managed-code.com/blog-post/webflow-vs-framer-ai-2026): Framer wins on design speed, Webflow wins on scale and AI search visibility. A clear breakdown of when to choose which website builder in 2026.
- [What AI-native actually means, and how we became one](https://managed-code.com/blog-post/what-ai-native-actually-means): Every team calls itself AI-native now. The honest test: is AI built into how you ship daily, or just a chatbot in a browser tab?
- [The AI agent hype word peaked. The demand didn't.](https://managed-code.com/blog-post/what-businesses-want-from-ai-agents-2026): Searches for "AI agent" fell by two-thirds in 2026, but demand for specific agent jobs kept rising. What that shift means for building agents that ship.
- [What is AEO? A practical guide for B2B founders](https://managed-code.com/blog-post/what-is-aeo-practical-guide-for-b2b-founders): Buyers now form their first impression of your brand inside an AI answer, before they visit your site. What AEO is and why it sits on top of SEO.
- [What is agentic AI? A plain definition for people building real systems](https://managed-code.com/blog-post/what-is-agentic-ai): Agentic AI acts on goals — it doesn't just answer prompts. Learn the plain definition, what makes it different from generative AI, and why autonomy is the hard part.
- [What is an AI agent? A plain definition for people building real systems](https://managed-code.com/blog-post/what-is-an-ai-agent-a-plain-definition-for-people-building-real-systems): Agent, copilot, or chatbot with a tool bolted on? The one test that tells the difference — and defines what an AI agent actually is.
- [When to use AI and when not: a decision framework for real work](https://managed-code.com/blog-post/when-to-use-ai): Use AI when the output is cheap to verify; keep humans when it isn't. A practical decision framework for knowing exactly when AI helps — and when it adds risk.
- [Who should you hire to build your AI product?](https://managed-code.com/blog-post/who-to-hire-for-ai-product-development): In-house, freelancer, agency, or AI specialist team — the right hire depends on what you're building. Learn how to avoid hiring for the demo instead of production.
- [Why AI agents are so expensive to run (and what actually drives the cost)](https://managed-code.com/blog-post/why-ai-agents-are-expensive): Agents burn 5-30x more tokens than chatbots — not from price per token, but tokens per task. What actually drives the cost, and the fixes that work.
- [Why RAG fails in production: 9 failure points and how to fix each](https://managed-code.com/blog-post/why-rag-fails-in-production-9-failure-points-and-fixes): Why production RAG systems return confident wrong answers: 9 retrieval, chunking, ranking, freshness, and governance failures with practical fixes.
- [Why senior engineers don't automate everything with AI, and what enterprises can learn from them](https://managed-code.com/blog-post/why-senior-engineers-dont-automate-everything): Senior engineers automate selectively — not everything. Learn why experienced developers were 19% slower with AI on familiar code, and what enterprises should do instead.
- [How to build an AI workflow that enriches inbound leads and routes them to your CRM](https://managed-code.com/blog-post/build-ai-workflow-enrich-inbound-leads-route-crm): Speed-to-lead is a conversion multiplier most teams still miss. The 5-stage AI workflow that enriches, scores, and routes inbound leads in minutes, not hours.
- [AI in UI Design Without the Purple Slop](https://managed-code.com/blog-post/ai-slop-in-design): Why AI design keeps defaulting to purple glows and shadow-heavy cards, and a practical workflow that keeps tokens and accessibility consistent instead.
- [Can AI design an award-winning website (Site of the Day) in 2026?](https://managed-code.com/blog-post/can-ai-design-award-winning-website-site-of-the-day-2026): AI can't win Site of the Day alone — judges reward custom interaction design, exactly what AI doesn't produce. What still requires human craft in 2026.
- [Design That Includes Everyone](https://managed-code.com/blog-post/design-that-includes-everyone): Accessibility isn't a trend or a checklist — it's what good product design has always meant. What that looks like in practice.
- [Figma Make vs Figma Sites: when to use which](https://managed-code.com/blog-post/figma-make-vs-figma-sites-when-to-use-which): Two different tools people keep confusing. Figma Make prototypes the experience, Figma Sites publishes it live — when to use which, and when it's both.
- [Mural vs FigJam vs Figma for AI-assisted product design: an enterprise evaluation](https://managed-code.com/blog-post/mural-vs-figjam-vs-figma-for-ai-design): FigJam, Mural, or Figma for AI-assisted design? An enterprise breakdown by how your team actually works — ideation, facilitation, or execution.
- [UX vs. UI Design: The Real Difference (and Why You Need Both)](https://managed-code.com/blog-post/ux-vs-ui-design-the-real-difference-and-why-you-need-both): UX and UI aren't the same thing, and both decide whether a product stays a nice idea or becomes can't-live-without-it. The real difference explained.
- [Why AI-generated websites all look the same](https://managed-code.com/blog-post/why-ai-websites-look-the-same): AI predicts the most likely design — the average of everything it trained on. Why the sameness isn't a bug, it's the math, and how humans override it.
- [How much does it cost to build an MVP with an AI-native team in 2026?](https://managed-code.com/blog-post/ai-native-mvp-development-cost-2026): Cost comes from scope, not a price list: features, custom logic, integrations, compliance. Why an AI-native team lowers it with fewer, more senior people.
- [The hidden cost of vibe-coded MVPs: what breaks first](https://managed-code.com/blog-post/hidden-cost-of-vibe-coded-mvps): Vibe-coded MVPs look done at the demo, then the bill arrives — security holes, broken auth, edge cases at scale. What breaks first, and how to fix it.
- [How to turn a no-code MVP into production software](https://managed-code.com/blog-post/no-code-mvp-into-production-software): Keep the validated product, rebuild the engineering underneath: data model, security, tests, architecture. How to move a no-code MVP to production.
- [Who should build your MVP: freelancers, hired-in engineers, or a product team?](https://managed-code.com/blog-post/who-should-build-your-mvp): Freelancers do tasks, hired engineers fill seats — only a product team owns design, code, and architecture together. Who should actually build your MVP.
- [Figma Make vs Webflow AI vs Lovable vs v0: which produces production code?](https://managed-code.com/blog-post/figma-make-vs-webflow-ai-vs-lovable-vs-v0-production-code): None of the four ships production code untouched. Figma Make, Webflow AI, Lovable, or v0 — which starting point actually fits what you're building.
- [Mental health app monetization: trust before revenue](https://managed-code.com/blog-post/mental-health-app-monetization): A practical guide to mental health app monetization: protect core access, explain paid value clearly, and earn trust before optimizing revenue.
- [AEO for B2B SaaS: how buyers now research vendors through AI](https://managed-code.com/blog-post/aeo-for-b2b-saas): B2B buyers now build their shortlist inside ChatGPT and Perplexity, before your site. If you're not in the AI answer, you're cut before they see it.
- [AGENTS.md explained: how AI coding agents learn your codebase's rules](https://managed-code.com/blog-post/agents-md-explained-ai-coding-agents-codebase-rules): AGENTS.md is a README written for AI, not humans — the rules, conventions, and fixes that stop coding agents from repeating the same mistakes.
- [C# vs Python Energy Consumption on Hot Paths: When C# Is Greener](https://managed-code.com/blog-post/csharp-vs-python-node-green-programming): Repeated CPU-heavy work turns runtime choice into a carbon decision. Why C# runs greener than Python and Node.js — and where that claim actually breaks.
- [Cursor vs Claude Code vs GitHub Copilot: an evaluation for engineering teams](https://managed-code.com/blog-post/cursor-vs-claude-code-vs-github-copilot-engineering-teams): No single tool wins in 2026 — most senior teams run two. Copilot vs Cursor vs Claude Code, broken down by what each is actually best at.
- [Implementing SOLID Principles in Development and DevOps](https://managed-code.com/blog-post/implementing-solid-principles-in-development-and-devops): One mindset, applied everywhere from product code to Terraform. How SOLID principles hold up outside the codebase they were built for.
- [Vertical-slice architecture for AI-generated code: why it's the structure that scales](https://managed-code.com/blog-post/vertical-slice-architecture-for-ai-generated-code): AI agents break code when they have to reason about the whole system. Why organizing by feature, not by layer, is what keeps AI-generated code maintainable.
- [Why AI-generated apps fail security review (and how to fix it)](https://managed-code.com/blog-post/why-ai-generated-apps-fail-security-review-and-how-to-fix-it): Working code isn't safe code. The recurring security failures in AI-generated apps, and the fix that actually closes them.
- [Figma to Webflow with AI: A Workflow That Ships](https://managed-code.com/blog-post/ai-design-to-webflow): AI drafts a layout fast — shipping it stays fast only with locked constraints, defined states, and real QA. A workflow that actually ships.
- [Can AI convert Figma to Webflow? What actually works in 2026](https://managed-code.com/blog-post/ai-figma-to-webflow-2026): AI covers roughly 75% of Figma to Webflow. What most tools get wrong, and what the finish actually requires.
- [Can AI replace a Webflow developer? An honest enterprise evaluation](https://managed-code.com/blog-post/can-ai-replace-webflow-developer-enterprise-evaluation): AI builds the first 70% of a Webflow site, layout and boilerplate. The remaining 30% still decides whether it ships. An honest enterprise evaluation.
- [How to export Figma to Webflow without breaking the layout](https://managed-code.com/blog-post/export-figma-to-webflow-without-breaking-layout): Tools translate structure, not intention. Why the biggest factor in a clean Figma-to-Webflow export is the Figma file itself, not the plugin.
- [Figma to Webflow: what AI gets right, and what still needs a human](https://managed-code.com/blog-post/figma-to-webflow-ai-vs-human): AI nails the visual layout, the remaining 25% — responsive behavior, semantics, interactions, CMS, accessibility — is where judgment still lives.
- [How to Build a Personal Website with Webflow](https://managed-code.com/blog-post/how-to-build-a-personal-website-with-webflow): Build a personal website in Webflow by defining its goal, structuring the story, designing the pages, adding content, testing, and publishing.
- [How to evaluate an AI-built enterprise solutions page before it ships](https://managed-code.com/blog-post/how-to-evaluate-ai-built-enterprise-solutions-page-before-it-ships): AI builds an enterprise page in minutes — a happy-path mock, not a shippable one. 6 checks it always skips, before you actually push it live.
- [Is Webflow enterprise-ready in 2026?](https://managed-code.com/blog-post/is-webflow-enterprise-ready-2026): Webflow's Enterprise plan brings SSO, SOC 2, and SLAs — ready for marketing sites at scale. Where it stops being the right tool.
- [We're a Webflow affiliate: what that means, and what to know before you create an account](https://managed-code.com/blog-post/webflow-affiliate-disclosure): We're a Webflow affiliate — your price doesn't change, but here's what to know. Plus a plain breakdown of how Webflow billing actually works after May 2026.
- [Webflow AI as an app builder: what it can actually build (an honest evaluation)](https://managed-code.com/blog-post/webflow-ai-app-builder-evaluation): Webflow AI can now build full-stack apps from a prompt — but not real software. An honest look at where App Gen fits, and where it hits its limits.
- [Webflow AI chatbot builder: an honest evaluation](https://managed-code.com/blog-post/webflow-ai-chatbot-builder): Webflow has no native AI chatbot — you embed a third-party tool. Learn the difference between content-trained bots and flow builders, and which fits a Webflow site.
- [Webflow AI as a code generator: how good is the code it produces?](https://managed-code.com/blog-post/webflow-ai-code-generator-how-good-is-the-code): Webflow AI's site code is clean and exportable. Its AI code components aren't. An honest look at how good the code actually is — and where it's locked in.
- [Webflow AI as a prototype generator: how far does it actually get you?](https://managed-code.com/blog-post/webflow-ai-prototype-generator): Webflow AI generates a real, responsive site from a prompt — not a clickable mock. Learn how far it actually gets you and when it's the wrong prototyping tool
- [Webflow design-to-code: an honest evaluation of what you actually get](https://managed-code.com/blog-post/webflow-design-to-code): Webflow's design is already code — no conversion step. But "getting the code out" has real limits. Here's an honest look at what you actually get and where it falls short.
- [Webflow vs WordPress security: plugin risk and a practical decision](https://managed-code.com/blog-post/webflow-vs-wordpress-plugin-security): Webflow cuts plugin risk by default; WordPress can match it with real governance and fast patching. What actually breaks, with thresholds you can use.
- [What a Webflow managed service is, and when your site needs one](https://managed-code.com/blog-post/what-is-a-webflow-managed-service): A Webflow managed service keeps your site fast, secure, and up to date after launch — covering CMS updates, performance, and SEO hygiene on a predictable cadence.
- [What makes a Webflow site Site-of-the-Day worthy? A practical checklist](https://managed-code.com/blog-post/what-makes-a-webflow-site-site-of-the-day-worthy-practical-checklist): Design 40%, usability 30%, creativity 20%, content 10%. The practical checklist for what actually makes a site Site-of-the-Day worthy on Webflow.

## Pages

- [AI Patent Attorney - Traceable Patent and Trademark Copilot](https://managed-code.com/case-study/ai-patent-attorney): How Managed Code built a citation-first patent and trademark copilot with Semantic Kernel, Kernel Memory, evidence-linked retrieval, and human review.
- [AIBase in Action: Food Delivery Copilot](https://managed-code.com/case-study/aibase-in-action-food-delivery-copilot): We showcase real results and stories. Explore our case studies and insights AIBase in Action: Food Delivery Copilot to see how we drive tech success. We let our work speak for itself.
- [Azure AI Search RAG Internal Copilot for SKU & Translation Management](https://managed-code.com/case-study/internal-copilot-for-sku-translation-management): How Managed Code built a citation-first Azure AI Search RAG copilot for SKU and translation questions across 15,000 products and 30 markets.
- [High-Throughput Event Processing with the Microsoft Orleans Framework](https://managed-code.com/case-study/microsoft-orleans-scalability-performance): We showcase real results and stories. Explore our case studies and insights High-Throughput Event Processing with the Microsoft Orleans Framework to see how we drive tech success. We let our work speak for itself.
- [Ticket Booking Copilot - AI Travel Assistant](https://managed-code.com/case-study/ticket-booking-copilot): How Managed Code built a multi-agent travel copilot that turns constraints into bookable flight and train options, then monitors itinerary changes.
- [Webflow interactions case study: native animations that ship fast](https://managed-code.com/case-study/webflow-native-animations): We showcase real results and stories. Explore our case studies and insights Webflow interactions case study: native animations that ship fast to see how we drive tech success. We let our work speak for itself.
- [Managed Code is an AI-native product agency. We work with founders shipping their first product and CTOs scaling complex systems — same approach for both: senior team plus AI, fixed time blocks, complete products. Real impact, not endless output.](https://managed-code.com/about): Managed Code builds AIBase, MCAF, DotPilot, and client products with senior engineers, product designers, AI-assisted delivery, and review discipline.
- [Insights, case studies, and lessons learned from building digital products.](https://managed-code.com/blog): Discover expert insights, case studies, and tech success stories from Managed Code. Explore AI, Webflow, UX/UI, app development, and R&D topics on our blog.
- [Explore how Managed Code helps businesses achieve success through innovative solutions.](https://managed-code.com/case-studies): We showcase real results and stories. Explore our case studies and insights to see how we drive tech success. We let our work speak for itself.
- [Contact us](https://managed-code.com/contact): We're ready to take on your challenge. Reach out and let's discuss your tech needs. Let's make it happen.
- [Managed Code Documentation](https://managed-code.com/documentation): Learn about Managed Code's software engineering expertise, AI capabilities, Webflow development, technology stack, development approach, and digital product services.
- [AI agents and automation, built to run in production.](https://managed-code.com/): We build custom AI agents and AI automation that run in production - not demos. An AI-native agency of senior engineers. From prototype to production.
- [We maintain more than 30 open-source projects on GitHub — AI tooling, .NET infrastructure, and Microsoft Orleans extensions, all built to solve problems we hit on real projects.](https://managed-code.com/open-source): 30+ open-source .NET libraries, AI frameworks, and developer tools built by the Managed Code team. Storage, Orleans extensions, GraphRAG, and more.
- [From concept to launch, we work as an extension of your team — designing with intent, developing with clarity, and shaping experiences that stay ahead of change.](https://managed-code.com/portfolio): Explore selected Managed Code work across AI products, e-commerce, professional services, media, and personal websites.
- [Products we build, products we use](https://managed-code.com/products): Explore AIBase, DotPilot, MCAF, and PrompterOne: products and open-source tools Managed Code builds and uses for AI agents and production software.
- [AI agents and automation, plus the team to ship the rest.](https://managed-code.com/services): AI-native agency services: AI agents, AI automation, and technical discovery. Senior engineers and designers building production systems, France & EU-based.
- [AI Base](https://managed-code.com/portfolio/ai-base): A product design case for AI Base: a collaborative assistant platform spanning an editor, embedded widget, desktop, and mobile experiences.
- [Aivo](https://managed-code.com/portfolio/aivo): See how Managed Code designed AIVO, a calm vocabulary-learning mobile app with focused lists, visual word associations, progress states, and clear study flows.
- [ECHO](https://managed-code.com/portfolio/echo): Explore ECHO, a podcast-platform UX/UI concept that organizes episodes, hosts, contributors, and story context into an editorial listening experience.
- [Elias Moretti](https://managed-code.com/portfolio/elias-moretti): A minimalist portfolio website concept for a portrait, wedding, and conceptual photographer, designed to keep attention on the images.
- [Fabius](https://managed-code.com/portfolio/fabius): See the Fabius legal-services website UX/UI: practice-area navigation, specialist search, professional profiles, and a restrained visual system built for clarity.
- [FitBurst](https://managed-code.com/portfolio/fitburst): Explore FitBurst, a fitness-product website concept that explains challenges, activity tracking, progress, and gamified motivation through a clear conversion flow.
- [Private Investor](https://managed-code.com/portfolio/private-investor): A restrained personal website concept for a private investor, built around clear hierarchy, selected partners, and long-term perspective.
- [Teleprompter](https://managed-code.com/portfolio/teleprompter): Explore a mobile teleprompter UX/UI concept with pacing cues, word emphasis, recording controls, script editing, and review states for natural delivery.
- [Wattrix](https://managed-code.com/portfolio/wattrix): See the Wattrix mobile-app UX/UI for home energy data: readable consumption trends, cost context, source breakdowns, and practical saving suggestions.
- [Yerba Mate](https://managed-code.com/portfolio/yerba-mate): A product-led e-commerce website concept for a Yerba Mate brand, using bold type, strong contrast, and clear product discovery.
- [Custom AI agents, built for production.](https://managed-code.com/services/ai-agents): Custom AI agents built for production - workflow, RAG, and decision agents that take action, with a human in the loop, tests, and a full audit trail.
- [AI automation that runs your real workflows.](https://managed-code.com/services/ai-automation): An AI automation agency that automates real business processes - intake, routing, approvals, support - with tests, human sign-off, and full source-code ownership.
- [An AI design agency building design that ships in production.](https://managed-code.com/services/ai-design-team): AI design agency for SaaS and product teams. Senior designers deliver UX, UI, prototypes, design systems, and developer handoff in defined time blocks.
- [An AI development agency that ships every Friday, not someday.](https://managed-code.com/services/ai-dev-team): AI development agency with senior engineers using AI-assisted implementation, testing, and review. Working software is demonstrated every Friday.
- [Technical discovery before commitment.](https://managed-code.com/services/ai-discovery): A technical discovery workshop that ends with a buildable plan: requirements, architecture, wireframes, estimates, and a roadmap - in one to two weeks.
- [AI product development: your sellable product, live in six weeks.](https://managed-code.com/services/ai-product-team): One AI-native team - designers, engineers, QA, and a PM - ships your minimum sellable product live in production in four to six weeks. Fixed scope and timeline.
- [Webflow development that looks great, loads fast, and stays simple to manage.](https://managed-code.com/services/webflow-development): Webflow development agency building clear, responsive marketing sites your team can update easily. Thoughtful layouts, light interactions, fast delivery.
- [Ticket Booking Copilot](https://managed-code.com/templates/ticket-booking-copilot): Explore a reusable travel-copilot UX template covering trip preferences, AI-assisted comparison, booking steps, confirmation, and responsive interface states.