# Yardwork

> Yardwork designs and builds what a company's team is missing, around the way it already works, and stays on after launch. The two current offers are managed AI agents and custom software development.

Yardwork is a service business, based in Berlin, for founders, owners and lean teams in the US and Europe that need custom work built around their existing tools. English pages are written for readers in the US and Europe; German pages for readers in Germany, Austria and Switzerland. Engagements cover scope, design and build, launch, and ongoing monitoring, maintenance and support. An agreed handover is also available. Framework choice and agent controls belong to the managed-agents offer; they are not requirements for every software project.

Last reconciled: 2026-10-27. Inventory: 89 bilingual blog pairs, three bilingual case studies, two primary offers and ten city pages. Each content entry below has English and German canonical links.

## Primary pages

- [Home](https://yardwork.dev/en): Custom solutions built around how a company works, with managed AI agents and custom software development as the two current offers. [Deutsch: Startseite](https://yardwork.dev/de).
- [Services](https://yardwork.dev/en/services): The two offers and the engagement: scope and price, build, launch and ongoing support. An agreed handover is also possible. [Deutsch: Leistungen](https://yardwork.dev/de/services).
- [Managed AI agents](https://yardwork.dev/en/services/managed-agents): AI agents built into a company's existing tools and kept running after launch. Sales, analytics, content and custom jobs, with access and review boundaries agreed before launch. [Deutsch: Managed KI-Agenten](https://yardwork.dev/de/services/managed-agents).
- [Custom software development](https://yardwork.dev/en/services/software-development): Customer-facing web apps, internal tools and AI-powered products, from scope and design to launch, hosting, fixes and new features. [Deutsch: Individuelle Softwareentwicklung](https://yardwork.dev/de/services/software-development).
- [Sales agents](https://yardwork.dev/en/services/sales): Agents for prospect research, outreach drafts, follow-up and CRM work. Client review before external sends is part of the current service design. [Deutsch: Vertriebsagenten](https://yardwork.dev/de/services/sales).
- [Analytics agents](https://yardwork.dev/en/services/analytics): Agents that read and reconcile business sources into answers and reports, with source references and no changes to live accounts. [Deutsch: Analytics-Agenten](https://yardwork.dev/de/services/analytics).
- [Organic search agents](https://yardwork.dev/en/services/content): Research, page decisions, drafting, client review, publishing and follow-up across search and AI citations. [Deutsch: Agenten für organische Suche](https://yardwork.dev/de/services/content).
- [Pricing](https://yardwork.dev/en/pricing): Implementation fees and ongoing operating costs, with scope and price agreed before work starts. The current starting point is published on this pricing page. [Deutsch: Preise](https://yardwork.dev/de/pricing).
- [About](https://yardwork.dev/en/about): How Yardwork closes the gap between knowing what a company needs and building systems that work, plus founder Ves Ivanov's background. [Deutsch: Über uns](https://yardwork.dev/de/about).
- [FAQ](https://yardwork.dev/en/faq): Buyer questions about both offers, delivery, connected tools, agent controls, ownership, scoped pricing and ongoing support, with a brief handover option. [Deutsch: Häufige Fragen](https://yardwork.dev/de/faq).
- [Writing](https://yardwork.dev/en/blog): Notes on AI and software, with 89 bilingual article pairs about choosing, building and operating systems for business work. [Deutsch: Artikel](https://yardwork.dev/de/blog).
- [Case studies](https://yardwork.dev/en/case-studies): Three projects: a cycling coaching platform, an agentic sales system and multi-agent business operations. Project facts belong to each case study. [Deutsch: Fallstudien](https://yardwork.dev/de/case-studies).
- [Locations](https://yardwork.dev/en/locations): Ten city pages for companies across Europe. Yardwork is based in Berlin; Berlin can include an in-person first meeting, with remote delivery elsewhere. [Deutsch: Standorte](https://yardwork.dev/de/locations).
- [Contact](https://yardwork.dev/en/contact): A 30-minute discovery call with founder Ves Ivanov, booked through the Cal.com embed, plus direct email contact. [Deutsch: Kontakt](https://yardwork.dev/de/contact).
- [Open-source tools](https://yardwork.dev/en/tools): A focused tools index, currently featuring AgentBridge, a local browser bridge with approval before agent actions. [Deutsch: Open-Source-Werkzeuge](https://yardwork.dev/de/tools).
- [AgentBridge](https://yardwork.dev/en/tools/agentbridge): A free browser extension and local bridge for approval of agent requests. Approved page and clipboard data goes to the connected agent; that agent's further processing depends on its setup. [Deutsch: AgentBridge](https://yardwork.dev/de/tools/agentbridge).
- [Privacy policy](https://yardwork.dev/en/privacy): Website data processing through Vercel, PostHog, Cal.com booking, Google Calendar on invitation-only appointment pages, and direct enquiries. [Deutsch: Datenschutzerklärung](https://yardwork.dev/de/privacy).
- [Imprint](https://yardwork.dev/en/imprint): Yardwork provider and contact information. [Deutsch: Impressum](https://yardwork.dev/de/imprint).

<!-- ai-tools:start -->
## Data: AI tools and your data

Yardwork's dataset of what 320 AI and work tools (338 plans) do with business data under EU rules: whether each trains on your data, where it stores it, whether it signs a DPA, its sub-processors, DPF status, retention, certifications, admin controls and price in Germany. Official vendor sources only; every answer quoted, dated and checked twice. Checked 2026-10-16 to 2026-10-20. Licence CC BY 4.0.

- [Data](https://yardwork.dev/en/data): Data Yardwork collects from official sources and keeps up to date. Every answer links to its source and check date, free to reuse with credit. [Deutsch: Daten](https://yardwork.dev/de/data).
- [Which AI tools are GDPR compliant? 320 tools checked](https://yardwork.dev/en/data/ai-tools): Which AI tools train on your data, keep it in the EU and include a DPA? 320 tools, 338 plans, from the vendors' own pages, checked twice. [Deutsch: Welche KI-Tools sind DSGVO-konform? 320 geprüft](https://yardwork.dev/de/data/ai-tools).
- [How we check what AI tools do with your data](https://yardwork.dev/en/data/ai-tools/method): Where the AI tools dataset comes from: official vendor sources only, a second check of every answer, dated and quoted. Statuses, corrections and licence. [Deutsch: So prüfen wir, was KI-Tools mit Ihren Daten machen](https://yardwork.dev/de/data/ai-tools/method).
- [Does AI train on your data? 81 of 338 plans do](https://yardwork.dev/en/data/ai-tools/no-ai-training): 81 of 338 AI and work tool plans train on your data unless someone switches it off. See which ones, and how to turn training off in each. [Deutsch: Trainiert KI mit Ihren Daten? 81 von 338 Tarifen](https://yardwork.dev/de/data/ai-tools/no-ai-training).
- [EU-hosted AI tools: 99 of 338 plans qualify](https://yardwork.dev/en/data/ai-tools/eu-hosting): EU-hosted AI, plan by plan: 99 of 338 plans keep your data in the EU by default or as an option. Many others only on enterprise plans. With sources. [Deutsch: KI mit Servern in der EU: 99 von 338 Tarifen](https://yardwork.dev/de/data/ai-tools/eu-hosting).
- [AI data processing agreements: which tools sign one?](https://yardwork.dev/en/data/ai-tools/dpa): 255 of 338 plans include a data processing agreement, most in their terms. Personal plans usually offer none. Which ones, and how to get each. [Deutsch: Auftragsverarbeitungsvertrag für KI-Tools](https://yardwork.dev/de/data/ai-tools/dpa).
- [AI sub-processors list for 338 plans](https://yardwork.dev/en/data/ai-tools/sub-processors): Which sub-processors handle your data behind each AI tool? 198 of 338 plans publish a list. Counts, links and which vendors keep theirs private. [Deutsch: Unterauftragsverarbeiter von KI-Tools: 338 Tarife](https://yardwork.dev/de/data/ai-tools/sub-processors).
- [EU–US Data Privacy Framework list: AI vendors](https://yardwork.dev/en/data/ai-tools/data-privacy-framework): Which AI and work tool vendors are on the official EU–US Data Privacy Framework list? 221 of 338 plans qualify. Checked against the list itself. [Deutsch: EU-US Data Privacy Framework Liste: KI-Anbieter](https://yardwork.dev/de/data/ai-tools/data-privacy-framework).
- [Assistants: AI training, EU hosting and DPA compared](https://yardwork.dev/en/data/ai-tools/assistants): ChatGPT, Claude, Copilot, Gemini, Mistral and others AI assistants: which train on your data, keep it in the EU and include a DPA, plan by plan. [Deutsch: Assistenten: KI-Training, EU-Hosting und AVV im Vergleich](https://yardwork.dev/de/data/ai-tools/assistants).
- [Writing and email: AI training, EU hosting and DPA compared](https://yardwork.dev/en/data/ai-tools/writing-and-email): AI writing and email tools such as Grammarly, DeepL and LanguageTool: which train on your text, keep it in the EU and include a DPA. [Deutsch: Schreiben und E-Mail: KI-Training, EU-Hosting und AVV im Vergleich](https://yardwork.dev/de/data/ai-tools/writing-and-email).
- [Meetings and notes: AI training, EU hosting and DPA compared](https://yardwork.dev/en/data/ai-tools/meetings-and-notes): AI note takers such as Otter, Fireflies, tl;dv and Fathom: which train on your calls, keep recordings in the EU and include a DPA. [Deutsch: Meetings und Notizen: KI-Training, EU-Hosting und AVV im Vergleich](https://yardwork.dev/de/data/ai-tools/meetings-and-notes).
- [Phone and voice: AI training, EU hosting and DPA compared](https://yardwork.dev/en/data/ai-tools/phone-and-voice): AI phone agents and voice AI such as Vapi, Retell, Synthflow and fonio: which train on your calls, keep them in the EU and include a DPA. [Deutsch: Telefon und Sprache: KI-Training, EU-Hosting und AVV im Vergleich](https://yardwork.dev/de/data/ai-tools/phone-and-voice).
- [Workspace and projects: AI training, EU hosting and DPA compared](https://yardwork.dev/en/data/ai-tools/workspace-and-projects): Project management and workspace tools such as Notion, Slack, Asana and monday.com: which train on your data, keep it in the EU and include a DPA. [Deutsch: Arbeitsbereich und Projekte: KI-Training, EU-Hosting und AVV im Vergleich](https://yardwork.dev/de/data/ai-tools/workspace-and-projects).
- [Automation: AI training, EU hosting and DPA compared](https://yardwork.dev/en/data/ai-tools/automation): Zapier, Make, n8n and others automation tools: which train on your data, keep it in the EU and include a DPA, from their own terms. [Deutsch: Automatisierung: KI-Training, EU-Hosting und AVV im Vergleich](https://yardwork.dev/de/data/ai-tools/automation).
- [Sales, marketing and support: AI training, EU hosting and DPA compared](https://yardwork.dev/en/data/ai-tools/sales-marketing-and-support): CRM, email marketing and support tools such as HubSpot, Pipedrive, Mailchimp and Zendesk: which train on customer data, keep it in the EU and sign a DPA. [Deutsch: Vertrieb, Marketing und Support: KI-Training, EU-Hosting und AVV im Vergleich](https://yardwork.dev/de/data/ai-tools/sales-marketing-and-support).
- [Code and app builders: AI training, EU hosting and DPA compared](https://yardwork.dev/en/data/ai-tools/code-and-app-builders): AI coding tools and app builders such as GitHub Copilot, Cursor, Lovable and Replit: which train on your code, keep it in the EU and include a DPA. [Deutsch: Code und App-Builder: KI-Training, EU-Hosting und AVV im Vergleich](https://yardwork.dev/de/data/ai-tools/code-and-app-builders).
- [Design, images and video: AI training, EU hosting and DPA compared](https://yardwork.dev/en/data/ai-tools/design-images-and-video): AI image generators, design and video tools such as Canva, Figma, HeyGen and Gamma: which train on your files, keep them in the EU and include a DPA. [Deutsch: Design, Bilder und Video: KI-Training, EU-Hosting und AVV im Vergleich](https://yardwork.dev/de/data/ai-tools/design-images-and-video).
- [Documents, HR and finance: AI training, EU hosting and DPA compared](https://yardwork.dev/en/data/ai-tools/documents-hr-and-finance): E-signature, HR and accounting tools such as Docusign, Personio and sevdesk: which use AI on your files, keep data in the EU and include a DPA. [Deutsch: Dokumente, HR und Finanzen: KI-Training, EU-Hosting und AVV im Vergleich](https://yardwork.dev/de/data/ai-tools/documents-hr-and-finance).
- [Research, forms and storage: AI training, EU hosting and DPA compared](https://yardwork.dev/en/data/ai-tools/research-forms-and-storage): Typeform, Dropbox, Calendly and others form, scheduling and storage tools: which train on your data, keep it in the EU and include a DPA. [Deutsch: Recherche, Formulare und Speicher: KI-Training, EU-Hosting und AVV im Vergleich](https://yardwork.dev/de/data/ai-tools/research-forms-and-storage).
- [Download: CSV](https://yardwork.dev/data/ai-tools.csv): Every answer with source URL, locator, quote, status and check date.
- [Download: JSON](https://yardwork.dev/data/ai-tools.json): The same data as JSON, grouped by tool and plan.
- 200 tool pages follow the pattern https://yardwork.dev/en/data/ai-tools/<tool> (for example https://yardwork.dev/en/data/ai-tools/chatgpt); the full list is in llms-full.txt and the sitemap.
<!-- ai-tools:end -->

## Case studies

- [An agentic sales system for a professional services agency](https://yardwork.dev/en/case-studies/bespoke-outreach-agent): A lean agency spent hours researching prospects and chasing follow-ups by hand. We built an approval-based outreach pipeline to replace that manual work. [Deutsch: Ein agentisches Vertriebssystem für eine Dienstleistungsagentur](https://yardwork.dev/de/case-studies/bespoke-outreach-agent).
- [A multi-agent operations system for a business owner](https://yardwork.dev/en/case-studies/multi-business-operations-agent): A multi-agent system built for one business owner with several ventures: a dedicated agent for each area of work, and an executive assistant agent that can direct them. [Deutsch: Ein Multi-Agenten-System für einen Unternehmer](https://yardwork.dev/de/case-studies/multi-business-operations-agent).
- [Performance: an AI coaching platform for professional cyclists](https://yardwork.dev/en/case-studies/performance-ai-coach): We took Performance from idea to launch for an entrepreneur: an AI coaching assistant that lives in messenger apps and helps cyclists genuinely improve. [Deutsch: Performance: eine KI-Coaching-Plattform für Profi-Radsportler](https://yardwork.dev/de/case-studies/performance-ai-coach).

## Managed-agent detail pages

These platform and industry pages remain live as detailed references. They are not additional primary offers and are not linked from the main navigation, footer or Services overview.

- [OpenClaw setup](https://yardwork.dev/en/services/openclaw): Setup and support for an OpenClaw assistant on hardware the client controls, with connected accounts, scoped access and review steps. [Deutsch: OpenClaw einrichten](https://yardwork.dev/de/services/openclaw).
- [Hermes Agent setup](https://yardwork.dev/en/services/hermes): Setup and support for Hermes on a chosen computer or server, with recurring jobs, retained working instructions and agreed boundaries. [Deutsch: Hermes Agent einrichten](https://yardwork.dev/de/services/hermes).
- [Grok Bot setup](https://yardwork.dev/en/services/grok-bot): Setup and support for Grok Bot in Cursor's cloud. Bots under one user share a computer, files and signed-in accounts; separate names do not isolate access. Product access and usage are separate from Yardwork's work. [Deutsch: Grok Bot einrichten](https://yardwork.dev/de/services/grok-bot).
- [Agents for agencies](https://yardwork.dev/en/services/agencies): Managed-agent work for marketing and creative agencies, covering reporting, client communication and pipeline tasks. [Deutsch: Agenten für Agenturen](https://yardwork.dev/de/services/agencies).
- [Agents for commercial real estate](https://yardwork.dev/en/services/commercial-real-estate): Managed-agent work for commercial real-estate teams, around their documents, contacts and recurring coordination. [Deutsch: Agenten für Gewerbeimmobilien](https://yardwork.dev/de/services/commercial-real-estate).
- [Agents for recruiting firms](https://yardwork.dev/en/services/recruiting-firms): Managed-agent work for recruiting desks, around candidate information, client requirements and follow-up. [Deutsch: Agenten für Recruiting-Unternehmen](https://yardwork.dev/de/services/recruiting-firms).
- [Agents for service businesses](https://yardwork.dev/en/services/service-businesses): Managed-agent work for service companies, around enquiries, scheduling, documents and follow-up. [Deutsch: Agenten für Dienstleistungsunternehmen](https://yardwork.dev/de/services/service-businesses).

## Locations

- [Berlin](https://yardwork.dev/en/berlin): Custom software and AI-agent work for Berlin companies, with an in-person first meeting available and ongoing support. [Deutsch: Berlin](https://yardwork.dev/de/berlin).
- [Munich](https://yardwork.dev/en/munich): Custom software and AI-agent work for companies in Munich, delivered remotely around existing tools and local working context. [Deutsch: München](https://yardwork.dev/de/munich).
- [Hamburg](https://yardwork.dev/en/hamburg): Custom software and AI-agent work for companies in Hamburg, delivered remotely around existing tools and local working context. [Deutsch: Hamburg](https://yardwork.dev/de/hamburg).
- [Frankfurt](https://yardwork.dev/en/frankfurt): Custom software and AI-agent work for companies in Frankfurt, delivered remotely around existing tools and local working context. [Deutsch: Frankfurt](https://yardwork.dev/de/frankfurt).
- [Cologne](https://yardwork.dev/en/cologne): Custom software and AI-agent work for companies in Cologne, delivered remotely around existing tools and local working context. [Deutsch: Köln](https://yardwork.dev/de/cologne).
- [London](https://yardwork.dev/en/london): Custom software and AI-agent work for companies in London, delivered remotely around existing tools and local working context. [Deutsch: London](https://yardwork.dev/de/london).
- [Dublin](https://yardwork.dev/en/dublin): Custom software and AI-agent work for companies in Dublin, delivered remotely around existing tools and local working context. [Deutsch: Dublin](https://yardwork.dev/de/dublin).
- [Amsterdam](https://yardwork.dev/en/amsterdam): Custom software and AI-agent work for companies in Amsterdam, delivered remotely around existing tools and local working context. [Deutsch: Amsterdam](https://yardwork.dev/de/amsterdam).
- [Vienna](https://yardwork.dev/en/vienna): Custom software and AI-agent work for companies in Vienna, delivered remotely around existing tools and local working context. [Deutsch: Wien](https://yardwork.dev/de/vienna).
- [Zurich](https://yardwork.dev/en/zurich): Custom software and AI-agent work for companies in Zurich, delivered remotely around existing tools and local working context. [Deutsch: Zürich](https://yardwork.dev/de/zurich).

## Blog — current article inventory

- [The Agent That Deleted a Production Database](https://yardwork.dev/en/blog/agent-blast-radius): A Cursor agent destroyed PocketOS's database and its backups in nine seconds. Its rules said never to. It quoted the rule while breaking it. [Deutsch: KI-Agent löscht Datenbank: Zugänge und Backups prüfen](https://yardwork.dev/de/blog/agent-blast-radius).
- [What an AI Agent Implementation Costs in Europe](https://yardwork.dev/en/blog/agent-cost-europe): Published quotes range from $2K to $400K. Eurostat puts the real labour cost base at €48.20 an hour across the EU — €24.00 in Bulgaria, €71.70 in Ireland. [Deutsch: Was die Implementierung eines KI-Agenten in Europa kostet](https://yardwork.dev/de/blog/agent-cost-europe).
- [Where Your AI Agent Needs a DPIA, and Where It Does Not](https://yardwork.dev/en/blog/agent-dpia): Every EU authority publishes its own DPIA list, and they differ. The Netherlands, Italy and Germany name employee monitoring. Five countries do not. [Deutsch: Wo Ihr KI-Agent eine DSFA braucht — und wo nicht](https://yardwork.dev/de/blog/agent-dpia).
- [AI Agent Liability in Europe: Who Actually Pays](https://yardwork.dev/en/blog/agent-liability-europe): The AI Act compensates nobody, the Product Liability Directive covers natural persons only, and the directive written to close the gap was withdrawn. [Deutsch: KI-Haftungsrichtlinie zurückgezogen: Wer für KI haftet](https://yardwork.dev/de/blog/agent-liability-europe).
- [Agentic AI Statistics 2026: What the Data Shows](https://yardwork.dev/en/blog/agentic-ai-statistics): In 2026, 88% of organizations use AI but only 12% of CEOs report both revenue gains and cost cuts from it. The gap is workflow redesign — not adoption. The 2026 agentic AI numbers, from Stanford HAI, PwC, Databricks, and NVIDIA. [Deutsch: Agentic-AI-Statistiken 2026: Was die Daten zeigen](https://yardwork.dev/de/blog/agentic-ai-statistics).
- [European AI Adoption by Business Size and Sector](https://yardwork.dev/en/blog/ai-adoption-european-enterprises): Eurostat puts EU AI use at 17% for small firms and 55.03% for large ones. See where adoption is concentrated and what the gap means for implementation. [Deutsch: KI-Nutzung in Europa nach Unternehmensgröße und Branche](https://yardwork.dev/de/blog/ai-adoption-european-enterprises).
- [AI Agent Adoption Statistics: 2025–2026 Survey Data](https://yardwork.dev/en/blog/ai-agent-adoption-stats): 79% of companies report AI agents already being adopted (PwC, 2025). Small business AI usage jumped from 39% to 55% in one year. Key statistics on adoption rates, outcomes, and trends by function. [Deutsch: KI-Agenten-Adoption: Umfragedaten 2025–2026](https://yardwork.dev/de/blog/ai-agent-adoption-stats).
- [What Your AI Agent Can Actually See](https://yardwork.dev/en/blog/ai-agent-data-access-permissions): An agent sees exactly what its OAuth scopes allow. gmail.send cannot read your mailbox. drive.readonly reads every file you own. Here is how to check yours. [Deutsch: Was Ihr KI-Agent wirklich sehen kann](https://yardwork.dev/de/blog/ai-agent-data-access-permissions).
- [Can AI Read Your Scanned Documents? The Limits](https://yardwork.dev/en/blog/ai-agent-document-extraction-limits): Textract needs text 15 pixels tall — about 8pt at 150 DPI. Below that, nothing helps. Six providers' published limits, and how to check your own documents against them. [Deutsch: Dokumente mit KI auslesen: Die Grenzen bei Scans und PDFs](https://yardwork.dev/de/blog/ai-agent-document-extraction-limits).
- [AI Agents for Client Onboarding: What Gets Automated](https://yardwork.dev/en/blog/ai-agent-for-client-onboarding): An AI agent for client onboarding can handle document requests, scheduling, CRM updates, and reminders. Define the approvals and exceptions your team retains. [Deutsch: KI-Agenten für das Kunden-Onboarding: Was automatisiert wird](https://yardwork.dev/de/blog/ai-agent-for-client-onboarding).
- [AI Agents for Lead Generation: Tasks and Limits](https://yardwork.dev/en/blog/ai-agent-for-lead-generation): An AI agent for lead generation handles first-contact response, follow-up sequencing, and qualification routing — not prospecting. The gain is in the gap between first contact and first conversation. [Deutsch: KI-Agenten für Leadgenerierung: Aufgaben und Grenzen](https://yardwork.dev/de/blog/ai-agent-for-lead-generation).
- [AI Agents for Reporting: Automating the Data Assembly Layer](https://yardwork.dev/en/blog/ai-agent-for-reporting): An AI agent for reporting can retrieve agreed fields, assemble a recurring report, and queue it for review. Check source freshness, exceptions, and measured effort. [Deutsch: KI-Agenten für Reporting: Datenmontage automatisieren](https://yardwork.dev/de/blog/ai-agent-for-reporting).
- [What AI Agent Governance Means for a Small Business](https://yardwork.dev/en/blog/ai-agent-governance): AI agent governance defines what agents can access, who can change that, and what triggers a review. It is different from approval workflows — and becomes necessary the moment you run more than one agent. [Deutsch: Was KI-Agent-Governance für kleine Unternehmen bedeutet](https://yardwork.dev/de/blog/ai-agent-governance).
- [What AI Agent Maintenance Actually Involves](https://yardwork.dev/en/blog/ai-agent-maintenance): AI agent maintenance is rewriting prompts when the business changes, patching integrations when tools update, and reviewing logs for silent edge cases. [Deutsch: Was die Wartung eines KI-Agenten wirklich bedeutet](https://yardwork.dev/de/blog/ai-agent-maintenance).
- [AI Agent Market Size 2026: What the Numbers Show](https://yardwork.dev/en/blog/ai-agent-market-size): The AI agent market reached roughly $10.9 billion in 2026 (Grand View), projected to hit $182.9 billion by 2033 at a 49.6% CAGR. Precedence projects $294.66 billion by 2035. What the 2026 research shows — and why firms report different figures. [Deutsch: KI-Agent-Marktgröße 2026: Was die Zahlen zeigen](https://yardwork.dev/de/blog/ai-agent-market-size).
- [AI Agent ROI: How to Measure It in a Service Business](https://yardwork.dev/en/blog/ai-agent-roi): Calculate AI agent ROI from time, response and error benefits after review and operating costs. A practical measurement guide for service businesses. [Deutsch: KI-Agenten-ROI: Messung für Dienstleister](https://yardwork.dev/de/blog/ai-agent-roi).
- [AI Agent ROI Statistics: 2026 Benchmarks and Studies](https://yardwork.dev/en/blog/ai-agent-roi-statistics): AI agent ROI statistics measure different things: personal productivity, company profit and modeled potential. Compare verified 2026 findings and calculate your own workflow return. [Deutsch: KI-Agenten-ROI-Statistiken: Benchmarks und Studien 2026](https://yardwork.dev/de/blog/ai-agent-roi-statistics).
- [How to Structure a Team of AI Agents](https://yardwork.dev/en/blog/ai-agent-team-structure): Structure an AI agent team by assigning trigger ownership and write authority per data field before the second agent launches, or it fails silently. [Deutsch: Wie Sie ein Team aus KI-Agenten strukturieren](https://yardwork.dev/de/blog/ai-agent-team-structure).
- [AI Agent vs. Hiring Cost in 2026: The Full Comparison](https://yardwork.dev/en/blog/ai-agent-vs-hiring-cost): In 2026, MIT found AI automation is economically viable in only ~23% of roles — for the other 77%, humans are still cheaper. A fully-loaded employee costs 1.25–1.40× salary; an agent's cost scales with task volume. The comparison only works per task. [Deutsch: KI-Agent vs. Einstellungskosten 2026: Der Vergleich](https://yardwork.dev/de/blog/ai-agent-vs-hiring-cost).
- [AI Resume Screening: What US and EU Law Require](https://yardwork.dev/en/blog/ai-applicant-screening): Screening applicants with AI is legal, but you answer for the result: discrimination law applies, the GDPR limits automated rejections, and the EU AI Act calls it high-risk. [Deutsch: KI in der Bewerberauswahl: Hochrisiko nach KI-Verordnung](https://yardwork.dev/de/blog/ai-applicant-screening).
- [AI-Written Content and SEO: What Google Penalizes](https://yardwork.dev/en/blog/ai-content-seo): Google does not penalize content because AI wrote it. It penalizes mass-produced pages without value. What its rules say, what the data shows, what to check. [Deutsch: KI-Texte und SEO: Was Google abstraft und was nicht](https://yardwork.dev/de/blog/ai-content-seo).
- [AI in Customer Service: 2025–2026 Statistics](https://yardwork.dev/en/blog/ai-customer-service-statistics): The widely-repeated '75% resolved by AI' figure has no disclosed source. What's verified: NBER found a 14% productivity gain for humans working alongside AI, and Gartner projects 80% autonomous resolution — by 2029, not now. [Deutsch: KI im Kundenservice: Statistiken 2025–2026](https://yardwork.dev/de/blog/ai-customer-service-statistics).
- [AI for Excel: What ChatGPT, Copilot and Claude Get Wrong](https://yardwork.dev/en/blog/ai-excel-analysis): On real business spreadsheets, the best AI model completed 34.89% of tasks in a June 2026 benchmark. File limits, how each tool calculates, and the checks before you trust a result. [Deutsch: KI für Excel: Wo ChatGPT, Copilot und Claude falsch rechnen](https://yardwork.dev/de/blog/ai-excel-analysis).
- [AI Fines in Europe 2026: Cases So Far and AI Act Penalties](https://yardwork.dev/en/blog/ai-fines-europe): Europe's largest AI-related fine, €824.99 million against Uber in 2026, punished automated decisions without human review. The cases and the AI Act's fines. [Deutsch: KI-Bußgelder in Europa 2026: Fälle und KI-Verordnung](https://yardwork.dev/de/blog/ai-fines-europe).
- [The AI Implementation Failure Rate in 2026, Explained](https://yardwork.dev/en/blog/ai-implementation-failure-rate): In 2026, 42% of companies abandoned most AI initiatives (up from 17%) and 88% of pilots never reach production. The stat measures pilots, not production implementations. The two causes that transfer to production are preventable before the build starts. [Deutsch: Wie viele KI-Projekte scheitern: Was die Studien messen](https://yardwork.dev/de/blog/ai-implementation-failure-rate).
- [AI Knowledge Base: Where Answers From Your Files Go Wrong](https://yardwork.dev/en/blog/ai-knowledge-base): Legal AI tools built on curated sources were wrong on 17–33% of questions in a 2024 Stanford study. Five places a company knowledge base fails, and a test. [Deutsch: KI-Wissensdatenbank im Unternehmen: Wo Antworten scheitern](https://yardwork.dev/de/blog/ai-knowledge-base).
- [AI Literacy Under the EU AI Act: What Article 4 Requires](https://yardwork.dev/en/blog/ai-literacy): Since July 2026, Article 4 of the EU AI Act asks businesses to support their staff's AI literacy, with no fixed level and no certificate. Who must comply, and what to do. [Deutsch: KI-Kompetenz: Was Art. 4 der KI-Verordnung jetzt verlangt](https://yardwork.dev/de/blog/ai-literacy).
- [Who Owns AI-Generated Content? Copyright and Clients](https://yardwork.dev/en/blog/ai-output-copyright): Your AI provider says you own the output, but copyright protects only what a human created. Who owns AI work in the US, the EU and the UK, and what to tell clients. [Deutsch: Wem gehören KI-generierte Inhalte? Urheberrecht und Kunden](https://yardwork.dev/de/blog/ai-output-copyright).
- [AI Phone Assistant: What It Costs and Where It Fails](https://yardwork.dev/en/blog/ai-phone-assistant): An AI phone assistant cost from about $0.07 a minute in October 2026. Prices in the US and Europe, where it fails, what the law requires, and a test. [Deutsch: KI-Telefonassistent: Kosten, Grenzen und Pflichten](https://yardwork.dev/de/blog/ai-phone-assistant).
- [AI-Written Quotes: What to Check Before You Send](https://yardwork.dev/en/blog/ai-quotes): AI drafts a trade quote from site notes in minutes, but an accepted quote fixes your price. What AI quoting tools do, where they go wrong, what to check. [Deutsch: Angebote mit KI erstellen: Was Sie vor dem Versand prüfen](https://yardwork.dev/de/blog/ai-quotes).
- [Which AI Providers Train on Your Data](https://yardwork.dev/en/blog/ai-provider-training-data-policies): Anthropic bars training on commercial accounts — until someone clicks thumbs up, which keeps that conversation five years. Eight products, read from the terms, by plan. [Deutsch: Welche KI-Anbieter mit Ihren Daten trainieren](https://yardwork.dev/de/blog/ai-provider-training-data-policies).
- [Enterprise AI ROI by Industry: 2026 Survey Data](https://yardwork.dev/en/blog/ai-roi-by-industry): In 2026, financial services, telecom, and retail/CPG lead AI ROI, yet 56% of all CEOs report no return at all. The gap tracks workflow structure, not the tool. AI ROI by industry, from NVIDIA, PwC, IBM, and Accenture 2026 data. [Deutsch: Enterprise-KI-ROI nach Branche: Umfragedaten 2026](https://yardwork.dev/de/blog/ai-roi-by-industry).
- [AI Spending Statistics: 2026 Global Data](https://yardwork.dev/en/blog/ai-spending-statistics): Gartner forecasts worldwide AI spending at $2.59 trillion in 2026 — up 47% year over year and roughly 41% of all IT spending. Enterprise GenAI in production reached $37 billion. The production gap is the story behind the headline numbers. [Deutsch: KI-Ausgaben 2026: Weltweite Zahlen von Gartner und IDC](https://yardwork.dev/de/blog/ai-spending-statistics).
- [AI and Technical Drawings: It Reads Text, Not Geometry](https://yardwork.dev/en/blog/ai-technical-drawings): On engineering drawings, the best general AI model answered 97.73% of title-block and table questions right but 46% of assembly questions in a 2026 benchmark. What AI misreads, specialist tools, and how to test yours. [Deutsch: Technische Zeichnungen mit KI lesen: Text ja, Geometrie kaum](https://yardwork.dev/de/blog/ai-technical-drawings).
- [Introducing AI at Work: When Employees Must Be Consulted](https://yardwork.dev/en/blog/ai-works-council): Whether employees get a say in new AI tools depends on where you are: union bargaining in the US, consultation in the EU, and Germany's works councils. What applies, and when. [Deutsch: KI im Betrieb einführen: Wann der Betriebsrat mitbestimmt](https://yardwork.dev/de/blog/ai-works-council).
- [Before You Hire an AI Agent Implementer](https://yardwork.dev/en/blog/before-you-hire): The businesses that reach a working agent fastest are the ones who documented their process before the first call. What preparation actually looks like — and why it cuts implementation time more than any other variable. [Deutsch: Bevor Sie einen KI-Agenten-Implementierer beauftragen](https://yardwork.dev/de/blog/before-you-hire).
- [What Moves When You Change AI Providers](https://yardwork.dev/en/blog/change-providers): Switching the AI model behind a running agent is not a settings change. The instructions, tool connections and testing all get redone. [Deutsch: Was sich ändert, wenn Sie den KI-Anbieter wechseln](https://yardwork.dev/de/blog/change-providers).
- [ChatGPT and the GDPR: Which Plan a Business Can Use](https://yardwork.dev/en/blog/chatgpt-gdpr): OpenAI signs a GDPR processing agreement for ChatGPT Business, Enterprise and the API only. On Free, Go, Plus and Pro, it is a second controller of your data. [Deutsch: ChatGPT DSGVO-konform nutzen: Tarife mit AV-Vertrag](https://yardwork.dev/de/blog/chatgpt-gdpr).
- [You Don't Need Clean Data to Start](https://yardwork.dev/en/blog/clean-data-myth): Start by checking the fields your AI agent actually needs. A full CRM cleanup can wait, but missing values, access boundaries, and fallback rules need testing. [Deutsch: Saubere Daten sind keine Voraussetzung](https://yardwork.dev/de/blog/clean-data-myth).
- [Client Confidentiality and AI: What Firms Must Check](https://yardwork.dev/en/blog/client-confidentiality-ai): Client confidentiality applies to AI tools. What lawyers, accountants and doctors must check under ABA Opinion 512, HIPAA, UK rules and the GDPR. [Deutsch: KI und Berufsgeheimnis: Was Kanzleien prüfen müssen](https://yardwork.dev/de/blog/client-confidentiality-ai).
- [Is Cold Email Legal? Rules for Sales AI in the US and EU](https://yardwork.dev/en/blog/cold-outreach-rules): Is cold email legal? In the US yes, under CAN-SPAM's rules; in Europe it depends on the country. What a sales AI may send by email, phone and LinkedIn. [Deutsch: Ist Kaltakquise mit KI erlaubt? E-Mail, Telefon, LinkedIn](https://yardwork.dev/de/blog/cold-outreach-rules).
- [Microsoft Copilot and the GDPR: Checks Before Switch-On](https://yardwork.dev/en/blog/copilot-gdpr): Copilot reads whatever each user can open in Microsoft 365. Three settings decide whether its processing leaves the EU, and one is on by default. [Deutsch: Microsoft Copilot DSGVO-konform nutzen: Prüfungen vorab](https://yardwork.dev/de/blog/copilot-gdpr).
- [Copilot Studio vs a Custom Agent: Pricing and Limits](https://yardwork.dev/en/blog/copilot-studio): Copilot Studio bills 1 to 10 credits per answer at $200 per 25,000, free for staff with Copilot licences. Costs, limits, and when a custom agent fits. [Deutsch: Copilot Studio oder eigener Agent: Kosten und Grenzen](https://yardwork.dev/de/blog/copilot-studio).
- [Custom Agent vs Off-the-Shelf: How to Choose](https://yardwork.dev/en/blog/custom-vs-off-the-shelf): Off-the-shelf agents bake in assumptions about approval chains, data formats, and output destinations. When those assumptions don't fit your workflow, workarounds compound — and maintaining them costs more than a purpose-built custom agent. [Deutsch: Custom Agent vs. Off-the-Shelf: Wie Sie entscheiden](https://yardwork.dev/de/blog/custom-vs-off-the-shelf).
- [GDPR Right to Erasure for AI Agents, in Practice](https://yardwork.dev/en/blog/data-erasure): A deletion request reaches more than the chat: the agent's action log, the search index it built, and the backups. One month covers all of it. [Deutsch: Recht auf Löschung bei KI-Agenten, in der Praxis](https://yardwork.dev/de/blog/data-erasure).
- [Does AI Data Residency Keep Your Data in Europe?](https://yardwork.dev/en/blog/data-residency): Azure's default processes prompts in any region. Google warns off its global endpoint. OpenAI charges 10% more for EU processing. [Deutsch: Hält Datenresidenz Ihre KI-Daten in Europa?](https://yardwork.dev/de/blog/data-residency).
- [EU AI Act: A Deployer Guide for Small Businesses](https://yardwork.dev/en/blog/eu-ai-act-deployer-guide): A practical EU AI Act intake for small businesses: identify your role, transparency duties, human controls, logs, and when you need legal advice. [Deutsch: EU AI Act: Leitfaden für Betreiber in kleinen Unternehmen](https://yardwork.dev/de/blog/eu-ai-act-deployer-guide).
- [European Alternatives to ChatGPT: Business Terms Compared](https://yardwork.dev/en/blog/european-chatgpt-alternatives): Four European AI assistants sell business seats. Mistral's Team plan trains on chats until an admin opts out; ChatGPT Business does not. As of 13 Oct 2026. [Deutsch: Europäische Alternativen zu ChatGPT für Unternehmen](https://yardwork.dev/de/blog/european-chatgpt-alternatives).
- [GDPR and AI Agents: What European Deployers Must Do](https://yardwork.dev/en/blog/gdpr-ai-agents-european-deployers): 53.61% of EU firms that considered AI cite unclear legal consequences. Here is which GDPR obligation each agent action triggers, and who is controller. [Deutsch: DSGVO und KI-Agenten: Was Betreiber wissen müssen](https://yardwork.dev/de/blog/gdpr-ai-agents-european-deployers).
- [Does an AI Agent Write German as Well as English?](https://yardwork.dev/en/blog/german-quality): Across 63 published EN/DE article pairs, German cost a median 21.5% more tokens while using fewer words. The quality gap is smaller than claimed. [Deutsch: Schreibt ein KI-Agent so gut Deutsch wie Englisch?](https://yardwork.dev/de/blog/german-quality).
- [Hermes Setup Guide: Deploy on Your Own Server](https://yardwork.dev/en/blog/hermes-setup-guide): A Hermes setup needs defined work, scoped tool access and tested review rules. Separate installation, local skills and external AI processing before launch. [Deutsch: Hermes Agent installieren: Setup auf dem eigenen Server](https://yardwork.dev/de/blog/hermes-setup-guide).
- [How Hermes Learns: Saving and Reusing Skills](https://yardwork.dev/en/blog/how-hermes-learns): Hermes can save reusable task instructions as skills. Learn where they live, how sharing works, and how to check whether a revision improves results. [Deutsch: Hermes Agent Skills: Speichern, prüfen, wiederverwenden](https://yardwork.dev/de/blog/how-hermes-learns).
- [AI Hallucination Rates 2026: How Reliable AI Agents Are](https://yardwork.dev/en/blog/how-reliable-are-ai-agents): AI models added unsupported content to 1.8% to 24.2% of document summaries in Vectara's October 2026 test. Error rates by model and task, why they differ, and how to test your own work. [Deutsch: KI-Modelle im Vergleich 2026: Wie oft sie falsch liegen](https://yardwork.dev/de/blog/how-reliable-are-ai-agents).
- [How to Brief an AI Agent for Your Workflow](https://yardwork.dev/en/blog/how-to-brief-an-ai-agent): A useful AI agent brief defines triggers, inputs, outputs, exceptions, and escalation. Test the acceptance criteria before connecting it to live work. [Deutsch: Wie man einen KI-Agenten brieft](https://yardwork.dev/de/blog/how-to-brief-an-ai-agent).
- [Custom AI Agent Development: The Five-Stage Build](https://yardwork.dev/en/blog/how-to-build-a-custom-agent): A custom agent build stalls at stage 1 — process documentation — more than any other stage. If the workflow isn't written down, it can't be built. Here is the full build process, stage by stage, with timelines and what to do when something breaks. [Deutsch: Custom-KI-Agent-Entwicklung: Der Fünf-Stufen-Build](https://yardwork.dev/de/blog/how-to-build-a-custom-agent).
- [How to Choose an AI Implementation Partner](https://yardwork.dev/en/blog/how-to-choose-an-ai-implementation-partner): The right implementation partner transfers knowledge and leaves you in control; the wrong one creates a dependency costing more in year two than the build. [Deutsch: Wie Sie den richtigen KI-Implementierungspartner auswählen](https://yardwork.dev/de/blog/how-to-choose-an-ai-implementation-partner).
- [How to Hire an AI Developer for a Small Business](https://yardwork.dev/en/blog/how-to-hire-an-ai-developer-for-small-business): Freelance AI developers charge $93–$160/hr, in-house engineers cost $134K–$193K a year. Neither is the real risk — a single point of failure is. Here's how to hire around it. [Deutsch: KI-Entwickler für kleine Unternehmen einstellen](https://yardwork.dev/de/blog/how-to-hire-an-ai-developer-for-small-business).
- [When Is a Process Ready for an AI Agent?](https://yardwork.dev/en/blog/how-to-know-if-a-business-process-is-ready-to-hand-to-an-ai-agent): The readiness bottleneck is not technology. It is whether the process has ever been specified clearly enough to run without the person who knows it. [Deutsch: Ist Ihr Geschäftsprozess bereit für einen KI-Agenten?](https://yardwork.dev/de/blog/how-to-know-if-a-business-process-is-ready-to-hand-to-an-ai-agent).
- [AI Agent Management: What It Takes After Launch](https://yardwork.dev/en/blog/how-to-manage-ai-agents): Managing an AI agent takes 30–45 minutes a week and looks nothing like managing software. Most are abandoned within 90 days because no one owned that time. [Deutsch: KI-Agenten-Management: Was nach dem Launch anfällt](https://yardwork.dev/de/blog/how-to-manage-ai-agents).
- [How to Onboard an AI Agent Into Your Team](https://yardwork.dev/en/blog/how-to-onboard-an-ai-agent): Onboard an AI agent with a named reviewer, a review cadence and workflow-specific success criteria. Use this adaptable 30-day example before reducing oversight. [Deutsch: Wie man einen KI-Agenten ins Team einführt](https://yardwork.dev/de/blog/how-to-onboard-an-ai-agent).
- [How to Scale an AI Agent System](https://yardwork.dev/en/blog/how-to-scale-ai-agents): Scaling AI agents means building ownership rules, not adding more agents. Without defined trigger and data ownership, each new one creates silent failures. [Deutsch: Wie man ein KI-Agenten-System skaliert](https://yardwork.dev/de/blog/how-to-scale-ai-agents).
- [How to Stay in Control of an AI Agent Working on Your Behalf](https://yardwork.dev/en/blog/how-to-stay-in-control-of-an-ai-agent-working-on-your-behalf): Control over an AI agent is not a monitoring habit — it is a design decision made before the agent runs. Permission scoping and approval gates define what the agent can do without you. [Deutsch: So behalten Sie die Kontrolle über Ihren KI-Agenten](https://yardwork.dev/de/blog/how-to-stay-in-control-of-an-ai-agent-working-on-your-behalf).
- [AI Agent Implementation Timeline: How Long It Takes](https://yardwork.dev/en/blog/implementation-timeline): A standard agent implementation for a service business takes 2–4 weeks from scoping to live. Integration complexity and process documentation drive timeline — not AI sophistication. [Deutsch: KI-Agenten-Implementierung: Wie lange sie dauert](https://yardwork.dev/de/blog/implementation-timeline).
- [Is Your AI Agent Actually Working?](https://yardwork.dev/en/blog/is-your-agent-working): Without defined success criteria there is no way to tell a running agent from a working one — and degradation stays invisible until it's expensive. [Deutsch: Funktioniert Ihr KI-Agent wirklich?](https://yardwork.dev/de/blog/is-your-agent-working).
- [Is Your Business Ready for AI Agents?](https://yardwork.dev/en/blog/is-your-business-ready-for-ai-agents): AI agent readiness needs a defined workflow, ownership, usable data and technical access. Assess the first job instead of relying on a generic failure rate. [Deutsch: Ist Ihr Unternehmen bereit für KI-Agenten?](https://yardwork.dev/de/blog/is-your-business-ready-for-ai-agents).
- [OpenClaw vs Hermes vs Grok Bot: None Is Safe by Default](https://yardwork.dev/en/blog/openclaw-hermes-grok-bot): OpenClaw, Hermes and Grok Bot compared using current plans, hosting, skills and approval controls. Choose by tested workflow fit and actual data boundaries. [Deutsch: OpenClaw, Hermes und Grok Bot: Freigaben ab Werk aus](https://yardwork.dev/de/blog/openclaw-hermes-grok-bot).
- [OpenClaw Setup Guide: Approval Rules Before Go-Live](https://yardwork.dev/en/blog/openclaw-setup-business): OpenClaw deploys on your own server in under a day. The step that decides whether it's safe — scoping permissions and naming an approver — takes longer. [Deutsch: OpenClaw einrichten: Anleitung mit Freigaberegeln](https://yardwork.dev/de/blog/openclaw-setup-business).
- [OpenClaw vs. Hermes: An AI Agent Tool Comparison](https://yardwork.dev/en/blog/openclaw-vs-hermes): OpenClaw and Hermes both connect tools and multi-step work. Compare their actual approvals, reusable instructions and operating requirements before choosing. [Deutsch: OpenClaw vs. Hermes: Vergleich zweier KI-Agenten-Tools](https://yardwork.dev/de/blog/openclaw-vs-hermes).
- [Running Multiple AI Agents at Once](https://yardwork.dev/en/blog/running-multiple-agents): Running multiple AI agents in the same environment creates trigger collisions and data conflicts that single-agent guides never cover. Here is what breaks, how to trace it, and how to prevent it. [Deutsch: Mehrere KI-Agenten gleichzeitig betreiben](https://yardwork.dev/de/blog/running-multiple-agents).
- [Small Business AI: Production Use vs Headline Adoption](https://yardwork.dev/en/blog/small-business-ai-adoption-statistics): 58% of US small businesses report using AI, but the Census Bureau's last narrowly-defined production reading put it at 8.8% — before the survey question changed. That gap is what separates experimenters from operators. [Deutsch: KMU und KI: Produktivbetrieb statt Schlagzeilen-Quote](https://yardwork.dev/de/blog/small-business-ai-adoption-statistics).
- [When an AI Agent Saves Time or Creates More Work](https://yardwork.dev/en/blog/the-difference-between-an-ai-agent-that-saves-time-and-one-that-creates-work): A poorly implemented AI agent doesn't fail by doing nothing. It fails by converting existing work into harder work. Here is what separates the two. [Deutsch: Wann ein KI-Agent Zeit spart oder neue Arbeit schafft](https://yardwork.dev/de/blog/the-difference-between-an-ai-agent-that-saves-time-and-one-that-creates-work).
- [What a Real AI Agent Implementation Actually Involves](https://yardwork.dev/en/blog/what-a-real-ai-agent-implementation-involves): Implementation is not a setup step. It is a sequence of decisions about workflow scope, system connections, controls, and real-conditions testing. [Deutsch: Was echte KI-Agenten-Implementierung umfasst](https://yardwork.dev/de/blog/what-a-real-ai-agent-implementation-involves).
- [What AI Agent Implementation Costs a Small Business](https://yardwork.dev/en/blog/what-ai-agent-implementation-actually-costs-for-a-small-business): Most implementation quotes cover the build and nothing else. Integration and maintenance together often cost as much again over three years. [Deutsch: Was ein KI-Agent kleine Unternehmen wirklich kostet](https://yardwork.dev/de/blog/what-ai-agent-implementation-actually-costs-for-a-small-business).
- [AI Agent Limitations: What They're Actually Bad At](https://yardwork.dev/en/blog/what-ai-agents-are-actually-bad-at): The failure mode most business owners don't expect: AI agents don't fail on hard tasks. They fail on vague ones. Here is what that looks like in practice — and how to screen for it before building. [Deutsch: Grenzen von KI-Agenten: Was sie wirklich schlecht können](https://yardwork.dev/de/blog/what-ai-agents-are-actually-bad-at).
- [What Approval Workflows Actually Do in an AI Agent System](https://yardwork.dev/en/blog/what-approval-workflows-do-in-an-ai-agent-system): Approval workflows define, before anything runs, which actions an agent can take without you — and which it cannot. They are gates, not safety nets. [Deutsch: Freigabe-Workflows in KI-Agentensystemen](https://yardwork.dev/de/blog/what-approval-workflows-do-in-an-ai-agent-system).
- [What Building Your Own AI Agents In-House Actually Takes](https://yardwork.dev/en/blog/what-building-your-own-ai-agents-in-house-actually-takes): In-house AI agent development is a staffing commitment with ongoing obligations, not a build project with a finish line. The economics rarely favour SMBs. [Deutsch: Was der Aufbau eigener KI-Agenten intern wirklich erfordert](https://yardwork.dev/de/blog/what-building-your-own-ai-agents-in-house-actually-takes).
- [What Human-in-the-Loop Actually Means in Practice](https://yardwork.dev/en/blog/what-human-in-the-loop-actually-means-in-practice): Human-in-the-loop is a design decision about which actions need a human before the agent proceeds — not a safety feature vendors switch on. [Deutsch: Was Human-in-the-Loop in der Praxis wirklich bedeutet](https://yardwork.dev/de/blog/what-human-in-the-loop-actually-means-in-practice).
- [What Is a Custom Agent? When to Build vs. Buy](https://yardwork.dev/en/blog/what-is-a-custom-agent): A custom agent is an AI agent built for one business's specific data, processes, and output requirements. The model is a commodity — the integration work, approval logic, and workflow design are what make it custom. Here is when to build, when to buy, and what it costs. [Deutsch: Was ist ein Custom Agent? Wann bauen, wann kaufen](https://yardwork.dev/de/blog/what-is-a-custom-agent).
- [What Is an AI Agent? Definition, Types, and How They Work](https://yardwork.dev/en/blog/what-is-an-ai-agent): An AI agent is software that takes actions on your behalf — reads inputs, makes decisions, and acts in external tools like Gmail, Slack, or HubSpot. Unlike chatbots, agents execute. Here is how they work, what they cost, and when they make sense. [Deutsch: Was ist ein KI-Agent? Definition, Typen und Funktionsweise](https://yardwork.dev/de/blog/what-is-an-ai-agent).
- [What Is Hermes? AI Agents and Reusable Skills Explained](https://yardwork.dev/en/blog/what-is-hermes): Hermes is Nous Research’s open AI agent with reusable skills and 20+ messaging platforms. Learn how its learning, permissions and deployment actually work. [Deutsch: Was ist Hermes Agent? Der KI-Agent von Nous Research](https://yardwork.dev/de/blog/what-is-hermes).
- [What Is OpenClaw? Self-Hosted AI Agent Framework Explained](https://yardwork.dev/en/blog/what-is-openclaw): OpenClaw connects messaging and tools on a computer you control. Learn what stays local, what goes to other services, and which approvals must be built. [Deutsch: Was ist OpenClaw und was kann es im Unternehmen?](https://yardwork.dev/de/blog/what-is-openclaw).
- [What Most Businesses Get Wrong When Implementing AI Agents](https://yardwork.dev/en/blog/what-most-businesses-get-wrong-when-implementing-ai-agents): Most AI agent implementations stall before they reach production. The reasons are predictable, preventable, and have nothing to do with the AI itself. [Deutsch: Was Unternehmen bei KI-Agenten am häufigsten falsch machen](https://yardwork.dev/de/blog/what-most-businesses-get-wrong-when-implementing-ai-agents).
- [What to Do When Your AI Agent Makes a Mistake](https://yardwork.dev/en/blog/what-to-do-when-your-ai-agent-makes-a-mistake): A well-implemented agent system is one where failures surface fast, stop automatically, and leave a trail clear enough to diagnose. [Deutsch: Was zu tun ist, wenn Ihr KI-Agent einen Fehler macht](https://yardwork.dev/de/blog/what-to-do-when-your-ai-agent-makes-a-mistake).
- [When to Let Your AI Agent Act Autonomously](https://yardwork.dev/en/blog/when-to-let-your-ai-agent-act-autonomously): The question is not whether you trust the agent. It is whether you can recover from what it does when it gets something wrong. Recoverability determines what should run without a human decision. [Deutsch: Wann Sie Ihren KI-Agenten autonom handeln lassen sollten](https://yardwork.dev/de/blog/when-to-let-your-ai-agent-act-autonomously).
- [Which Workflows to Automate First](https://yardwork.dev/en/blog/which-workflows-first): Ranking workflows by integration overlap — not business value — cuts total build cost in half and makes each new agent faster than the last. [Deutsch: Welche Workflows zuerst automatisieren](https://yardwork.dev/de/blog/which-workflows-first).
- [Why AI Agent Prototypes Never Make It to Production](https://yardwork.dev/en/blog/why-ai-agent-prototypes-never-make-it-to-production): Most AI agent prototypes stall for one reason: the implementation work that turns a demo into a running system never gets done. [Deutsch: Warum KI-Agent-Prototypen nie live gehen](https://yardwork.dev/de/blog/why-ai-agent-prototypes-never-make-it-to-production).
- [Connecting AI Agents to Real Systems Is Harder Than It Looks](https://yardwork.dev/en/blog/why-connecting-ai-agents-to-real-systems-is-harder-than-it-looks): Connecting AI agents to real systems requires working access, reliable field matching, and tested recovery rules. Those connections need attention after launch. [Deutsch: Warum KI-Agenten-Integrationen schwieriger sind als erwartet](https://yardwork.dev/de/blog/why-connecting-ai-agents-to-real-systems-is-harder-than-it-looks).
- [Why Most AI Agent Projects Stall After Ninety Days](https://yardwork.dev/en/blog/why-most-ai-agent-projects-stall-after-ninety-days): Most AI agent projects don't fail at the build. They stall after go-live, when ownership disappears and no one's job is keeping the agent running. [Deutsch: Warum KI-Agenten-Projekte nach 90 Tagen stocken](https://yardwork.dev/de/blog/why-most-ai-agent-projects-stall-after-ninety-days).
- [Your Second AI Agent Is Harder Than the First](https://yardwork.dev/en/blog/your-second-ai-agent): The second build surfaces every undocumented decision the first one made, and those decisions become constraints you did not know you had chosen. [Deutsch: Der zweite KI-Agent ist schwieriger als der erste](https://yardwork.dev/de/blog/your-second-ai-agent).

## Machine-readable and utility references

- [Detailed site reference](https://yardwork.dev/llms-full.txt): Per-page reference using surviving bilingual article and case-study frontmatter.
- [Entity map](https://yardwork.dev/entity-map.md): Canonical brand, people, offers, audiences and topic associations.
- [Machine-readable pricing](https://yardwork.dev/pricing.md): Scoped pricing model; the pricing page is authoritative for the published starting point.
- [Sitemap](https://yardwork.dev/sitemap.xml): All indexable localized canonical page URLs.
- [AgentBridge privacy](https://yardwork.dev/agentbridge/privacy.html): Standalone extension policy and current contact email.