--- title: "RAG vs fine-tuning vs agents | Lookup, weights, worker | Len P. van der Hof" description: RAG looks up a named passage. Fine-tuning changes weights for a stable skill. An agent holds a goal and uses tools under a charter. They are three jobs… image: "https://lenvanderhof.com/media/generated/blog-hero-rag-vs-finetune-vs-agents-v1.86b37f8dd3d7.wide.webp" --- [AI Systems](https://lenvanderhof.com/en/blog/category/ai-systems/) Research guide # RAG vs fine-tuning vs agents: three jobs, not a ladder A vector store is not a trained model. A trained model is not a worker. A worker without a ceiling is a demo. Len P. van der HofPublished 3 September 2026Updated 9 September 20266 min read Lookup. Weights. Worker. Three benches. Direct answer RAG, fine-tuning, and agents are three jobs. RAG, retrieval-augmented generation, looks up a named passage in a corpus you gave it, then writes. Fine-tuning changes model weights so a stable skill fires without fetching a document. An AI agent holds a goal, uses tools, and continues across steps under a written charter. They are not a ladder of sophistication. Fine-tuning does not make the model know Tuesday's PDF. RAG does not give the model a ceiling. An agent that cannot open a source still needs retrieval or a refuse. What is RAG owns the method. What is an AI agent owns the worker. This page is the comparison. The RAG Engineer is live. Fine-tuning win-rates are not this page. ## Key takeaways - RAG = named passage. Fine-tuning = weights. Agent = goal plus tools plus a ceiling. - Not a ladder. Pick the job that failed, not the slide that is fashionable. - Fine-tuning is not a source of truth. RAG is not a charter. - This page is the three-way. The definition pages stay next door. A chatbot can inherit an API key and still be unable to open the source behind its sentence. That failure is why **RAG, fine-tuning, and agents** must be separated by job, not by status. **RAG vs fine-tuning vs agents.** Three jobs. Not a maturity model. Doing all three is allowed. Stacking them because a slide said “start with RAG, then fine-tune, then agents” is how you buy a retainer and still cannot open the file. [What is RAG?](https://lenvanderhof.com/en/blog/what-is-rag/) owns the lookup. [What is an AI agent?](https://lenvanderhof.com/en/blog/what-is-an-ai-agent/) owns the worker. This page is the comparison: which artefact you should be holding on Friday. ## Three artefacts **RAG.** Retrieval-augmented generation. Look up a named passage in a corpus you gave the system, then write. The artefact is an openable locator: file, page, chunk, date. If you cannot open it, you do not have RAG. You have extra text in a prompt. [Evaluate the reasoning, not the fluency](https://lenvanderhof.com/en/blog/evaluate-reasoning-not-fluency/) is the same test, scored on the trace instead of the prose. A vector store is furniture in that cabinet. Embeddings help find candidates. They do not prove the sentence came from a passage you would accept in a dispute. **Fine-tuning.** A weight change so a stable skill fires without fetching a document: a tone, a format, a classifier that does not move every Tuesday. The artefact is a trained checkpoint with an eval that still holds on held-out cases. Fine-tuning does not make the model know last week’s contract. A changed PDF is a retrieval problem. Treating fine-tuning as a filing cabinet is how stale clauses survive a model upgrade. If the skill still needs a named file, you are not looking at a fine-tune job. You are looking at a lookup you tried to bake into weights. **Agent.** A system that holds a goal, uses tools, and continues across steps under a charter. The artefact is seven written fields and a named person who can fail the work. [Using AI agents effectively](https://lenvanderhof.com/en/blog/using-ai-agents-effectively/) is the operating loop once the worker exists. An agent may *call* retrieval as a tool. That does not turn the agent into RAG, and it does not replace the lookup test. An agent without a ceiling is a demo that can spend. ## One object, three benches Walk one question through the three benches. Do not skip the bench that failed. Someone asks: how many days of notice do we need to cancel the supplier contract? A chatbot answers from folklore. The number may even be a common one. It is not *your* clause. Giving that window an API key does not make the number true. Fine-tuning can teach the model to answer in house tone, or to emit a JSON object with `days`, `clause`, and `confidence`. Useful. The weights still do not contain Tuesday’s signed amendment. If you fine-tuned on last year’s PDF, the tidy JSON will be last year’s number. RAG searches the signed contracts you indexed, finds clause 12.3, and the answer names that clause. You can open the file. If the file is last year’s draft, you will see that too. That is the whole advantage. An agent is a different job. “Chase the cancellation this week, using the sheet and the mail tool, and stop if legal has not signed the notice” is an agent, but only if someone wrote the stop. The agent still needs retrieval, or a refuse, before it drafts the letter. A worker that cannot open the source is still a chatbot with extra tools. Three other objects, so the split does not collapse into contracts. **Ticket type.** “Is this billing, outage, or a feature request?” A classifier that does not need a document is a fine-tune candidate. The eval is held-out tickets, not a vector store. **Weekly handbook.** “What is the refund window today?” The handbook moves. That is RAG, or a refuse. Fine-tuning the old handbook into the model is how a dead policy survives a launch. **Overdue invoices.** “Chase unpaid invoices this week, draft only, no new discount.” That is an agent under a charter. Retrieval may supply the amount. The charter supplies the ceiling. ## Not a ladder Doing all three is allowed. Doing them in a fashionable order is not a strategy. Fail the lookup: you need RAG, or a refuse, not a bigger model. Fail a stable format that never needs a document: fine-tuning can be the job. A weekly handbook is not that job. Fail the ceiling: you need a charter, not another index. The mixing tells I see in the same week: - “We fine-tuned on our docs.” The docs changed on Thursday. The checkpoint did not. That is a filing cabinet with no date stamp. - “The agent has RAG.” Meaning a plugin list. Nobody can name the chunk. That is a worker with a search box, not retrieval. - “We added agents.” The system still answers one turn and cannot refuse a tool call. That is a chatbot that inherited a key. A vector store is furniture. GRAIN, on this site, is the inspectable retrieval loop. Fine-tuning is not GRAIN. An agent without Inspect still answers from the closest wrong paragraph. Do not restage the five moves here. One ownership line is enough: **Navigate owns freshness**, not the next lookup. TTL, re-ingest, change detection. If the index still holds last month’s PDF, ranking harder will not save you. That is a clock problem, not a model-size problem. ## A proposal you can fail If a vendor sells one retainer for all three, ask which artefact you will hold on Friday: a locator you can open, a checkpoint with an eval, or a charter with a named reviewer. Missing all three, you bought a demo. Ask the room to fill four lines on one live system. 1. Which job failed last week: lookup, a stable skill, or the ceiling? 2. What artefact will exist on Friday? 3. Who can open it, or fail it? 4. What does the system refuse when the artefact is missing? If line two is “a better model,” you have not chosen a job. [The RAG Engineer](https://lenvanderhof.com/books/the-rag-engineer/) is live in English and Dutch paperback, hardcover, and Kindle. It does not turn fine-tuning into a source of truth. Write which job failed. Then stop stacking rungs. ## What this page is not This page is not the RAG definition. That page is the named passage. This page is not the agent definition. That page is the three tests and the charter. This page is not a fine-tuning cookbook and not a bake-off. Training stacks and agent runtimes rename themselves. The three jobs do not. Pick the job that failed. Hold the artefact. Keep the other two off the slide until they have a Friday object of their own. Cite this:RAG vs fine-tuning vs agents: three jobs, not a ladder.Len P. van der Hof. [https://lenvanderhof.com/en/blog/rag-vs-finetuning-vs-agents/](https://lenvanderhof.com/en/blog/rag-vs-finetuning-vs-agents/) · Published 3 September 2026. ## Terminology - [RAG](https://lenvanderhof.com/glossary/rag/) - [AI agent](https://lenvanderhof.com/glossary/ai-agent/) - [GRAIN](https://lenvanderhof.com/glossary/grain/) ## Sources 1. [What is RAG?](https://lenvanderhof.com/en/blog/what-is-rag/) 2. [What is an AI agent?](https://lenvanderhof.com/en/blog/what-is-an-ai-agent/) 3. [Using AI agents effectively](https://lenvanderhof.com/en/blog/using-ai-agents-effectively/) 4. [Evaluate the reasoning, not the fluency](https://lenvanderhof.com/en/blog/evaluate-reasoning-not-fluency/) 5. [The RAG Engineer](https://lenvanderhof.com/books/the-rag-engineer/) ## Further reading - [What is RAG?](https://lenvanderhof.com/en/blog/what-is-rag/) - [What is an AI agent?](https://lenvanderhof.com/en/blog/what-is-an-ai-agent/) - [The RAG Engineer](https://lenvanderhof.com/books/the-rag-engineer/) About the author ## [Len P. van der Hof](https://lenvanderhof.com/en/authors/len-p-van-der-hof/) Entrepreneur, AI Innovator and Venture Builder Len P. van der Hof builds practical AI systems, digital ventures and evidence-informed tools for founders. ```json { "@context": "https://schema.org", "@graph": [ { "@type": "Person", "@id": "https://lenvanderhof.com/#person", "name": "Len P. van der Hof", "alternateName": [ "Len van der Hof", "L.P. van der Hof", "Leendert Pieter van der Hof" ], "honorificSuffix": "MSc", "url": "https://lenvanderhof.com/", "image": [ "https://lenvanderhof.com/photos/len-portrait-1.jpg", "https://lenvanderhof.com/photos/len-portrait-2.jpg", "https://lenvanderhof.com/photos/len-portrait-3.jpg", "https://lenvanderhof.com/photos/len-portrait-4.jpg", "https://lenvanderhof.com/photos/len-speaking.jpg", "https://lenvanderhof.com/photos/len-hero.jpg" ], "jobTitle": "Entrepreneur, AI Innovator and Venture Builder", "description": "Len P. van der Hof, MSc, is a Dutch entrepreneur and AI innovator in Zwijndrecht. He builds ReasonKit, MindSesh, Undominated.ai, books under his name, the fiction imprint LPH98.lifestyle, and technology ventures through LPH98.ventures. Eleven titles in Systems for the Strategic Self are available now, in English and Dutch.", "address": { "@type": "PostalAddress", "addressLocality": "Zwijndrecht", "addressCountry": "NL" }, "alumniOf": { "@type": "CollegeOrUniversity", "name": "Rotterdam School of Management, Erasmus University" }, "knowsAbout": [ "Artificial intelligence", "AI agents", "Agentic AI systems", "LLM routing", "SEO", "Generative engine optimization", "Answer engine optimization", "Venture building", "Founder performance", "Founder psychology", "Evidence-based decision-making" ], "sameAs": [ "https://www.linkedin.com/in/lenvanderhof/", "https://x.com/LenvanderHof", "https://www.youtube.com/channel/UCTG20buKqYYbitqqf7l3zJA", "https://www.instagram.com/Lenvanderhof/", "https://www.threads.com/@lenvanderhof", "https://github.com/Lenvanderhof", "https://huggingface.co/LPH98", "https://www.npmjs.com/~lenvanderhof", "https://www.goodreads.com/author/show/70983905.Len_P_van_der_Hof", "https://www.amazon.com/author/lenvanderhof", "https://www.bol.com/nl/nl/b/len-p-van-der-hof-msc/609879394/", "https://bsky.app/profile/lenvanderhof.com", "https://mastodon.social/@Lenvanderhof", "https://crates.io/users/Lenvanderhof", "https://cursor.com/@Lenvanderhof", "https://medium.com/@Lenvanderhof", "https://gitlab.com/Lenvanderhof", "https://hub.docker.com/u/lenvanderhof/", "https://dev.to/lenvanderhof", "https://www.facebook.com/Lenvanderhof", "https://soundcloud.com/Lenvanderhof" ], "affiliation": [ { "@id": "https://lenvanderhof.com/#publisher" }, { "@id": "https://lenvanderhof.com/#mindsesh" }, { "@id": "https://lenvanderhof.com/#lifestyle" } ] }, { "@type": "WebSite", "@id": "https://lenvanderhof.com/#website", "url": "https://lenvanderhof.com/", "name": "Len P. van der Hof", "description": "Len P. van der Hof, MSc, is a Dutch entrepreneur and AI innovator in Zwijndrecht. He builds ReasonKit, MindSesh, Undominated.ai, books under his name, the fiction imprint LPH98.lifestyle, and technology ventures through LPH98.ventures. Eleven titles in Systems for the Strategic Self are available now, in English and Dutch.", "inLanguage": [ "en", "nl" ], "publisher": { "@id": "https://lenvanderhof.com/#person" } }, { "@type": "Organization", "@id": "https://lenvanderhof.com/#publisher", "name": "LPH98.ventures", "url": "https://lph98.ventures", "founder": { "@id": "https://lenvanderhof.com/#person" } }, { "@type": "Organization", "@id": "https://lenvanderhof.com/#mindsesh", "name": "MindSesh", "url": "https://mindsesh.net", "founder": { "@id": "https://lenvanderhof.com/#person" } }, { "@type": "SoftwareApplication", "@id": "https://lenvanderhof.com/#reasonkit", "name": "ReasonKit", "url": "https://reasonkit.sh", "creator": { "@id": "https://lenvanderhof.com/#person" } }, { "@type": "SoftwareApplication", "@id": "https://lenvanderhof.com/#undominated", "name": "Undominated.ai", "url": "https://undominated.ai", "creator": { "@id": "https://lenvanderhof.com/#person" } }, { "@type": "Organization", "@id": "https://lenvanderhof.com/#lifestyle", "name": "LPH98.lifestyle", "url": "https://lph98.lifestyle", "founder": { "@id": "https://lenvanderhof.com/#person" } }, { "@type": "ImageObject", "@id": "https://lenvanderhof.com/en/blog/rag-vs-finetuning-vs-agents/#primaryimage", "url": "https://lenvanderhof.com/media/generated/blog-hero-rag-vs-finetune-vs-agents-v1.86b37f8dd3d7.wide.webp", "contentUrl": "https://lenvanderhof.com/media/generated/blog-hero-rag-vs-finetune-vs-agents-v1.86b37f8dd3d7.wide.webp", "representativeOfPage": true }, { "@type": "BreadcrumbList", "@id": "https://lenvanderhof.com/en/blog/rag-vs-finetuning-vs-agents/#breadcrumb", "itemListElement": [ { "@type": "ListItem", "position": 1, "name": "Home", "item": "https://lenvanderhof.com/" }, { "@type": "ListItem", "position": 2, "name": "Blog", "item": "https://lenvanderhof.com/en/blog/" }, { "@type": "ListItem", "position": 3, "name": "AI Systems", "item": "https://lenvanderhof.com/en/blog/category/ai-systems/" }, { "@type": "ListItem", "position": 4, "name": "RAG vs fine-tuning vs agents: three jobs, not a ladder", "item": "https://lenvanderhof.com/en/blog/rag-vs-finetuning-vs-agents/" } ] }, { "@type": "WebPage", "@id": "https://lenvanderhof.com/en/blog/rag-vs-finetuning-vs-agents/#webpage", "url": "https://lenvanderhof.com/en/blog/rag-vs-finetuning-vs-agents/", "name": "RAG vs fine-tuning vs agents: three jobs, not a ladder", "description": "RAG looks up a named passage. Fine-tuning changes weights for a stable skill. An agent holds a goal and uses tools under a charter. They are three jobs. Mixing them is how a chatbot inherits an API key and still cannot open the source.", "isPartOf": { "@id": "https://lenvanderhof.com/#website" }, "primaryImageOfPage": { "@id": "https://lenvanderhof.com/en/blog/rag-vs-finetuning-vs-agents/#primaryimage" }, "breadcrumb": { "@id": "https://lenvanderhof.com/en/blog/rag-vs-finetuning-vs-agents/#breadcrumb" }, "inLanguage": "en-GB" }, { "@type": "BlogPosting", "@id": "https://lenvanderhof.com/en/blog/rag-vs-finetuning-vs-agents/#article", "mainEntityOfPage": { "@id": "https://lenvanderhof.com/en/blog/rag-vs-finetuning-vs-agents/#webpage" }, "headline": "RAG vs fine-tuning vs agents: three jobs, not a ladder", "description": "RAG looks up a named passage. Fine-tuning changes weights for a stable skill. An agent holds a goal and uses tools under a charter. They are three jobs. Mixing them is how a chatbot inherits an API key and still cannot open the source.", "datePublished": "2026-09-03T07:00:00.000Z", "dateModified": "2026-09-08T23:00:00.000Z", "author": { "@id": "https://lenvanderhof.com/#person" }, "publisher": { "@id": "https://lenvanderhof.com/#person" }, "image": [ "https://lenvanderhof.com/media/generated/blog-hero-rag-vs-finetune-vs-agents-v1.86b37f8dd3d7.square.webp", "https://lenvanderhof.com/media/generated/blog-hero-rag-vs-finetune-vs-agents-v1.86b37f8dd3d7.landscape.webp", "https://lenvanderhof.com/media/generated/blog-hero-rag-vs-finetune-vs-agents-v1.86b37f8dd3d7.wide.webp" ], "articleSection": "AI Systems", "inLanguage": "en-GB" } ] } ```