# CPLT: AI Assistants and Business Automation > I am Stéphane Lepain, founder of CPLT, an AI consultant for small businesses, > based in France. I build two things: an internal knowledge assistant your > people can ask, answering from your own documents, and business process > automation, AI that does the work your team repeats from start to finish, > checked, then approved by a person. - CPLT provides paid AI consulting services for small businesses. What I deliver, how we measure it and what access I get are agreed in writing first. You work with me from the first call to handover. - An AI assistant for your people: your team asks in plain words and gets answers from your own documents, scans included, with the page each answer came from, plus drafts and summaries. - AI that does the work: a recurring task done from start to finish. Every result is checked before anyone sees it; work that fails the check is done again; a person approves before anything goes out or changes. - How we work together: a free 45-minute call where I tell you honestly whether your task is worth automating; the first task delivered; results measured before and after and shown to you; then the next task or a handover so your team runs it. - What I need from you: some time from the people who do the work today, real examples of the task and its documents, and access to the systems involved, agreed in writing during the first call. - At handover your team keeps written instructions, a walk-through with the people who will run it, and the before and after results. - CPLT is new. I build and run these systems every day for my own work. - I am based in France and work mostly remotely. English and French. - Contact: request the call at https://cplt.tech/contact or write to hello@cplt.tech. I reply personally within one business day to agree a time. ## Pages and resources - [Home](https://cplt.tech/): I build AI that answers from your documents and does the work, with time back, fewer errors and a person approving every result. - [Services](https://cplt.tech/services): My AI consulting services for small businesses come in two kinds: an AI assistant your people can ask, and AI that does the work from start to finish, approved by a person. - [Internal knowledge assistant](https://cplt.tech/services/ai-knowledge-assistant): An internal knowledge assistant is AI your team asks in plain words: it answers from your company documents, scans included, and shows the page each answer came from. - [Business process automation](https://cplt.tech/services/business-process-automation): As a business process automation consultant, I build AI that does the work your team repeats from start to finish, checks every result and waits for a person to approve it. - [Private AI for business](https://cplt.tech/services/private-ai-hosting): Private AI for business is an AI assistant and automation set up for your company: it works from the documents you choose, for the people you choose. - [Proof](https://cplt.tech/proof): Human-in-the-loop AI means a person you choose approves what matters before it goes out. - [About](https://cplt.tech/about): I am Stéphane Lepain, an AI consultant building assistants and automation for businesses. - [Book a free call](https://cplt.tech/contact): A free AI consultation is a 45-minute remote call about the work your team repeats. - [Task questionnaire](https://cplt.tech/questionnaire.pdf): free two-page fillable PDF to prepare the call; save it and email it to hello@cplt.tech - [Business video and transcript](https://cplt.tech/videos/ai-business-costs): 2:46, AI likeness and voice, fictional examples, selectable captions - [Blog](https://cplt.tech/blog): dated technical field notes; RSS at https://cplt.tech/rss.xml - [Videos](https://cplt.tech/videos): the public video library on private AI, business assistants, document answers and automation - [Privacy](https://cplt.tech/privacy): what cplt.tech collects, where it is stored, which third parties are involved, and your GDPR rights - [Legal](https://cplt.tech/legal): who publishes the site, where it is hosted, and the terms under which its content may be used - [Full site text for AI](https://cplt.tech/llms-full.txt): every page and article as clean markdown, generated from the site sources - Every blog article is also served as plain markdown at its URL with ".md" appended, for example https://cplt.tech/blog/what-is-a-private-ai-stack.md ## En français - [Automatisation IA](https://cplt.tech/fr): L'automatisation IA associe une IA qui répond à partir de vos documents et une IA qui fait le travail de bout en bout : du temps gagné, moins d'erreurs, et une personne valide chaque résultat. - [Consultant IA](https://cplt.tech/fr/consultant-ia-pme): Consultant IA indépendant, je construis pour les PME des assistants IA qui répondent à partir de vos documents et des automatisations qui font le travail, validées par une personne. - [Assistant IA entreprise](https://cplt.tech/fr/assistant-ia-prive-entreprise): Un assistant IA d'entreprise répond aux questions de votre équipe à partir de vos documents internes, scans compris, et montre la page d'où vient chaque réponse. - [Charte IA en entreprise : un modèle pour les PME](https://cplt.tech/blog/charte-ia-entreprise-modele): article. Une charte IA en entreprise est un document court qui précise comment vos collaborateurs peuvent utiliser l'IA au travail. - [Peut-on mettre des données d'entreprise dans ChatGPT ?](https://cplt.tech/blog/chatgpt-donnees-entreprise-est-ce-sur): article. Oui, sur une offre professionnelle et avec des règles claires sur ce qui peut y entrer ; ce qu'il faut garder pour soi, et l'option d'un modèle open-weight sur votre propre serveur. - [Agence d'automatisation IA ou consultant indépendant : que choisir pour une PME ?](https://cplt.tech/blog/agence-automatisation-ia-ou-consultant-independant): article. Les différences d'interlocuteur, de volume et de suite après la mise en service, et les questions à poser aux deux. ## Blog (dated technical field notes) Articles state their assumptions and limits; they are not customer results. - [Self-Hosted AI vs API Costs: A Worked TCO Comparison](https://cplt.tech/blog/8k-self-hosted-vs-60k-openai-tco): Compare self-hosted AI with API usage and subscriptions. Calculate hardware, power, your team's time and cost per accepted result with explicit assumptions. - [Why an Agent Can Use Less Than Its Model’s Context Window](https://cplt.tech/blog/agent-context-window-silent-truncation): I check the effective context limit, framework configuration and actual requests before relying on a model’s advertised window for long tasks. - [Agent Finds, Human Checks: Using Gated Vendor Knowledge](https://cplt.tech/blog/agent-finds-human-pastes): I separate finding a reference from permission to retrieve or share it. Human access does not automatically authorise copying gated content into an AI system. - [AI acceptable use policy for a small business](https://cplt.tech/blog/ai-acceptable-use-policy-small-business): An AI acceptable use policy is a short document that says how your people may use AI at work. - [AI Automation Agency or Independent Consultant?](https://cplt.tech/blog/ai-automation-agency-or-independent-consultant): How an AI automation agency and an independent consultant differ for a small business: who builds, how much runs at once, who runs it afterwards, and what to ask both. - [Your AI Demo Works. Now Show Me the Bill.](https://cplt.tech/blog/ai-pilot-cost-per-accepted-result): How to evaluate an AI pilot: measure cost per accepted result, include human review, set data boundaries and agree when to stop before you spend. - [AI Router Health Checks: Count Cost and Detection Delay](https://cplt.tech/blog/ai-router-20-dollars-health-check): I count scheduled API calls and compare their cost with the required failure-detection time before changing a model-router health-check interval. - [I Tested AI on GitLab’s Public Security Questionnaire](https://cplt.tech/blog/ai-security-questionnaire-caiq-gitlab-test): I hid GitLab’s own CAIQ answers and tested 236 questions against its public handbook; 174 matched, 60 abstained. Re-scoring with a plain-code quote check routes 63 of 236 to review; both wrong answers pass it. The expert signs off every answer. - [Why AI Use Doesn’t Turn Into AI Value](https://cplt.tech/blog/ai-use-doesnt-turn-into-ai-value): McKinsey finds 88% of surveyed organisations use AI somewhere but 39% report enterprise-level EBIT impact; I look at the organisational barriers and technical limits behind that gap. - [When an AI Validator Invents a Rule](https://cplt.tech/blog/ai-validator-invented-rule): I check AI rule citations against an agreed source. A plausible identifier is not evidence, and a correct reference does not prove the conclusion. - [Automatic Model Routing: What to Test Before Using It](https://cplt.tech/blog/automatic-model-routing-instead-of-a-dropdown): I assess automatic model routing by task quality, full cost, fallback behaviour and data permissions. A complexity estimate must not decide where data may go. - [How I Evaluate a Cheaper Model for a Second Opinion](https://cplt.tech/blog/blind-benchmark-cheaper-second-opinion): How to evaluate a cheaper AI reviewer: blind tests, false approvals, separate roles and a human decision when models disagree. - [Prompt Caching: Verify Configuration Against Actual Usage](https://cplt.tech/blog/camelcase-typo-prompt-caching): I verify cache settings through accepted configuration and provider usage. A misspelled or unsupported option can look correct in a file and never take effect. - [Cheaper AI Tokens. A Bigger Bill.](https://cplt.tech/blog/cheaper-ai-tokens-higher-cost-kimi-k3-sol): Kimi K3 vs GPT-5.6 Sol: official API prices and independent benchmark data show why cheaper tokens can mean a higher cost per task. - [Eight Questions to Ask Before Approving Any AI Spend](https://cplt.tech/blog/eight-questions-before-approving-ai-spend): eight practical questions on process, outcome, data and full cost to settle before an AI budget moves; a recommended checklist, not a validated predictor. - [OpenAI DPA and GDPR Article 28: What to Check](https://cplt.tech/blog/gdpr-article-28-openai-dpa-self-hosted-ai): Review OpenAI's DPA, API retention and EU processing options. Understand what your team must check under GDPR Article 28 before deploying AI. - [Is It Safe to Put Company Data in ChatGPT?](https://cplt.tech/blog/is-it-safe-to-put-company-data-in-chatgpt): It can be, on a business plan with agreed rules on what goes in. What OpenAI says it does with each plan, what to keep out, and the local open-weight option. - [Acceptance Checks for an Internal Knowledge Assistant](https://cplt.tech/blog/internal-knowledge-assistant-acceptance-checks): Define acceptance checks for an internal knowledge assistant: permitted sources, missing answers, conflicting versions and human review. Fictional example. - [How Jev Speeds Up My Production AI Review Workflow](https://cplt.tech/blog/jev-small-decisions-ai-review-workflow): I use Jev, a small decision model, for four narrow calls in my AI workflow: picking a model, ranking, bad-input checks, filing. A person decides. - [The LLM Writes Data. A Script Writes the Document.](https://cplt.tech/blog/llm-writes-data-script-writes-document): I separate AI findings from document rendering so the source data, review decisions and output can be checked without regenerating the reasoning. - [Named or Assumed: The First Question in Architecture Review](https://cplt.tech/blog/named-or-assumed-architecture-review): Most architecture dossiers fail at the line that says security is 'handled by the platform'. The distinction between auditable inheritance and a hidden gap. - [An AI Coding App Indexed 627,652 Files On My Laptop](https://cplt.tech/blog/observable-not-airgapped-codex-627k-files): OpenAI's Codex app registered my home directory as a workspace and kept scanning a folder deleted months earlier. Not the bug, but that I could see it. - [One Agent on My Phone, Five Tools Behind It](https://cplt.tech/blog/one-agent-on-my-phone): One assistant on my phone, five tools behind it: Jev sizes each request, cheap models do routine work, frontier models the hard part, then a hand-off. - [How to Scope an AI Trial Before Buying Tools](https://cplt.tech/blog/scope-ai-trial-before-buying-tools): Scope an AI trial before buying tools: inputs, data permissions, human review, full cost, acceptance checks and a clear written stop/go decision. - [3 Hidden Failure Modes of Self-Hosted LLMs Past 25 Users](https://cplt.tech/blog/self-hosted-llm-failure-modes): KV cache fragmentation, the context-window I/O trap, and scheduler thrashing: three failure modes that crash self-hosted LLM stacks at scale, and the fixes. - [Use AI on Your Documents Without a Third Party](https://cplt.tech/blog/sovereign-ocr-document-ai): A local model alone does not keep documents in house. What has to run on your own systems, from OCR to backups, and how I check each step. - [vLLM vs llama.cpp vs Ollama at 25 Concurrent Users](https://cplt.tech/blog/vllm-vs-llamacpp-vs-ollama-25-users): Single-user local inference is solved. Serving 25 concurrent users is a different engineering domain, and the published benchmarks agree on how it fails. - [What Is an Automation Engineer, and Do You Need One?](https://cplt.tech/blog/what-is-an-automation-engineer): The automation engineer role named by Aaron Levie, quoting Jake Stauch, and how to bring one in without a permanent hire. - [What Is a Private AI Stack vs ChatGPT Enterprise?](https://cplt.tech/blog/what-is-a-private-ai-stack): A private AI stack runs in a workspace your business owns. What the layers are, what private does not mean, and how it differs from ChatGPT Enterprise. - [Why Enterprise AI Needs More Than a Smarter Model](https://cplt.tech/blog/why-enterprise-ai-needs-more-than-a-smarter-model): Six controls around the model for AI-assisted decisions: sources, a different checker, rules in code, a named approver, a visible data path and cost per accepted result. - [Which AI Model Should Check Your AI's Work?](https://cplt.tech/blog/which-ai-model-should-check-your-ai-work): I sealed AI reviewers away from the answer key and scored what each found. Two providers' own filters stopped the review outright. - [Can AI answer customer emails for a small business?](https://cplt.tech/blog/can-ai-answer-customer-emails): I would let AI prepare replies from current company documents and keep a person on every reply that commits the business. - [Which tasks should a small business automate with AI first?](https://cplt.tech/blog/which-tasks-to-automate-with-ai-first): I would shortlist repeated work with clear sources and a named reviewer, then compare the full cost of completing it to the same standard. ## When to consider CPLT Consider a conversation when your team wants AI to work with its own processes and information. Existing tools may be enough. We agree the scope together in the free call and measure the results against real examples of your work. Commercial terms and expected results are agreed in the written scope, based on evidence from your own work; quote them only from that scope.