ChatGPT in business: from draft to finished work
A practical method to get the most out of ChatGPT: choose a task, prepare the context, check the result and build it into the team’s work.
Read the guide ↗GUIDES / UNDERSTAND TO DECIDE
The right questions.
Answers you can use.
Install local AI. Get your tools working together. Make better use of ChatGPT or Claude. Each guide starts from a concrete problem, with examples and a possible next step.
Installation, hardware and the hybrid option.
TEAMSWe have the tool. Now what?Uses, review and adoption.
INTEGRATIONI want to connect my software.Workflows, agents, APIs and MCP.
Claude for teams: organizing use and access
Move from individual experiments with Claude to shared use: business examples, documents, roles, review and handover.
Read the guide ↗Installing local AI in business: the useful steps
From first use to running local AI: model, interface, hardware, documents, network access, backup and offline testing.
Read the guide ↗AI agents in business: chatbot, workflow or agent?
Understand what an AI agent can bring to a business and tell apart conversation, defined automation and tool choice by a model.
Read the guide ↗Integrating AI into a CRM or ERP without replacing everything
How to prepare an AI integration: map the exchanges, manage identifiers and duplicates, test a workflow, then hand over operations.
Read the guide ↗RAG and access rights: who can find which document?
Design an AI document search (RAG) that takes into account identities, permissions, versions and document withdrawal.
Read the guide ↗RAG, fine-tuning and memory: three ways to adapt an AI
Choose between document search, model adaptation and application memory, depending on the problem to solve in your business.
Read the guide ↗MCP in business: connecting an agent to its tools
Understand the Model Context Protocol, its role in an AI integration and the checks to plan before connecting business software.
Read the guide ↗Which machine for local AI in business?
Size an AI workstation or server from the task, memory, users, latency and operations.
Read the guide ↗Local AI speeds, machine by machine
Community-published speeds for the same small model: Apple chips, DGX Spark, Ryzen AI and common graphics cards, with the source of each figure.
Read the guide ↗Local AI or cloud: choosing a useful hybrid architecture
Compare local AI, cloud and hybrid setups for a business by data, performance, running costs and continuity.
Read the guide ↗Changing AI models without breaking existing work
Compare a new model at task level: formats, tools, answers, latency, cost, data and rollback to the previous version.
Read the guide ↗Measuring the gain from AI automation without forgetting the corrections
Compare time and quality before and after an automation: representative cases, review, exceptions, costs and maintenance load.
Read the guide ↗Forward Deployed Engineering: AI integration in the field
What Forward Deployed Engineering means for a business: observing the work, building within real constraints and handing over the AI integration.
Read the guide ↗Visibility in Google and AI search engines: the useful basics
Distinguish indexing, ranking and citation in AI answers. Accessible pages, useful information, internal links and Search Console monitoring.
Read the guide ↗Autonomous agent or supervised workflow: choosing the level of control
Choose between rules, workflow and AI agent according to the decisions to be made, then place approvals where an error really changes the outcome.
Read the guide ↗Auditing AI flows: following a case and its data
Find the destinations, accounts and copies in a flow that uses AI, then choose the fixes that make it understandable.
Read the guide ↗Automating quotes and invoices: from the source document to a checked draft
Keep extraction, reference data, calculation and approval separate to prepare quotes or invoices without inventing missing details or duplicates.
Read the guide ↗Before you buy AI
The integration, data and skills difficulties seen in business surveys, and the choices they suggest.
Read the guide ↗AI agent cost: count the cases actually accepted
Work out the cost of an AI agent workflow, including calls, retries, human work and operations, then relate it to the results you can actually use.
Read the guide ↗Train a team in AI: learn a task you can pass on
Build an AI workshop around a business deliverable, compare the results and leave the team examples it can take over and maintain.
Read the guide ↗AI governance: turn responsibilities into real access
Define who can read, prepare and execute, with identifiable accounts, useful audit trails and a testable way to interrupt the AI workflow.
Read the guide ↗AI in an SME: choosing the first workflow and the people
Select a first AI use in an SME from existing work, the access available and the people who will prepare, use and accept the result.
Read the guide ↗A company intelligence of your own: organizing business memory
Build a governed business memory: sources, validated decisions, rights, dates and corrections, without confusing knowledge with model training.
Read the guide ↗Updating cases when new documents arrive
Organize a documented review: effective dates, new documents, compared proposals and validation before the reference changes.
Read the guide ↗An AI machine you can operate remotely, under your control
Use a local or hybrid AI machine from authorized workstations and plan for on-demand remote assistance, with limited, revocable access.
Read the guide ↗Remote maintenance of local AI: access and responsibilities
Prepare support for a local AI installation: temporary access, scope of work, data visibility, recovery and on-site handover.
Read the guide ↗An AI model is not a backup of your business
Separate model weights, documents, business memory and configuration to prepare a version of your AI installation that can really be restored.
Read the guide ↗Small local model and invoices: preparing extraction, keeping control
Test local invoice extraction with sourced fields, amount checks and validation, without replacing accounting software or giving tax advice.
Read the guide ↗