IAfluence · prepared on 2 October 2026
A one-off request: start with a general-purpose assistant
Drafting text, exploring an idea or rephrasing does not necessarily need a dedicated application. A general-purpose assistant may work if it can receive the data and a person checks its answer.
Your documents are the reference: consider RAG
EchoChat’s documented need is to query website content. Exploration and indexing provide the model with information from that corpus. This illustrates an architecture, not proof that every answer is correct. Test questions with known answers, conflicting sources and missing information.
Actions in tools are needed: consider an agent
Reading, choosing a step and calling a tool requires more controls than a text answer. Define possible actions, permissions, operation records and human approvals. Cognitive Enterprise OS explores orchestration; its prototype status does not prove operational client deployment.
For an SME: test one task before expanding
Choose a repeated task, define an acceptable result and collect representative examples. Compare the current method and prototype on the same examples. Measure quality, errors, running cost and checking time before claiming any benefit.
MCP: connect tools with explicit permissions
MCP is a protocol connecting AI applications to external sources or tools. It may simplify integration but does not determine user permissions or result quality. In a project, those rules are designed and tested separately.
Before getting started
What is an AI consultant?
An AI consultant helps define a need, assess tools and data, choose an approach and test its suitability. At IAfluence, advice may precede training, a proof of concept or development of a solution.
How much does an AI consultant cost?
IAfluence offers focused consulting at EUR 75/hour with no ongoing commitment, in 1, 2, 3 or 5-hour sessions. This is not a development price: training and custom solutions are quoted separately.
How much does an AI proof of concept cost?
There is no single price. Cost depends on scope, data, access and integrations. The first discussion defines what must be demonstrated before a quote is prepared.
What is RAG?
RAG retrieves passages from a collection and supplies them to a language model to prepare an answer. It helps when answers depend on your documents, but sources and answer quality must be checked.
What is an AI agent?
An agent organises steps and calls tools to perform a task. Permissions, allowed actions and approval points must be defined; autonomy does not guarantee reliability.
How can a small business integrate AI?
Choose a repeated task, define an acceptable result and collect representative examples. Compare the current method and prototype on the same examples. Measure quality, errors, running cost and checking time before claiming any benefit.
How does ChatGPT differ from a business AI assistant?
ChatGPT is a general-purpose assistant. A business assistant is designed for a defined task, using an organisation’s documents, tools, rules and controls. Adding an interface is not enough: access, sources and expected results must be tested.
When should you use RAG?
When an answer needs to rely on an identified document collection and retrieving relevant passages has practical value. A small collection can sometimes be supplied directly to the model; RAG adds a retrieval step that needs evaluation.
How long does an AI prototype take to build?
There is no universal timeframe. A prototype for one task with accessible data differs from a tool connected to several systems. IAfluence defines scope, access requirements, tests and the schedule before committing to a deadline.
Sources and documented examples
Examples refer to IAfluence prototypes, not client outcomes.