What actually makes AI work across business systems?
Chatbots are familiar. Connecting an AI capability to business records, decisions and operational processes introduces different questions about information, integration, reliability, accountability and cost.
Several technologies, each with a different job.
Businesses do not have to adopt all of these capabilities. The design should depend on the process and the consequence of an error, not on how autonomous the technology appears.
| Building block | What it does in plain English | Question for the business |
|---|---|---|
| Model and structured output | Interprets language or documents and returns a response that other software can use. | How will the output be checked? |
| Retrieval and context | Supplies relevant, authorised information, such as a current policy or product record. | Which source is authoritative? |
| APIs and connectors | Allow systems to read or exchange business information under defined access permissions. | What may be read, changed or disclosed? |
| Workflow orchestration | Runs repeatable steps, manages exceptions and records what happened. | Who owns the process when it fails? |
| Agent capabilities | May select permitted tools or next steps where the route is not predictable. | Is that additional freedom necessary? |
| Security and oversight | Constrains access and actions, provides human approvals and monitors outcomes. | Who can authorise and review decisions? |
Practical rule: when the steps are predictable, a defined workflow with AI used for selected tasks is usually preferable to an agent with broad discretion. The technology should reflect the task rather than become the objective.
Beyond the Chatbot
The 14-page white paper explains the technologies behind integrated AI, workflows and agents, including examples, risk considerations and a plain-English technical glossary cross-referenced throughout the main text.
- Founders: implications, cost and control
- Executives: governance and commercial decisions
- Technology leaders: integration and architecture choices
Architecture follows the business requirement.
Technical Foundations complements the seven Practical AI articles and the three tools used to assess opportunities, readiness and benefits.
