Answers before you buy, and support once you're using AI SMITH.

An AI agent is software designed to understand a request, use your approved company knowledge, and carry out or prepare part of a business workflow — such as answering a question, drafting a follow-up, or preparing a summary — with a person approving sensitive steps. See our Platform page for the full explanation.

A chatbot typically replies inside a single conversation. A simple automation follows a fixed rule with no judgement. An AI agent is designed to work across a workflow — gathering context, drafting outputs, handing off to another agent or a person, and being measured against a business outcome — while still respecting defined boundaries and human approval.

It's the coordinated environment where multiple specialised agents share the same approved company knowledge, connect to your workflows and tools, and stay under human governance — rather than operating as separate, disconnected tools. Full definition on the Platform page.

Start with the workflow that repeats often, has a clear owner, and can be measured — not necessarily the most impressive one. Visit Solutions for an outcome-based selector, or book a consultation and we'll help you identify it together.

Yes. Most businesses start with one workflow or one agent, prove the value, and expand only once the evidence supports it.

Access to the workflow owner, the documents and rules the agent should follow, and a way to measure your current baseline. The exact list depends on your workflow and is confirmed during the Assess stage — see How It Works.

Yes — agents are designed to work from your approved SOPs, policies, product facts, and templates. You review and approve what's included before it's used, and you can update it as your business changes.

Connections fall into three categories: verified (already confirmed for your account), custom (built specifically for your systems), and planned (not yet available). We do not claim an agent connects to every system — we confirm what's genuinely available during scoping.

Sensitive actions — pricing, external communication, financial postings, refunds, and employment or contractual decisions — always require a person's approval. Each agent page sets out its specific human-control boundaries; see Platform for the general governance model.

Agents are designed to recognise uncertainty and escalate to a person rather than guess. When something is wrong, the correction is captured and used to improve the workflow. We don't claim error-free output — human review is part of the design, not an afterthought.

Against the baseline you start with — time, quality, throughput, service, or commercial impact, depending on the workflow. See the value framework on the Pricing page and the measurement stage in How It Works.

Complexity, volume, how ready your knowledge already is, custom connections, number of languages, risk and review requirements, service level, and overall deployment scope. See Pricing & Plans for the full explanation — pricing is determined by scope, not a fixed package.

Partner roles (Entrepreneur, City, Strategic) involve an ongoing, defined commercial relationship with responsibilities and enablement. The Referral Program is for occasional introductions from customers, professionals, or advocates, recognised under published rules, without a territory or delivery role.