Every few years a label arrives that is simultaneously a real shift and a marketing sticker. “Agentic AI” is this cycle’s: underneath the noise is the most useful change in what software can do for a small firm in a decade, the move from AI that answers to AI that works. This page is the map of the cluster; the four guides beneath it go deep.
The shift in one example
Ask a chatbot “which of our invoices are overdue?” and, at best, you get advice about credit control. An agent, connected to your systems and given the job, checks the ledger, identifies the overdue accounts, drafts each chasing email in your voice, queues them for a human’s approval and reports what it did. Same intelligence, different posture: the model plans steps, uses tools, and carries the task rather than the conversation.
That is the whole definition worth having: planning, tool use, multi-step execution, under supervision. Everything else is elaboration, and the cluster provides it in a deliberate order:
- What is an AI agent? The plain-English anatomy and the vocabulary vendors will use at you.
- Agentic workflows in practice. The five patterns where agents genuinely pay in a service firm.
- Governance and risk. The six controls that keep an acting system defensible.
- Getting started. The first-agent playbook, from choosing the process to the month-end decision.
Why it matters to service firms
Professional service firms are made of exactly the work agents are good at absorbing: recurring, rule-shaped, judgement-adjacent processes. Enquiry triage. Client onboarding chases. Reconciliation. Report assembly. Monitoring the sources that matter and surfacing only what needs a person. The agentic workflows guide works through these patterns concretely, including what this site’s own operations automate daily, because these pages are written from a business that runs on agents, not from a briefing document.
The part the vendors skip
Agents fail differently from chatbots: a chatbot’s mistake is a wrong sentence, an agent’s mistake is a wrong action. That asymmetry is why governance is the core of this subject rather than its appendix:
- Permissions bounded: an agent’s tool access defines its blast radius
- Actions logged: every step inspectable after the fact
- Approval gates exactly where consequences live
- Caps on spend and volume, because agents can also fail by succeeding too much Gartner’s much-cited prediction that over 40% of agentic AI projects will be cancelled by the end of 2027 for cost and unclear value describes avoidable failure, and the risk and governance guide is the avoidance manual, written for firms that answer to regulators.
The sane adoption path
The single most useful sentence in this cluster: a first agent’s every output is a draft a person approves. It does the full job, nothing acts on its own, and a month of real traffic produces the evidence at zero action risk. Firms that skip this step supply the cancellation statistics.
The pattern that works is unheroic: pick one narrow process that hurts, give an agent the job with every output held as a draft for approval, keep a human on the gate, measure honestly, then widen. The getting started guide turns that into a step-by-step path, including the build-versus-buy question and what a realistic first month looks like.
Related reading elsewhere in the knowledge base: multi-agent routing for how larger agent systems are structured, and the internal AI assistant playbook for the adjacent build most firms want first.
Where I fit in
My business is run this way: agents build the software, watch the systems, draft the content and chase the routine work, under written specifications, security gates and human review. That system is productised as the Software Factory, and the same discipline scaled down is how a first agent lands safely in your firm via a Kick-starter. If you want to see agents doing real work before believing any of this, the Labs page is the evidence.