Agents vs chatbots

The practical difference between a chatbot and an agent is not conversational style. A chatbot primarily returns a reply. An agent works toward a goal by reading authorized systems, maintaining task state, selecting tools, producing artifacts, asking for approval, and recording what happened. That extra capability creates value and risk.

When a chatbot is enough

Use a bot when the job is answering bounded questions from approved material, collecting simple intake, or directing a person to the correct resource. Frequently asked questions, opening hours, policy lookup, and basic product navigation do not need autonomous planning. A smaller surface is easier to test and safer to operate.

Even a bot needs source freshness, fallback behavior, and a human escalation path. If an answer can materially affect a contract, payment, health, employment, or account state, retrieval alone does not remove accountability.

When the work requires an agent

An agent becomes useful when completion spans systems and time: qualify a lead, create a CRM proposal, draft mail, wait for approval, update status, and schedule follow-up. It has a role, memory scoped to the task, connector permissions, and a success condition. It creates files and events that a reviewer can inspect.

Writes default to Ask. CRM changes, outbound mail, publishing, payments, and status transitions are proposed before execution. Credentials are brokered server-side. The role runs in a sandbox and its behavior is versioned in git.

Migrate in 30 days without discarding FAQs

Week one inventories existing intents, answer sources, escalation paths, and analytics. Keep the high-performing FAQ surface. Week two selects one multi-step intent and defines its state machine, tools, and approval points. Week three builds the connector and tests historical cases, failures, retries, and permissions. Week four shadows live requests, then enables one approved action.

Measure answer resolution separately from task completion. For the agent path, track successful steps, approvals, corrections, abandonment, escalation, and unintended writes. Do not rename a chatbot “agent” because it calls one API, and do not introduce planning where deterministic routing is enough.

The right architecture often contains both: a simple conversational front door and narrowly scoped roles behind it. Complexity should follow the work, not the category label.

Run it as a role, not a prompt

The durable implementation is a role inside the company operating system. Its instructions, checklists, examples, and connector definitions live in git, so every change has an author, review, and rollback path. Each run gets a sandbox and an auditable record. The team can use the best model for each step instead of tying the workflow to one vendor. Models can change; the role, tests, permissions, and history remain.

Connections are brokered server-side. CRM, document, mail, and accounting credentials never sit in a prompt or a browser extension. Read access is scoped to the records needed for the task. Writes default to Ask: the agent prepares the proposed update, message, or file, then an accountable person approves it. Self-hosting is available when policy, residency, or network boundaries require it.

Make the first pilot measurable

Choose one queue with enough volume to observe within two weeks. Record the current cycle time, rework rate, backlog, and escalation rate before the first run. Test historical cases, including awkward and incomplete ones, before touching live work. During the pilot, compare accepted outputs, corrected outputs, false escalations, and time returned to the team. A useful role becomes more reliable because corrections are committed back to its skill and evaluation set.

Keep the boundary explicit. The agent can collect evidence, apply a checklist, draft, route, and update systems after approval. A named employee owns exceptions and consequences. That division is what turns model capability into dependable operations without pretending that probability is judgment.

This work runs on Neuro OS. To scope a first role, get started.

Explore Neuro OS →