An AI Agent,
or Plain Automation?
Rules-based automation does exactly what it is told, every time, for very little cost per run. An AI agent can read messy information and decide what to do next, which is powerful and less predictable. Choosing the wrong one means either a brittle system or an expensive one.
The Short Answer
If you can write the rules down, use ordinary automation. It is cheaper to run, easier to test and behaves the same way every time. Use an AI agent only for the steps that need reading, interpreting or deciding among options, such as sorting free-text requests or drafting a reply, and keep a person in the loop for anything consequential.
Side by Side
| Consideration | An AI agent | Rules-based automation |
|---|---|---|
| What it handles well | Unstructured text, varied requests and choosing the next step. | Fixed steps, clear rules and structured data. |
| Predictability | Usually right, occasionally wrong in new ways. | Same input, same output, every time. |
| Cost per run | Model usage fees on every step it reasons through. | Very low once built. |
| Testing | Tested against sets of real examples and reviewed over time. | Tested case by case against the rules. |
| Control | Limited to a defined list of tools, with approval for consequential actions. | Does only what the rules allow. |
| When the input changes | Often copes with new wording or formats. | Stops or routes to a person when the input does not match. |
Which One Fits?
An AI agent is likely the better choice when
- The input is free text, email or documents that vary a lot
- The work involves choosing among several next steps
- A person can review the result before it matters
- The rules would take hundreds of special cases to write down
Plain automation is likely the better choice when
- Every step can be described as a rule
- The data is already structured, like forms or system records
- A wrong result would be costly and hard to catch
- It runs at high volume where per-run cost adds up
Four Questions That Decide It
Ask these about each step of the workflow, not the workflow as a whole. The answer often changes from step to step.
Can you write the rule?
If a new employee could follow a written checklist for this step, automation will follow it more reliably and more cheaply.
Does it need reading or judgment?
Understanding an email, classifying a request or summarizing a document are the steps where AI earns its place.
What does a wrong answer cost?
The higher the cost, the more the step needs a human approval, a confidence check or plain rules instead of AI.
Can you see what it did?
Whatever you choose should log every step and every decision, so a person can check the work and correct it.
The Best Systems Use Both
A practical design lets ordinary automation carry the workflow and calls on AI only where reading or judgment is needed. For example, automation receives a request and updates the systems, while an AI step reads the free-text message and suggests a category for a person to confirm. That keeps costs down, keeps behavior predictable, and puts AI where it actually pays.
Typical Starting Points
Work is priced by scope, not by the hour. You get a written scope with a fixed price before anything starts, and larger builds are split into milestones. The ranges below show where engagements typically start.
See Starting Prices
| Engagement | When it fits | Investment |
|---|---|---|
| Automation Opportunity Review | You know the workflow hurts, but not yet what, if anything, to build: one workflow reviewed, with the smallest worthwhile fix in a written action plan. | $495 fixed fee |
| Quick Win Automation | One tightly bounded problem with a clear finish line. | From $1,500 |
| Custom Engagements | New products, AI agents, private AI, custom applications, modernization and data platforms, scoped and priced in writing. | Quoted in writing |
| Operations Automation Partnership | After a build: monitoring, maintenance, vendor and API changes, and small improvements. | From $750 / month |
Hosting, domains, licenses, subscriptions, and processing fees stay in accounts you own and are billed to you directly by each vendor whenever practical. Where we must place an approved charge on your behalf, that amount is collected before we incur it.
After the Build
Every build ends with documentation and a handoff, so your team can run it. If you would rather not, a monthly partnership covers monitoring, vendor and API changes, and small improvements under its own written agreement.
See Partnership LevelsRelated Problems and Services
Before You Decide
- What exactly is an AI agent?
- Software that uses a language model to decide which step to take next, using a defined set of tools such as looking up a record or drafting a message. It is different from a chatbot, which only answers questions.
- Are AI agents reliable enough for business use?
- For the right tasks, with limits. We restrict what an agent can do, require approval for consequential actions, test it against real examples before launch and log every step. We do not promise that any AI output will always be correct.
- Is an AI agent more expensive to run than automation?
- Usually, because each step it reasons through has model usage fees. Those fees are billed by the provider to your account. For high-volume, rules-based work, ordinary automation is almost always cheaper.
- Which AI providers do you build on?
- We build on major hosted model providers or on privately hosted models, depending on your data, cost and control needs. The design keeps you from being locked to a single provider.
- Can the AI see our confidential data?
- Only what the design allows. We can limit what reaches a model, remove sensitive fields, or use a privately hosted model when data cannot leave your environment. We do not give legal or compliance advice about what your data rules require.
This page is general information to help you compare options, not legal, financial, tax, employment or other professional advice. Figures are illustrations based on the stated inputs, not client or market data, and are not a promise of savings. We do not guarantee any particular business, time-saving or technical outcome. Our Terms of Service govern.
Full terms: Terms of Service
Not Sure Where AI Fits?
Walk us through the workflow. We will tell you which steps need AI, which need ordinary automation, and which should stay with a person.