Applied AI · Practical guide

When should you automate a process, and when should you use AI?

Clear rules can be handled with conventional software. Tasks that depend on patterns may need a model. Choosing carefully avoids complexity that adds no value.

Explore the guide
Conceptual schematic. No client data or metrics.

The central idea

Start with what the task needs.

Automation means performing a task with less manual intervention. AI is one possible tool, not a requirement. Describe the input, the decision and what should happen when something goes wrong before choosing technology.

At T2G we start with the task, its data and operating conditions. We agree on a scope that can be checked and explain technical decisions throughout the work.

Step by step

Four decisions before automating.

Four points to organise the work. Each diagram supports the explanation; the process adapts to its context.

  1. 01 / 04

    Observe the real task

    Identify what a person receives, checks and delivers. Include exceptions such as an incomplete document or a new client. If the process changes depending on who performs it, clarify the workflow first.

    Input · Decision · Outcome
  2. 02 / 04

    Separate rules from interpretation

    An unambiguous condition is often easier to test and maintain as a rule: flag missing records or assign a date. Interpreting an image or classifying documents with varying formats may require learning from examples.

    Clear rule · Variation · Model
  3. 03 / 04

    Compare with a simple baseline

    Try an organised manual workflow, a rule or a small automation before training a model. Compare review effort, types of error and maintenance. The useful solution may combine rules with a model estimate.

    Manual · Rules · AI
  4. 04 / 04

    Decide what a person reviews

    Agree when the process should stop, who handles uncertain cases and how results are corrected. A limited initial scope lets you check whether the automation helps in real conditions.

    Automate · Review · Correct

An explained example

Classifying and reviewing accounting documents

01 / Starting point

A person opens each file, identifies the client and checks the period’s records.

02 / An approach

Rules can check dates and required fields. A model could suggest a document category, with ambiguous cases routed for review. AI does not need to take over the whole task.

For your decision

What needs to be clear.

01

Rules

Stable, testable conditions understood by the business.

02

Models

Variation that can be learned from representative examples.

03

People

Exceptions, consequential decisions and process validation.

Common questions

Questions before getting started.

Does automation mean replacing people?

It can organise work and reduce repetitive effort while preserving human review and responsibility where needed.

Can we start without AI?

Yes. Clear workflows and simple rules may address much of the problem while creating better organised data for later evaluation.

Keep exploring

Connect this guide with your challenge.

Let’s discuss your context

Let’s define a useful first step.

We review the problem, available information and constraints with you to assess where it makes sense to start.

Book a diagnostic session directly with the partners