GrowFast Learning

AI in practice

Start with a business problem. Use this free guide to see when a prompt, a workflow, or an agent fits the work.

A short guide to applied AI

  1. Prompt an LLM

    An LLM generates a response from your instructions and the context you supply. Use it for a single task where someone can review the output.

  2. Run a workflow

    A workflow runs a defined sequence of steps. It is useful when the inputs and rules repeat and people know which exceptions to review.

  3. Delegate to an agent

    An agent can choose steps and use permitted tools to work toward a goal. Give it a defined scope, approved sources, and clear points where a person must decide.

Before choosing a tool, define the gap. What decision or task takes too long? Which records can the tool use? Who will check the result? What will show that the change helped?

Try it with an estimating workflow

A contractor wants to know which AI tools estimators use and who can keep the work running if the lead leaves. Compare three ways to investigate.

Prompt an LLM

Draft a summary of one estimating log. The estimator checks the summary against the source before using it.

Discuss this example with Stephen

Questions about the guide

Which approach should I try first?

Choose the simplest approach that fits the work. Start with one task, define how you will check the result, and build from there.

Read the related section
Where does a person need to be involved?

A person should review results that affect quotes, commitments, access, or other consequential decisions. Define who reviews exceptions and when work should stop.

Read the related section
What could GrowFast find in the first week?

With a tool list and recent records, an initial review might surface a duplicate subscription or an estimating workflow only one person can access. Results depend on scope and evidence.

Read the related section