Skip to main content

Practical AI Education

Learn

Three practical lessons on using AI in everyday business work.

Move from answers to repeatable work.

Start with a useful question, then give the tool the context and boundaries needed to help with real work.

Assistant, automation, or agent?

Assistant
A person asks for help drafting a guest reply, then reviews and sends it.
Automation
A received form creates a task and confirmation through defined rules.
Agent
A system chooses steps toward a goal, such as researching a case, within approved tools and review boundaries.

Lessons by Andrew Heldenmuth, founder and operator of Ascension AI.

Three lessons

  1. Lesson 02

    Getting AI to work from your business, not the internet

    Generic answers come from generic context. AI gets genuinely useful when it works from how your business does things: your policies, your pricing, your customers, your records.

    There are three ways to get context in, in escalating order of usefulness. Paste it by hand each time: works, but nothing is repeatable. Encode it once as maintained instructions and documents a whole team reuses: now the team has a consistent brief and review process for that task. Connect AI to the systems where the business already lives, with real permissions: now nobody feeds it anything, and answers reflect what is actually in your records.

    Each step up trades convenience for governance: the more AI can see, the more deciding what it may see becomes a management job.

    Explore connected business systems →

    Connected context, in practice

    “What changed with this customer since March?”AI reads CRM, email thread, meeting notesOne synthesized answer, with sourcesA person verifies before acting on it

  2. Lesson 03

    Assistants, automation, and agents are three different things

    The words get used interchangeably, and they should not be. An assistant helps a person do a task: the person initiates everything and reviews everything. Automation runs a defined process when a condition occurs: an invoice arrives, data gets extracted, the record updates, no one started it. An agent is delegated a goal and plans its own steps across tools, which is powerful and materially riskier, because oversight becomes the whole job.

    The practical rule: use the lightest one that fits. A weekly report wants automation, not an agent. Drafting a difficult email wants an assistant, not automation. Delegating “investigate this claim and assemble the file” is agent work, and it requires the access, guardrails, and audit trail to match.

    This is the spine of the Ascension framework: the ladder is mostly the climb from assistant to automation to agents, with control growing at each step.

    Explore workflow automation →

    Same business, three modes

    Assistant: “draft the reply to this tenant”Automation: invoice in, extracted, filed, flagged if oddAgent: “assemble the damage claim; flag what’s missing”

Where to take it from here.

  • AI Basics for Business Owners

    Six short lessons. Practical business tasks. One repeatable workflow you can review and trust.

  • Want the full map?

    Ascension, the seven-level maturity framework: where your company sits and which step is worth taking next.

  • Want something to keep?

    Free frameworks and diagrams for running AI work in a business, readable on the page or as a download.

  • Want your team taught in person?

    Lunch and learns and workshops use your team’s real workflows. Lunch and learns: starting at $500; workshops and longer programs are scoped to the team.