Practical ways to use AI as a thought partner at work.
Session resources with Ben Sondgeroth · Exploring EdTech
Thanks for joining me. Use this page to revisit the presentation, try the customer email exercise, or pick one way to use AI in your work this week.
AI as a thought partner
We started with familiar examples like suggested text and streaming recommendations. Then we looked at generative AI, which creates text, images, and other content based on patterns it has learned and the instructions you give it. You can use it to sketch a first draft, explore another perspective, or spot a question you missed. You still decide what to use and check whether it holds up.
Try the customer email exercise
- Start with “Write an email about our upcoming meeting.” Notice which details the draft has to guess.
- Try the more specific prompt below. Compare the two drafts and see which details make the second one easier to use.
- Follow up with “Read this as the customer. What might still be unclear?” Use the response to decide what to revise.
Draft a friendly email under 120 words to a returning customer. We’re meeting Thursday at 10 a.m. to review a proposal. Ask them to bring their two main questions. Include a subject line and flag any missing details. Use only the facts I gave you.
When you adapt the prompt, give the AI context about the audience and situation, be specific about what it should include, and ask for a structure you can review quickly.
Get more from the second prompt
Once you have a draft, give the AI a specific job for the next turn. If the first email feels stiff, try “Make this warmer and more direct. Keep the meeting time, purpose, and request intact.” If you’re worried the recipient may miss the next step, try “Read this as a busy customer. What would you think I’m asking you to do?”
You can also ask the AI to take on a role. In the coffee shop example, the role of social media manager gives the tool a perspective and a sense of the audience. You still supply the facts and decide whether the result sounds like your business.
Help a locally owned coffee shop draft three friendly Facebook posts for a fictional fall latte launch. Give each post a different angle. Leave placeholders for the launch date and price, and flag any claims that would need checking.
Turn rough ideas into something you can use
- Expand notes after a meeting. Ask for a recap with separate sections for decisions, action items, and open questions. Try “Mark any missing owner or deadline as ‘to confirm.’” Compare the recap with your notes before circulating it.
- Talk through an idea on your phone. Use dictation to explain what you’re trying to do in your own words. Try “Organize what I just said into a short outline. Ask me two questions about anything that’s unclear.” Review the transcription, especially names and numbers.
- Review documents. Use approved or public files. Try “Compare these two proposals in a table. Include price, delivery timing, and exclusions. Point me to the section for each item and mark anything the proposal leaves open.” Open the cited sections and check the figures yourself.
- Explore spreadsheet data. Start with a small, shareable sample. Ask for two possible trends and the figures behind them. Check totals, date ranges, and missing rows in the original file before presenting a finding.
Create visuals for outreach and planning
Give the image tool a purpose, audience, and layout. For example, ask for “A warm, realistic image for a fall coffee shop flyer. Show two lattes on a wooden counter, use soft morning light, and leave clear space at the top for the event title.” Add the final wording yourself and check the image for distracting details. The same approach works for a packaging mockup or an early idea for a website banner.
Choose a setup for work you’ll revisit
- ChatGPT Projects. Use a project to keep chats about a continuing task together, such as planning a workshop. Give it a short brief with your audience, goals, and preferred formats. Keep unrelated work in separate projects so the context stays useful.
- Custom GPTs. Consider one when you want to repeat a defined workflow, such as drafting event descriptions in a consistent format. Write down the inputs it needs and the rules it should follow. Test it with a complete example and one that’s missing details to see whether it asks before guessing.
- NotebookLM. Create a notebook with a small set of approved sources, such as an onboarding guide and a public FAQ. Ask “What should a new employee do in their first week? Show me where each step comes from.” Open the cited passages. Use a generated FAQ or audio overview as a starting point for review.
- AI agents. An agent can work through several steps and may use connected tools. Start with a bounded research or planning task. Specify which sources it can use and where it should stop for review, especially before it sends anything or makes a change.
Before you share the result
Use only information you have permission to share with the tool. In our supplier example, a confident paragraph made claims about delivery times, savings, and certifications. Check claims like those against the original documents, and open any citations before relying on them. Read the finished draft in your own voice before sending it.
For practice, take the fictional supplier claim that packaging cuts shipping costs by 35%. Ask AI to list what you would need to verify that number. You might need the original shipping bills, the proposed packaging costs, and the volume and time period used in the comparison. Check whether the answer assumes all customers get the same delivery terms or leaves out an important perspective. If a detail remains unconfirmed, flag it before sharing.
Keep the conversation going
Have a question from the session? I’d be glad to hear from you.
