BoulevardLabs
THE METHOD

Reconcile the record · instrument the definitions · ship the smallest thing that runs · close the loop.

HOW WE DELIVER Talent gives you direct access to our specialists.
ALL CAPABILITIES
NAVAL AVIATION From an Access database to a Certificate-to-Field in under 90 days $1.85M → $4.5M INSTALLATION READINESS 200 installations, one screen 200+ SITES · 24/7/365
GHOST WORKS No Soup Sandwich · ODCS Internship 90-DAY CLOCK · INTERNALLY FUNDED EXPLORE GHOST WORKS →
PAST PERFORMANCE

Where a customer can't be named, the component and scale are characterized instead.

ALL WORK
CAREERS

Engineering, data science, and Ghost Works.

OPEN ROLES INTERNSHIPS

Hands-on projects and practical challenges, including our Drone Development internship.

EXPLORE INTERNSHIPS →
AI Basics · Lesson 2 of 6

Asking well

10 MINUTES · AUDIO, READING, EXERCISE AND QUIZ

How to write requests that get useful results the first time, and how to improve them when they don't.

Watch the overview · about 1.5 minutes · captions on screen and in the player · AI-generated narration
Listen to this lessonAbout 5 minutes · narration is AI-generated
Read the narration transcript

Welcome to lesson two, Asking well. In lesson one, you saw that an AI assistant produces whatever best fits the patterns in front of it. That means the request you give it shapes everything that comes back. This lesson is about writing requests that get useful results the first time, and improving them when they don't.

Picture handing a task to a smart new hire who started this morning. They're quick and willing, but they know nothing about your team, your audience, or what good looks like here. If you say, write something about the new expense policy, you'll get something generic. AI assistants are the same. Most disappointing results come from requests that are too short. The fix isn't clever wording. It's giving the same context you'd give a person.

There are five things worth including. You don't need all five every time, but the more you include, the better the first draft tends to be. First, context. What's going on, and who it's for. For example, our team is moving to a new timesheet system. Training is March third, the old system closes March fourteenth, and the new one opens March seventeenth. Most staff are not technical. Second, role. The perspective you want, like, write as an HR generalist explaining this to employees. Third, task. The specific thing to produce, such as a short announcement email. Fourth, format. The shape of the output. Under one hundred fifty words, a subject line, and three bullets for the key dates. And fifth, constraints. What to avoid or include. No jargon, and don't promise anything about pay timing.

Below this section is an interactive prompt builder that uses that same timesheet example. Each of the five parts has its own toggle. With everything off, the request is just, write something about the new timesheet system, and the builder warns you to expect a generic message built on guesses. Turn the parts on one at a time, and watch both the request and the likely result change. With all five on, you get a strong first draft. You'd still check the dates and the tone against your notes before sending. Pause the audio here and try a few combinations. The same five parts work for almost any task.

The assistant doesn't know your facts, but it's very good at working with facts you supply. Paste in the meeting notes, the draft you already wrote, or the bullet points from your manager, as long as they contain nothing sensitive and your organization allows it. Then ask it to work from that material only. Saying, use only the information below, helps keep it from filling gaps with invented details. And if something is missing, ask it to say so rather than guess.

Your first request rarely needs to be perfect. A few habits make the back-and-forth much more productive. Try asking for questions first. Before you start, ask me up to three questions that would help you do this well. Iterate instead of starting over. Something like, good, now make the second paragraph shorter and the tone a little warmer. The assistant keeps the earlier context within the same conversation. Ask it to list any assumptions it made, which often reveals where it guessed. And ask for options. Asking for three versions of a subject line is faster than rewriting one three times.

When a request produces a result you like, save it. Keep a simple document of reusable prompts with blanks to fill in, like the audience, the deadline, or a spot that says paste notes here. Over time, this becomes your personal toolkit, and you can share the good ones with your team. Lesson six builds on this idea.

The worked example on the page is a complete request. It helps an operations manager tell department heads about a new supply ordering process. It covers context, role, task, format and constraints, and it supplies the facts. Pay attention to the last two sentences. They tell the assistant to use only the details given, and to list anything important that seems missing instead of inventing it. That gives you a short list of things to confirm. The Try it exercise is called Before and after. Pick a routine, non-sensitive writing task, like a note announcing an office closure. In one new conversation, type a one-line request and save the result. In a second new conversation, write a fuller request using all five parts, and end with, ask me any questions you need first. Answer its questions, read the draft, and ask for one specific revision instead of starting over. Then compare the first and final results side by side. The final version should need noticeably less editing. If it does, save that fuller request as a reusable prompt, with blanks for the details that change.

To sum up. Give it what you'd give a new colleague: context, role, task, format and constraints. Supply your own material, and ask it to flag what's missing. Refine in the same conversation, and save the requests that work. Take the short quiz below, then move on to lesson three, Checking its work.

Treat it like a capable new colleague

Imagine handing a task to a smart new hire who started this morning. They are quick and willing, but they know nothing about your team, your audience or what "good" looks like here. If you say "write something about the new expense policy," you will get something generic.

AI assistants are the same. Most disappointing results come from requests that are too short. The fix is not clever wording. It is giving the same context you would give a person.

Five things to include

You do not need all five every time, but the more of them you include, the better the first draft tends to be.

  • Context: what is going on and who it is for. "Our team is moving to a new timesheet system. Training is March 3, the old system closes March 14 and the new one opens March 17. Most staff are not technical."
  • Role: the perspective you want. "Write as an HR generalist explaining this to employees."
  • Task: the specific thing to produce. "Draft a short announcement email."
  • Format: the shape of the output. "Under 150 words, a subject line, three bullets for the key dates."
  • Constraints: what to avoid or include. "No jargon. Don't promise anything about pay timing."

Loading the interactive example…

Give it your material

The assistant does not know your facts, but it is very good at working with facts you supply. Paste in the meeting notes, the draft you already wrote or the bullet points from your manager, as long as they contain nothing sensitive and your organization allows it. Then ask it to work from that material only.

Saying "use only the information below" helps keep it from filling gaps with invented details. If something is missing, ask it to say so rather than guess.

Make it a conversation

Your first request rarely needs to be perfect. A few habits make the back-and-forth much more productive.

  • Ask for questions first: "Before you start, ask me up to three questions that would help you do this well."
  • Iterate instead of starting over: "Good. Make the second paragraph shorter and the tone a little warmer." The assistant keeps the earlier context within the same conversation.
  • Ask it to show assumptions: "List any assumptions you made." This often reveals where it guessed.
  • Ask for options: "Give me three versions of the subject line" is faster than rewriting one three times.

Save what works

When a request produces a result you like, save it. Keep a simple document of reusable prompts with blanks to fill in, such as [audience], [deadline] or [paste notes here]. Over time this becomes your personal toolkit, and you can share the good ones with your team. Lesson 6 builds on this idea.

Worked example

A complete request

You are helping an operations manager write to department heads. Context: our office is changing its supply ordering process. Starting the first of next month, all orders go through a shared online form instead of email, and orders placed by Wednesday arrive the following Monday. Task: draft a short email announcing the change. Format: a subject line, then no more than 120 words, with the two key rules as bullets. Constraints: friendly and direct, no jargon, and don't mention any vendor names. Use only the details I've given. If anything important seems missing, list it after the draft instead of inventing it.

This request covers context, role, task, format and constraints, and it supplies the facts. The last two sentences matter most: they tell the assistant not to fill gaps with made-up details and to flag what is missing, which gives you a short list of things to confirm.

Try it · 5 to 10 minutes

Before and after

  1. Pick a routine, non-sensitive writing task, such as a reminder about a team meeting or a note announcing an office closure.
  2. In a new conversation, type a one-line request, such as "Write an email about the office closure." Save the result.
  3. Start a second new conversation. Write a fuller request using context, role, task, format and constraints. End with "Ask me any questions you need first."
  4. Answer its questions, then read the draft.
  5. Ask for one specific revision, such as a shorter version or a different tone, instead of starting over.
  6. Compare the first and final results side by side.

How you’ll know it worked: The final version should need noticeably less editing than the first one. If it does, save the fuller request as a reusable prompt with blanks for the details that change.

Use only non-sensitive information, in a tool your organization allows.

Key takeaways

  • Give it what you would give a new colleague: context, role, task, format and constraints.
  • Supply your own material and ask it to flag what is missing.
  • Refine in the same conversation, and save the requests that work.

Quick check

1. You ask an assistant to "summarize the project status" and get a vague, generic paragraph. What is the most likely cause?
The assistant does not know your project. Without your notes, the audience and the format you want, a generic answer is the best it can do.
2. The draft is close but the tone is too formal. What is usually the most efficient next step?
Within the same conversation the assistant keeps the earlier context, so a targeted revision request builds on what already works.
3. Why add "use only the information below, and tell me if anything is missing" to a request?
This instruction steers it to work from your material and flag gaps rather than guess. It helps, but it does not guarantee accuracy, so you still review the result.
Share this lesson

The Boulevard Consulting Group, LLC, DBA BoulevardLabs. SBA 8(a) certified, Native American-owned small business.

FOLLOW
2611 S. Clark St., Suite 430, Arlington, VA 22202 · UEI DYV5P3KPC474 · CAGE 70AG9