When I was a kid, I loved the Amelia Bedelia books. She's
the cheerful housekeeper who follows instructions to the letter and misses the
meaning entirely. Tell her to "dust the furniture," and she sprinkles it with
dust. Ask her to "draw the curtains," and she sketches them on paper. The
stories are funny when you're seven. They're less funny when you realize modern
AI behaves the same way.
AI isn't dumb — far from it. It can draft code, summarize
meetings, write marketing copy, even create prototypes for tools your business
needs. But it's also literal. It doesn't know what you meant; it only knows
what you said. If you're vague, it guesses. If you're contradictory, it
improvises. And like Amelia Bedelia, it will do all of this with total
confidence, smiling the whole way.
This is why so many business owners in Greater Boston try AI
once and walk away disappointed. They paste a vague request like "make me a
report" and get something generic back. "AI doesn't work for us," they say,
when really the problem is the input. In computer science there's an old phrase
for this: garbage in, garbage out. Even the best AI in the world can't give you
what you need if you don't tell it clearly what you want.
There's a famous example of this in the real world, and it
didn't involve AI at all. In 1999, NASA lost the $125 million Mars Climate
Orbiter because one team used metric measurements and another used imperial. A
spacecraft that was supposed to study the Martian atmosphere burned up instead.
A tiny miscommunication, a catastrophic result. Most small business prompts
aren't as high-stakes as interplanetary missions, but the principle is the
same: unclear instructions cost time and create avoidable headaches.
The fix isn't complicated. It's learning to write better
prompts — not more complicated ones, just clearer ones. A framework we use with
clients is called GCES: goal, context, expectations, source. Think of it as
translating what's in your head into a language AI understands. First, define
the goal — exactly what you want. Instead of "write me an email," try "write a
friendly follow-up email to a prospective client after our first meeting,
inviting them to schedule a call." Add context — who the recipient is, why it
matters. Set expectations — tone, length, format, even color scheme if
relevant. Finally, provide a source — a previous email, brand guidelines, or an
example of the style you want.
When you combine these pieces, something changes. The AI
stops guessing. The first draft comes back closer to what you envisioned. A
downtown Boston law firm used this approach to draft client reminders; what
used to sound stiff and impersonal now reads like it came from their actual
attorneys. A wealth management firm in Newton applied GCES to quarterly
summaries, instructing AI to highlight tax implications in plain English. A
dental office in Quincy used it to generate staff onboarding checklists, complete
with tone adjustments to keep it friendly but professional. Same AI, same
capabilities, radically different results — all because of better instructions.
It's tempting to think prompting is just common sense, but
there's more to it. Many new users make the same mistakes. They overload their
prompts, cramming every possible detail into one giant request. The AI responds
with something clunky and hard to edit. Or they expect perfection on the first
try, not realizing prompting works best as a back-and-forth. They type, they
review, they refine. The conversation gets better with each round. And almost
everyone forgets the golden rule: never paste sensitive data. No client names,
no financial info, no personal health details. Use placeholders like "Client
Name" or "Project X." AI doesn't need the real thing to draft something useful.
Getting this right has ripple effects beyond the immediate
task. Clear prompts create consistency. If everyone in your office — partners,
associates, assistants — uses the same framework, the output looks and feels
like it comes from the same voice. That's critical for professional services
firms, especially in law, finance, or healthcare, where tone and accuracy
matter as much as speed. Over time, those prompts can be saved and shared,
forming a kind of internal playbook. Suddenly, your team has its own library of
instructions that anyone can reuse — a quiet form of managed service you built
yourself.
There's another unexpected benefit: culture. When employees
see their prompts turning into useful tools, they start contributing ideas.
"Could we automate this report?" "Could we make a checklist for that process?"
The technology becomes less intimidating and more collaborative. We've seen
this firsthand in small agencies around Boston. A marketing team in Cambridge
gave its account managers the freedom to prototype with AI. Within weeks,
someone built a content scheduler that fit their exact workflow; another
automated client follow-ups. Not every experiment worked, but the ones that did
saved hours and sparked more creativity.
Prompting may sound tactical, but it's strategic too. The
businesses that master it are better positioned to take advantage of vibe
coding — describing entire tools, not just individual tasks. And those who
invest in learning it now will have a head start as AI becomes more embedded in
daily work.
Try this: take a prompt you've used before — maybe something
like "summarize this meeting" — and rewrite it using GCES. Make it specific:
"Summarize this meeting for the partners who weren't there, focusing on client
concerns and next steps, in three bullet points using plain language." Compare
the results side by side. Chances are, the clearer prompt saves you time and
produces something you can actually send.
Good prompting is like packing for a hike. If you take the
time to plan, to bring what you need and leave what you don't, the trip ahead
is smoother. In Camp Vibes, we call this "packing smart." The next leg of the
journey is about safety — how to keep your AI experiments secure and protect
your clients' data — but for now, start here. Pack well. It'll make the whole
hike easier.
