Setting Realistic Expectations
What AI actually delivers on a deadline, and how to keep your team and clients calibrated
Two ways agencies get this wrong
The first failure mode is overselling: telling clients or your own team that “AI will handle it,” then shipping something unreviewed and wrong. The second, quieter failure is undermining adoption before it starts — one bad first experience and a skeptical account lead writes off the whole effort. Both come from the same root cause: nobody set a concrete definition of what “good” looks like before starting.
Claude drafts. Your team decides.
Every workflow in this course follows the same shape: Claude produces a strong first draft from real inputs, a human with agency and client context reviews and edits it, and the human’s name goes on the final output. That’s not a hedge — it’s the actual production model. Setting this expectation up front, in writing, prevents both failure modes at once.
Key principle: Never present Claude’s output to a client as final without a human review pass. The review pass is not optional overhead — it’s where your agency’s judgment, relationship knowledge, and quality bar actually get applied.
Set the bar before you start, not after you’re disappointed
Before rolling out any workflow from this course to your team, answer three questions together:
| Question | Example answer |
|---|---|
| What does “good enough to send to a human reviewer” look like? | Covers all required sections, matches our tone, cites real numbers from the source data — not final copy |
| Who reviews, and what are they checking for? | Account lead checks for accuracy and client-specific nuance; not a grammar pass, that’s assumed |
| What’s the fallback if the draft misses badly? | Flag it, don’t force-fit it — go back to source notes and try a more specific prompt |
Calibrate your team with a side-by-side, not a lecture
Skeptical team members change their mind faster from one good example than from any amount of persuasion. Run this exercise in a team meeting:
Here are the raw notes from last week's client kickoff call
(pasted below). Draft a status-update email to the client covering:
what we completed this week, what's in progress, any blockers,
and next week's plan. Match the tone of our last three status
emails (also pasted below) - direct, no filler, specific numbers
where we have them.
[paste kickoff notes]
[paste 3 past status emails as tone reference]Have the team compare the draft against what a team member would have written from scratch in the same ten minutes. The honest comparison — not a rigged demo — is what builds real buy-in.
Set client-facing expectations too
You don’t need to announce “we use AI” on every deliverable, but you should have a ready answer if a client asks directly. A simple, honest line works: “We use AI tools to speed up drafting and reporting — every deliverable that reaches you is reviewed and finalized by your account team.” That’s true, it’s reassuring, and it matches the actual workflow you’re building in this course.
Try it: Write your agency’s one-sentence answer to “do you use AI on our account?” before you need it. Share it with your account leads so everyone gives the same answer.