Voice

Your agency’s voice. Not Claude’s.

Generic AI sounds like generic AI. Scout learns the specific way your agency writes - sentence patterns, emphasis, what you cut, what you always lead with - and applies it to every draft and every answer. By week six, the edits are mostly typos.

Not a vague brand thing

Four concrete signals Scout tracks.

Sentence cadence - short and punchy vs flowing and exploratory.
Emphasis - what you always lead with (the headline metric, the strategic priority, the recommendation).
Skipped sections - what you cut even when the data’s there. Some agencies never report on impressions; some always do.
Vocabulary - “paid social” vs “Meta ads.” “Customer” vs “buyer.” “ROAS dipped” vs “ROAS softened.” The small word choices that make writing sound like yours.
The training loop

Every edit is signal.

First draft Scout produces is a baseline - Claude’s default with your agency’s prior reports as reference. Every edit you make becomes training data. The phrases you cut, the words you swap, the headings you reorder.

Voice version increments every few weeks. v6 to v7 to v8. By v6, most agencies stop rewriting entirely and just tighten typos.

VOICE v5
VOICE v6
+4 LEARNED
“Performance was healthy this week.”
+“Solid week - ROAS held at 3.4× on $4.2k spend.”
“The Klaviyo flow is performing well.”
+“Welcome flow doing $8.40/recipient - best of the quarter.”
“We recommend optimising the funnel.”
+“Push budget into the top two creatives; pause the bottom one.”
Learned from 6 weeks of edits by Sam P. and Mira K.
Driftline
Punchy, casual
“Killer week - ROAS at 3.4×, cold creative is finally clicking.”
Lead with revenueNever use “leverage”OK to use “killed it”
Halcyon
Measured, formal
“Performance held steady this week, with ROAS at 3.4× on stable cold acquisition.”
Lead with ROASNo first-personAlways quote period over period
When a client needs different

Agency default. Client override.

Most clients get the agency’s default voice. Some don’t - a more formal client, a more casual one, a client that needs a specific phrase included in every report. Voice overrides handle this at the client level without polluting the agency default.

Six weeks of voice training

Same client. Same week’s data. Two drafts.

Week 1 draft: a competent, generic write-up. Week 6 draft: indistinguishable from a draft the account manager wrote. Same numbers, same client - different voice.

WEEK 1
Generic Claude voice

Performance was strong this week with notable improvements in paid social efficiency.

We recommend continuing to optimise creative variants and monitoring the trend over the coming weeks.

WEEK 6
Your agency’s voice

Solid week. ROAS held at 3.4× on $4.2k spend - cold creative finally pulling its weight.

Pushing budget into the top two; pausing the bottom one Monday. Welcome flow is doing the heavy lifting on email.

FAQ

Common questions about voice.

How long until Scout sounds like us?

Most agencies are comfortable sending drafts with minor edits by week two, and stop rewriting entirely by around week six. Every edit you make is training signal.

What does it actually learn from?

Your prior reports as a baseline, then every edit on top - the phrases you cut, words you swap, headings you reorder, and what you consistently lead with or skip.

Can one client have a different tone?

Yes. The agency default applies everywhere; per-client voice overrides handle a more formal client, a more casual one, or a required phrase - without changing your default.

Does the whole team share one voice?

Yes. The voice profile is agency-level, so whichever account manager writes this week’s update, the client hears one consistent voice.

See your voice in your first draft.