Using the PersonaBot Effectively — Trimble Content & Creative Team

psychology

What the PersonaBot actually is (and isn't)

Build the right mental model before you prompt

The most important thing to understand: The PersonaBot is a predictive text system, not a database or search engine. It generates the most contextually appropriate response based on patterns — it can't "look things up," and telling it to "not hallucinate" is like telling a calculator not to give wrong answers. What matters is what you put in.

The antidote to bad or fabricated output is better input — not instructions to behave. Every section of this playbook is about giving the bot the context it needs to generate accurate, on-brand content.

layers Three layers of knowledge

Every PersonaBot draws from three knowledge sources. Knowing which layer answers your question helps you understand how confident the output should be.

Foundation

Charter-derived instructions

Audience, personas, voice, positioning, product knowledge. Stable and authoritative — this is always the base.

Current

Initiative-Strategy & Initiative-Content files

Active campaigns, approved terminology, copy blocks, FAQs, current priorities. Takes precedence over the charter for anything time-sensitive.

Skill

Content skill files

Format-specific instructions for landing pages, email, ads, blog posts, and more. Loaded automatically when the right content type is triggered.

fact_check It's advisory, not authoritative

Every output is a draft or recommendation until a human validates it. The bot won't flag its own gaps unprompted — you need to know when to cross-check. If the bot generates content that doesn't appear in the living files, treat it as a starting point, not finished copy.

block What the bot won't do

It will not speak for other initiatives, make final creative decisions, access real-time market data, guarantee campaign performance, or share confidential product roadmap details. Staying within its scope keeps outputs reliable.

rule

The golden rules of prompting

Six principles to internalize before you type

1

Tell it what you're making, not just what you want Naming the deliverable unlocks the right skill file and format guidance. "Write an email" is vague. "Write a mid-funnel nurture email targeting a CFO at a mid-market general contractor who's evaluating switching from a generic ERP" is actionable — and produces dramatically better output.
2

Assume it already knows the audience — don't re-explain it The charter is loaded. You don't need to explain who a Controller is, what profit fade means, or how a BIM Coordinator thinks. Use the persona names directly (CFO, Operations Director, BIM Coordinator, Estimator) and the bot will draw on everything it already knows about them.
3

Name the funnel stage Top-of-funnel, mid-funnel, and bottom-of-funnel require different tone, urgency, and CTAs. TOFU = awareness and education. MOFU = consideration and differentiation. BOFU = decision and conversion. If you don't specify, the bot will guess — and may miss your intent entirely.
4

Specify the format, channel, and length "Short-form LinkedIn ad, under 150 characters, no questions" is actionable. "Social post" is not. Be precise about the container — a landing page hero, a 3-email nurture series, a Google display ad, a 600-word blog intro — and the output will match the real-world constraint.
5

Use negative constraints — tell it what to avoid "No jargon," "don't open with a question," "avoid bullet points," "no passive voice" steer away from patterns you already know you hate. These work surprisingly well and are one of the most underused prompting techniques. Pair a negative constraint with a positive one: "Don't open with a statistic — lead with the buyer's pain instead."
6

When it drifts, tell it what felt wrong — then ask it to retry Be specific about the problem: "That felt too casual for a CFO audience, try again with more authority" works. "Don't do that" does not. The bot uses your correction as a signal for the rewrite — the more specific your feedback, the more precisely it can adjust.

One more thing: Iterate in the same conversation thread. The bot holds your full session context, so "tighten the hero and make the CTA more urgent" in the same chat consistently outperforms starting a new conversation from scratch.

draw

Getting style-compliant content

Why style guides get missed — and how to make sure they don't

The bot is instructed to pre-load the Trimble brand editorial guidelines and the Viewpoint content style guide before writing any content. But the order in which it does this matters — and users can actively reinforce it.

bolt Trigger content creation explicitly

Use the exact content type name to reliably activate the right skill file. The bot matches trigger phrases — so "write a blog post" is more reliable than "write something for the blog," and "write a landing page" activates the landing page skill while "write a webpage" might not. Available triggers:

Landing page

Whitepaper refresh

Ad copy

Email series

Blog post

check_circle Check for approved language first

Before requesting new copy, ask the bot whether approved language already exists. If it appears in the Initiative-Content file, the bot should use it verbatim. Any content generated beyond the living files is a draft — not approved output.

chat Try asking

"Does the Initiative-Content file have any approved language or copy blocks about [topic]?"

autorenew When style feels wrong, name the guide

If the output doesn't sound right, reference the specific guide by name. This re-activates the style instruction in context and gives the bot a precise target for the rewrite.

chat Try saying

"This doesn't feel like Trimble brand voice — please re-check the Trimble brand editorial guidelines and rewrite."

"This reads too casually for the Viewpoint content style. Please review the Viewpoint style guide and revise the tone."

Precedence rule: When style guides conflict, the bot follows: Trimble brand editorial guidelines → Viewpoint content style guide → skill file instructions. If something still feels off after a retry, check whether the content type skill file is being triggered correctly (see above).

description

Request templates

Fill-in-the-blank prompts you can copy and adapt

person_search Audience question

"From the initiative charter, describe how a [CFO / Controller / Ops Director / Estimator / etc.] at a [company type] typically evaluates [topic or solution type]."

campaign Campaign validation

"We're planning [describe concept]. Rate its alignment with this initiative's strategic goals 1–10, explain your reasoning, and tell me what would make it a 9 or 10."

edit_note Content creation

"Write a [content type] for [funnel stage — TOFU / MOFU / BOFU] targeting [persona] at [company type]. Focus on [value driver or pain point]. [Add any negative constraints here.]"

forum Messaging test

"We have two options: (1) [Option A] and (2) [Option B]. Which aligns better with this initiative's positioning for a [persona] audience? Pick a winner and explain the reasoning."

fact_check Style / approved language check

"Does the Initiative-Content file have any approved language or copy blocks about [topic]?"

"Review this draft against the Trimble brand and Viewpoint style guides. Flag anything that needs to change."

source_notes Source check

"Which source did you draw that from — the initiative charter, Initiative-Strategy, or Initiative-Content? If it's not in one of those, flag it as a draft inference."

Pro tip: Give the bot something to react to. Paste in an existing brief, draft, or asset rather than starting from scratch. "Here's our current landing page — rewrite the hero for a [persona] audience" consistently outperforms a blank-slate prompt.

tune

When output feels wrong

A quick reference for the most common problems

Problem

Output ignored the style guide

Fix

Name the guide explicitly — "Please re-check the Trimble brand editorial guidelines and rewrite." Don't just say "fix the style."

Problem

Output includes details you don't recognize or can't verify

Fix

Ask "Which source did that come from?" If it can't cite the charter or a living file, treat it as a draft inference — not fact.

Problem

Output feels too generic, not initiative-specific

Fix

Add specificity: persona name, vertical (MEP, GC, Heavy Civil), funnel stage, and the specific product or solution you're featuring.

Problem

Tone is off — too informal, too feature-heavy, or doesn't sound like Trimble

Fix

Say exactly what felt wrong: "Too informal for a CFO," "Too feature-focused — lead with the business outcome," or "Doesn't sound like Trimble — re-check the brand guide."

Problem

Output is generating copy for a different initiative

Fix

Make sure you're in the correct GEM. Each PersonaBot is scoped to one initiative and won't speak for others.

Problem

Content isn't using the right format or structure

Fix

Use the exact trigger phrase for the content type (landing page, blog post, ad copy, email series) to activate the correct skill file.

Remember: Gemini can't read linked documents — upload files directly or paste content into the conversation. A Google Drive link won't work; the bot can't open it.