The experiment that changed our workflow

The right AI persona
changes everything.

We gave the same marketing brief to four AI models. The prompts were identical. The results were worlds apart — and it came down to one thing: which personas were in the room.

Brief Kids Juice Box Launch
Models tested 4
Prompt Identical across all
SCROLL

The setup

Same prompt. Same brief. Same product. Four completely different levels of strategic thinking.

We asked each model to act as a seasoned brand strategist and develop a full marketing strategy — competitive analysis, buyer personas, and a 4-week omnichannel launch campaign — for a natural, no-added-sugar juice box for kids aged 4–9.

The difference wasn't the question. It was who was answering it. Model 4 ran as a multi-agent system — assembling a team of specialised AI personas (brand strategist, devil's advocate, market analyst, campaign planner) to interrogate the brief from multiple angles before converging on a response. The results speak for themselves.

7.2
ChatGPT · /10
6.8
Gemini · /10
8.1
Multi-Model · /10
9.0
Model 4 (Personas) · /10

Response by response

What each model actually said

These are real extracts from each response — same brief, different depths of thinking.

ChatGPT

Competitive analysis

"Major brands like Capri Sun and Innocent Drinks are already reformulating products around lower sugar and wellness claims. Most brands focus on 'less bad', lower sugar, or organic credentials."

What's missing: A clear view of where to attack. The competitive map lists weaknesses but doesn't identify a time-sensitive or structural opening. It's an accurate summary — not a strategic diagnosis.
Gemini

Greenwashing warning (2026)

"In 2026, parents are weary of greenwashing. Ensure the packaging clearly states 'No Natural Flavours' (which are often lab-made). Transparency is your loudest marketing tool."

What's sharp: This is the most market-aware single observation in any response. But Gemini ends with a question rather than a completed strategy — the brief was never fully delivered.
xanber

The go / no-go gate

"The entire strategy rests on the product passing the kid palate test across diverse taste preferences and geographies. A 70% preference threshold is the go/no-go gate. This assumption must be tested before launch, not assumed."

What's unique: Only response that treats product-market fit as an open question. Introduces confidence calibration and A/B testing of the lead benefit claim — genuine strategic rigour missing elsewhere.
★ WINNER
xanber with Different Persona

The competitive window

"Hiring patterns and product sequencing at major FMCG players suggest R&D investment is flowing toward teen and adult functional beverages, not kids' formats. For an agile challenger brand, this is a structural window — not a permanent gap. You have approximately 18–24 months before a well-resourced player pivots here."

What changes the game: This observation transforms the brief from a launch plan into a time-sensitive strategic bet. No other model identified a competitive window with this specificity. The devil's advocate persona forced the team to pressure-test every assumption.
Key insight

The multi-persona approach doesn't just produce more output — it produces a fundamentally different type of thinking. When a devil's advocate persona interrogates the brand strategist's assumptions, when a market analyst pressure-tests the competitive claims, the result isn't a longer answer. It's a sharper one.

Head to head

The full comparison

Scored across five dimensions a marketing specialist actually cares about.

Dimension ChatGPT Gemini xanber xanber · Personas
Strategic depth Surface-level. Lists gaps without a thesis. Sharper than ChatGPT. Misses structural framing. Strong. Introduces risk-calibration logic. Identifies a time-limited 18–24 month window. Reframes the entire brief.
Persona quality Archetypal. Busy Parent / Health Mom / Value Seeker. Vivid. Allergy-Aware Guardian is distinct and useful. Sharp tensions defined. Skepticism explicitly addressed. Each persona named per tactic throughout the campaign — not just a setup section.
Campaign tactics Complete 4-week plan. Week-by-week clarity. Creative. "Superpower Capes" is memorable. Incomplete overall. Sound mechanics. Pre-launch influencer seeding is smart. Trust → Trial → Advocacy spine. Each week's tactic linked to a persona and conversion goal.
Differentiation thinking "No added sugar" as a differentiator — already table stakes. Greenwashing flag is genuinely sharp and 2026-specific. A/B test of hidden veggies vs. Vitamin C as lead claim — rare clarity. Names child as repeat-purchase gatekeeper. Dual-gatekeeper framing is original.
Risk & caveats raised None flagged. Greenwashing risk only. Taste-test gate · Regional flavour variation · Subscription conversion uncertainty. UGC consent for children · Retail sell-through threshold · Influencer credibility-before-reach principle.
Completeness of brief Fully delivered. Clean and formatted. Ends with a question. Brief not fully completed. Fully delivered. Dense but comprehensive. Fully delivered + Phase 2 setup + retailer debrief framework.
Overall score 7.2 / 10 6.8 / 10 8.1 / 10 9.0 / 10

The takeaway

Selecting the right AI personas isn't a nice-to-have. It's the strategy.

When you brief a single AI persona, you get one perspective — however capable that perspective is. When you assemble a team of specialised thinkers, you get challenge, rigour, and the kind of non-obvious insights that change how a client thinks about their business.

The model mattered less than the method. The prompt that produced a 9.0 wasn't more detailed — it was more deliberate about who was in the room.

Multi-persona prompting Devil's advocate thinking AI team design Strategic AI use FMCG marketing