Préciser votre public

L’audience : la variable d’invite la moins exploitée

L’ancrage de l’audience par l’âge

import anthropic

client = anthropic.Anthropic(api_key='sk-ant-your-key-here')

ages = ['a 10-year-old', 'a 16-year-old high school student', 'a college graduate with no CS background']

for age in ages:
    response = client.messages.create(
        model='claude-opus-4-5',
        max_tokens=80,
        messages=[{
            'role': 'user',
            'content': f'Explain what a computer CPU does in 2 sentences for {age}.'
        }]
    )
    print(f'Audience ({age}):')
    print(response.content[0].text.strip())
    print()

Spécifier l’audience selon le rôle

import openai

client = openai.OpenAI(api_key='sk-your-key-here')

audiences = [
    'a product manager who understands user flows but has no coding background',
    'a CFO evaluating whether to approve a $200k infrastructure investment',
    'a first-year developer who knows Python but has never worked with APIs'
]

topic = 'why our system needs a message queue (like RabbitMQ or Kafka)'

for audience in audiences:
    response = client.chat.completions.create(
        model='gpt-4o',
        max_tokens=100,
        messages=[{
            'role': 'user',
            'content': f'Explain {topic} for {audience}. 2 sentences maximum.'
        }]
    )
    print(f'Audience: {audience[:55]}...')
    print(response.choices[0].message.content.strip())
    print()

Pour les dirigeants non techniques

import anthropic

client = anthropic.Anthropic(api_key='sk-ant-your-key-here')

response = client.messages.create(
    model='claude-opus-4-5',
    max_tokens=200,
    messages=[{
        'role': 'user',
        'content': (
            'Explain why we should migrate from our custom authentication system '
            'to Auth0 or Okta.\n\n'
            'Audience: non-technical CEO and CFO who will make the go/no-go decision.\n'
            'Format: 3 bullet points — one per business benefit.\n'
            'Language: no technical jargon. Frame entirely in business value. '
            'Each bullet: max 20 words.'
        )
    }]
)
print(response.content[0].text)

Pour les ingénieurs logiciels expérimentés

import openai

client = openai.OpenAI(api_key='sk-your-key-here')

response = client.chat.completions.create(
    model='gpt-4o',
    messages=[{
        'role': 'user',
        'content': (
            'Explain the trade-offs between using PostgreSQL advisory locks vs Redis distributed locks '
            'for preventing duplicate job processing.\n\n'
            'Audience: senior backend engineer with 8+ years experience, familiar with '
            'CAP theorem, knows Redis and PostgreSQL internals.\n'
            'Format: 4-bullet trade-off analysis — no intro, no conclusion. '
            'Assume expert-level knowledge throughout. No definitions needed.'
        )
    }]
)
print(response.choices[0].message.content)

Pour les professionnels du marketing

import anthropic

client = anthropic.Anthropic(api_key='sk-ant-your-key-here')

response = client.messages.create(
    model='claude-opus-4-5',
    max_tokens=300,
    messages=[{
        'role': 'user',
        'content': (
            'Explain our new AI-powered search feature to our marketing team.\n\n'
            'Audience: marketing professionals who run email campaigns, SEO, and social media. '
            'They understand marketing KPIs but not ML or engineering.\n'
            'What they need to know:\n'
            '1. What customer problem it solves (in customer language, not tech language)\n'
            '2. How to message it in campaigns (key benefit in one memorable sentence)\n'
            '3. Which customer segment benefits most (for targeting)\n'
            'The feature: semantic search that finds relevant results even when keywords do not match.'
        )
    }]
)
print(response.content[0].text)

Ancrer l’audience dans ses connaissances

import openai

client = openai.OpenAI(api_key='sk-your-key-here')

anchored_audiences = [
    'a data analyst who uses Excel and SQL daily but has never used Python',
    'a Python developer who knows pandas and numpy but has never used machine learning',
    'an ML engineer who knows sklearn but has never worked with deep learning frameworks'
]

for audience in anchored_audiences:
    response = client.chat.completions.create(
        model='gpt-4o',
        max_tokens=80,
        messages=[{
            'role': 'user',
            'content': (
                f'Explain what a neural network is in 2 sentences for {audience}. '
                f'Build on what they already know — use it as an analogy anchor.'
            )
        }]
    )
    print(f'Audience: {audience[:60]}...')
    print(response.choices[0].message.content.strip())
    print()

L’audience et le format du contenu

import anthropic

client = anthropic.Anthropic(api_key='sk-ant-your-key-here')

aggregated_prompt = (
    'We launched two-factor authentication (2FA) for our SaaS app. '
    'Generate one short piece of communication for each of the following audiences. '
    'Label each section clearly.\n\n'
    '1. EXECUTIVES (2 bullets, business risk/benefit framing)\n'
    '2. ENGINEERS (numbered steps to enable and test 2FA in dev)\n'
    '3. END USERS (2-sentence friendly explanation of what to do next time they log in)\n'
    'Total: 120 words maximum across all three sections.'
)

response = client.messages.create(
    model='claude-opus-4-5',
    max_tokens=250,
    messages=[{'role': 'user', 'content': aggregated_prompt}]
)
print(response.content[0].text)

Modèles de spécification de l’audience

import openai

client = openai.OpenAI(api_key='sk-your-key-here')

# Role + concern pattern: most useful for persuasive content
skeptic_audience = (
    'For a CISO (Chief Information Security Officer) at a regulated financial institution '
    'who is skeptical about AI tools handling sensitive customer data. '
    'They prioritize: data residency, audit trails, and regulatory compliance.'
)

response = client.chat.completions.create(
    model='gpt-4o',
    messages=[{
        'role': 'user',
        'content': (
            f'Audience: {skeptic_audience}\n\n'
            f'Write a 3-bullet summary of why our AI document processing tool is safe to use. '
            f'Focus entirely on their specific concerns. '
            f'Each bullet: max 20 words. No marketing language.'
        )
    }]
)
print(response.choices[0].message.content)

Documents destinés à plusieurs audiences

import anthropic

client = anthropic.Anthropic(api_key='sk-ant-your-key-here')

response = client.messages.create(
    model='claude-opus-4-5',
    max_tokens=400,
    messages=[{
        'role': 'user',
        'content': (
            'Write a new feature release note for our API v2 launch. '
            'Structure for mixed audiences:\n'
            '## For Leadership (2 bullets: business impact only)\n'
            '## For Developers (2 bullets: breaking changes + migration path)\n'
            '## For End Users (1 sentence: what changes for them)\n'
            'Feature: API v2 processes requests 3x faster, adds webhook support, '
            'but deprecates the /v1/upload endpoint.'
        )
    }]
)
print(response.content[0].text)

La boucle de rétroaction de l’audience

import openai

client = openai.OpenAI(api_key='sk-your-key-here')

# Audience refinement loop
def generate_for_audience(topic, audience_spec):
    response = client.chat.completions.create(
        model='gpt-4o',
        max_tokens=100,
        messages=[{
            'role': 'user',
            'content': f'Audience: {audience_spec}\n\nExplain in 2 sentences: {topic}'
        }]
    )
    return response.choices[0].message.content.strip()

topic = 'why we need database backups'

# Iteration 1: too broad
v1 = generate_for_audience(topic, 'business professional')
print('v1 (too broad):', v1[:120])

# Iteration 2: refined with role + concern
v2 = generate_for_audience(topic, 'non-technical e-commerce founder who lost sales data once before and is skeptical of IT advice')
print('v2 (refined):', v2[:120])

Vérification des connaissances

Préciser votre audience — récapitulatif