Speak to Everyone Without Losing Yourself

Personalization and Audience Segmentation

The Email That Went to Everyone (And Felt Like It)

Their marketing manager, frustrated with stagnant numbers, started using ChatGPT Plus to rewrite the same core newsletter content in three distinct versions: one for first-time buyers who seemed gift-driven, one for repeat buyers focused on outdoor performance gear, and one for lapsed customers who hadn't purchased in over a year. She didn't change the products being featured. She changed the framing, the opening line, the call-to-action, and the tone. The gift-driven version led with warmth and occasion. The performance version led with technical specs and trail conditions. The lapsed-customer version opened with a direct acknowledgment: "We noticed you haven't been out with us in a while." Same products. Three completely different conversations.

The results after 90 days were hard to ignore. Open rates across the segmented sends climbed to between 24% and 31% depending on the segment. Click-throughs on the performance segment hit 6.8%. The lapsed-customer version recovered 4% of dormant subscribers, not a massive number in isolation, but worth tens of thousands of dollars in recovered revenue when multiplied across a list that size. The team hadn't built a new AI system. They hadn't hired a data scientist. One marketing manager, using a $20/month tool, rewrote the same email three times with AI assistance and produced measurably better outcomes. That's the core tension this lesson explores: personalization used to require enormous resources. Now it requires clear thinking.

Click to enlarge

The Principle Behind the Result

What Cotopaxi's marketing manager figured out, almost by accident, is something direct marketers have known for decades but rarely had the tools to act on at scale. People don't respond to messages aimed at "everyone." They respond to messages that feel like they were written for them specifically. The problem has always been capacity. Writing three versions of everything, for every campaign, for every channel, was simply not feasible for a team of two or three people managing dozens of campaigns per quarter. AI doesn't change the principle. It removes the capacity constraint. Suddenly, writing five versions of a subject line, or three versions of a landing page intro, or four variations of a sales follow-up email, takes twenty minutes instead of two days.

Click to enlarge

How a Hospital System Learned to Stop Talking Like a Hospital

Intermountain Health, a large nonprofit hospital system operating across seven western US states, faced a communication problem that will sound familiar to anyone in healthcare, insurance, or any heavily regulated industry. Their patient outreach content was accurate, compliant, and almost completely unread. Appointment reminder emails had open rates below 15%. Preventive care campaign materials, the kind designed to get patients to schedule annual checkups or flu shots, were generating almost no response. The content team knew the information was important. Patients clearly didn't feel that way. The language was clinical, the tone was institutional, and every message sounded like it had been written by a committee, because it had been.

A communications manager on their team began experimenting with Claude Pro to reframe the same health information for different patient segments. Younger patients (18-35) received messages written in a direct, time-conscious tone that emphasized convenience: "Your annual checkup takes 45 minutes. Here's how to book it without calling." Older patients received warmer, more detailed messaging that acknowledged concerns and offered phone support prominently. Parents of young children got content that connected preventive care to their kids' wellbeing. None of this required new medical information. It required someone to think clearly about who was reading the message and then use AI to write it appropriately. The communications manager described it as "finally having a writing partner who doesn't get tired of revisions."

This example matters because it shows segmentation isn't just a marketing tactic for selling products. It applies anywhere you're trying to move people toward an action, scheduling an appointment, completing a training module, responding to a survey, attending a meeting, or reading a report. The underlying skill is the same: identify who you're talking to, understand what they care about, and frame your message in terms of their world. AI makes executing that skill faster. But the thinking has to come from you. No AI tool knows that your 55-plus patient segment prefers phone callbacks unless you tell it.

A Real Estate Team That Stopped Sending the Same Listing Email to Everyone

A boutique real estate agency in Austin, Texas with a four-person team was managing a buyer contact list of around 2,400 people. Every time a notable listing came in, they'd send the same property announcement to everyone. It was efficient. It was also producing unsubscribes, because a first-time buyer looking at $350,000 condos doesn't want to hear about a $1.2 million lakefront property, and vice versa. The lead agent started using Gemini (Google's AI, built into Google Workspace) to write segmented versions of property announcements directly inside Gmail and Google Docs. She kept a simple reference document with five buyer profile descriptions, first-timers, move-up buyers, investors, downsizers, and relocation clients, and pasted the relevant profile into her prompt alongside the property details.

The result was five emails written in about 25 minutes, each emphasizing different aspects of the same property. The investor version led with rental yield estimates and proximity to employment centers. The downsizer version emphasized single-floor living, low maintenance, and neighborhood walkability scores. The first-timer version opened with financing context and what the monthly payment would look like at current rates. The property didn't change. The audience's reasons for caring about it were completely different. Her unsubscribe rate dropped by more than half in the first quarter. More importantly, response rates from serious buyers, the ones actually ready to book showings, increased significantly because the emails felt relevant rather than generic.

What an HR Manager at a Manufacturing Company Figured Out

Personalization isn't only a marketing problem. Diane, an HR manager at a 600-person manufacturing company in the Midwest, was responsible for rolling out a new employee benefits enrollment period. Every year, the same problem: low enrollment in the supplemental benefits options, confusion about the health savings account, and a flood of calls to her desk in the final week before the deadline. She'd always sent one all-hands email and posted one flyer in the break room. This year, she used Microsoft Copilot (built into her company's Microsoft 365 subscription) to write four different versions of the enrollment announcement: one for hourly workers focused on cost and simplicity, one for salaried employees emphasizing the tax advantages of the HSA, one for employees nearing retirement age highlighting the supplemental life insurance options, and one for new hires who'd never gone through enrollment before.

Diane didn't have a CRM or a marketing platform. She had Outlook, a spreadsheet with employee data sorted by category, and Copilot. She wrote each version in a Word document using Copilot to adjust the language, tone, and emphasis for each group, then sent them through Outlook to the relevant distribution lists her IT team had already set up. Calls to her desk in the final enrollment week dropped by 40% compared to the previous year. Supplemental benefit enrollment went up by 12 percentage points. She hadn't done anything technically complex. She had thought clearly about the four different groups receiving her message, and used AI to execute four different conversations rather than one.

Putting It Into Practice: Your Segmentation Starting Point

The most common mistake professionals make when trying AI-assisted personalization for the first time is starting with the tool instead of starting with the audience. They open ChatGPT and type "write me a personalized email", and then wonder why the output feels generic. The tool produced generic content because it received generic instructions. Personalization starts before you touch the keyboard. It starts with a clear answer to three questions: Who specifically is receiving this? What do they already know or believe? What do they need to feel or understand in order to take the action I want? When you can answer those three questions for a specific group, you have everything the AI needs to write something that actually works.

A practical starting point for most professionals is to identify between two and four distinct groups within your current audience, customers, employees, prospects, patients, students, or whatever population you communicate with regularly. You don't need sophisticated data analyzis to do this. You need honest observation. Think about the last ten conversations you had with people from your audience. Were there patterns in what they asked about? In what they seemed to misunderstand? In what motivated them versus what left them cold? Those patterns are your segments. Write a two or three sentence description of each one, their situation, their primary concern, and what success looks like for them. That description becomes the context you feed into every AI prompt for that segment.

Once you have those segment descriptions written down, the workflow becomes repeatable and fast. You write your core message once, the facts, the offer, the key information. Then you run it through AI once per segment, each time pasting in the relevant segment description alongside the core content and asking the tool to rewrite it for that specific audience. The AI handles the adaptation. You review, adjust where needed, and send. What used to take a full day of writing and editing now takes an hour. More importantly, the output is genuinely different for each group, not just cosmetically tweaked, but substantively reframed around what each segment actually cares about. That's the shift from broadcast to conversation.

Click to enlarge

Key Lessons So Far

1. Personalization isn't about using someone's first name in a subject line. It's about framing the same information through the lens of what a specific group of people actually cares about.
2. AI removes the capacity constraint that made personalization impractical for small teams. Writing three or four versions of a message now takes roughly the same time as writing one used to.
3. The quality of AI-assisted personalization depends entirely on the quality of your audience description. Vague input produces vague output. Specific, human insight about your audience is your competitive advantage.
4. Segmentation applies far beyond marketing. HR communications, internal updates, patient outreach, client proposals, and sales follow-ups all benefit from the same approach.
5. You don't need a large budget, a CRM, or technical skills to start. ChatGPT Plus ($20/month), Claude Pro ($20/month), Microsoft Copilot (included in many Microsoft 365 plans), and Google Gemini (included in Google Workspace) are all capable of producing genuinely segmented content.
6. The most effective segmentation starts with two to four clearly defined groups, not dozens. Start simple, get results, then add complexity if the situation warrants it.
7. Your core message, the facts, the offer, the information, stays the same across segments. What changes is the framing, the emphasis, the tone, and the call to action.

From Segments to Stories: Making Personalization Feel Human

In 2022, Spotify's marketing team faced a problem that sounds almost too good to complain about: they had too much data. Hundreds of millions of listeners, thousands of genres, and listening habits that shifted by time of day, mood, and season. Their email campaigns were performing adequately, open rates around industry average, decent click-throughs. But 'adequate' wasn't the goal. The team started using AI to cluster listeners not just by genre preference, but by behavioral patterns: when people listened, how long, what they skipped, what they replayed. The result wasn't just better targeting. It was a fundamental shift in how they wrote copy. Each segment got messaging that felt like it was written by someone who actually knew them.

The principle Spotify extracted from this work is one that applies far beyond streaming. Segmentation is only half the job. The other half is translating what you know about a segment into language, tone, and framing that resonates specifically with them. A 45-year-old podcast listener who commutes 90 minutes daily has completely different emotional triggers than a 22-year-old who listens to lo-fi beats while studying. Same product. Radically different message. Spotify's team used AI not just to sort people into buckets, but to generate the actual words that would speak to each bucket. That's the leap most marketing teams haven't made yet.

This is where AI tools like ChatGPT Plus and Claude Pro earn their subscription cost. They don't just help you identify who your segments are, they help you write for them, at scale, without losing the human texture that makes content feel genuine rather than algorithmic. The craft is in how you describe your audience to the AI. The more specific and behavioral your description, the more targeted and useful the output. Vague inputs produce generic content. Specific inputs produce content that feels like it was written by someone who has been listening to your customers for years.

The HR Director Who Stopped Writing One Job Post

Rachel Nguyen runs talent acquisition for a mid-sized logistics company in Ohio. For years, she wrote one job description per role and posted it everywhere. LinkedIn, Indeed, the company website. The applications she got were hit or miss. Then she started thinking about her candidates the way a marketer thinks about customers. A warehouse operations manager role attracts at least three distinct candidate profiles: recent graduates looking for their first leadership role, experienced shift supervisors wanting a step up, and career changers from adjacent industries like retail management. Each group reads job postings differently. Each has different fears about applying. Each responds to different proof points.

Rachel started using Claude Pro to write three versions of each job description, one for each candidate profile. She'd describe the segment in detail in her prompt, including what they likely worried about (am I qualified enough? will this company value my non-traditional background?) and what they were hoping to find (growth path? team culture? stability?). The AI generated distinct versions that shared the same factual requirements but framed them entirely differently. Applications increased 34% in the first quarter. More importantly, the quality of fit improved, candidates who applied felt the role had been written for someone like them, because it had been.

The Consultancy That Stopped Sending the Same Proposal to Everyone

Meridian Advisory, a boutique management consultancy based in Chicago, had a proposal acceptance rate of about 22%. Their proposals were thorough, well-designed, and completely interchangeable. A proposal sent to a CFO looked almost identical to one sent to a Chief People Officer, even when the project scope was the same. The partners knew this was a problem but assumed fixing it would require hours of custom writing per proposal. When they started using Microsoft Copilot integrated into their Word and PowerPoint workflow, they found a different answer: they could generate segment-specific proposal language in minutes, as long as they were precise about who they were writing for.

Their approach was systematic. Before drafting any proposal, a partner would spend ten minutes writing what they called a 'client lens', a paragraph describing the decision-maker's role, priorities, likely objections, and how they would measure success. That paragraph became the first thing pasted into Copilot before any proposal content was generated. The AI then framed the entire proposal around that lens. A CFO-targeted proposal led with ROI, risk mitigation, and cost timelines. A CPO-targeted proposal for the same project led with employee experience, retention impact, and change management. Same consulting work. Completely different story. Their proposal acceptance rate climbed to 31% within two quarters.

Click to enlarge

A High School Principal Who Learned to Write for Three Different Parent Audiences

David Okonkwo is the principal of a public high school in Atlanta with a student body that spans genuinely different socioeconomic and cultural backgrounds. For years, school communications went out as one-size-fits-all newsletters. Then David started noticing something: parent engagement varied dramatically by community group, even when the information being shared was identical. He began mapping his parent audience into rough segments, parents of college-bound seniors, parents of students in vocational tracks, and parents of recent immigrant families navigating the American school system for the first time. Each group had completely different information needs, different levels of prior knowledge, and different communication preferences.

Using Google Gemini, David started drafting three versions of key communications, enrollment deadlines, test prep resources, and graduation requirements. He'd write one master version of the message, then prompt Gemini to rewrite it for each parent segment with specific instructions: plain language and translated context for immigrant families, detailed college application timelines for college-track parents, and clear career pathway information for vocational-track families. The effort per communication went from two hours to forty minutes. Parent attendance at information nights increased across all three groups. The content hadn't changed, the framing had. That's the whole game.

Putting Segmented AI Content Into Your Actual Workflow

The practical challenge with personalization isn't understanding why it matters, it's building a workflow where it actually happens consistently rather than occasionally. Most marketing and communications teams have good intentions about segmentation that collapse under deadline pressure. When you have 45 minutes to get a campaign email out the door, writing three versions sounds like a luxury. This is where AI shifts the equation. Generating a second or third version of a piece of content takes three minutes once your segment descriptions are ready. The bottleneck moves from writing to thinking, specifically, thinking clearly about who you're talking to before you start.

A practical starting point is to pick one content type you produce regularly, weekly emails, social posts, proposals, or internal announcements, and commit to running it through a two-segment test for the next four weeks. Write your standard version first, then prompt your AI tool to rewrite it for a second, distinctly different audience. Pay attention to what changes: the examples used, the tone, the level of detail, which benefits get emphasized. You'll quickly develop an intuition for which elements are truly universal and which are segment-specific. That intuition is what eventually makes segmented content creation fast rather than effortful.

The teams that do this best treat their AI tool like a skilled writer who needs good direction, not a vending machine that produces content on demand. The direction is the segment description. The better your description, the more specific, behavioral, and emotionally grounded, the better the output. Over time, your segment library becomes a genuine asset: a set of audience portraits that inform not just AI prompts but briefs, campaign strategies, and hiring criteria for content roles. What starts as a prompt-writing habit becomes a sharper organizational understanding of who you're actually serving.

Click to enlarge

Key Principles From This Section

1. Segmentation without tailored copy is just sorting, the value comes when your messaging actually changes based on who's reading it.
2. AI tools generate segment-specific content most effectively when given behavioral audience descriptions, not just demographic labels.
3. The same offer, product, or message needs different framing for different decision-makers, same facts, different story.
4. A 'segment description' pasted into your prompt before any content request is the single highest-impact habit you can build for AI-assisted personalization.
5. Building a reusable Audience Segment Library turns a one-time effort into a permanent workflow asset that compounds over time.
6. Segmented content creation becomes fast, not slow, once your segment descriptions exist and your prompting instinct develops.
7. The comparison between segments teaches you which parts of your message are truly universal and which are audience-specific, that's strategically valuable knowledge beyond just the content itself.

Section Takeaways

• Spotify, Meridian Advisory, and David Okonkwo's school all achieved measurably better results by combining AI tools with precise audience descriptions, not by using AI harder, but by briefing it better.
• Different professional roles. HR, consulting, education, marketing, all face the same underlying personalization challenge: one message trying to do the work of many.
• AI tools like ChatGPT Plus, Claude Pro, Microsoft Copilot, and Google Gemini can generate segment-specific versions of any content type in minutes, given the right prompt structure.
• The comparison table in this section gives you a ready-to-use reference for how to adjust tone, proof points, and framing based on your decision-maker's role.
• Your next step is practical: pick one content type, build two segment descriptions, and run the two-version test this week.

From Spray-and-Pray to Surgical: Making Personalization Work at Scale

In 2019, Spotify's marketing team faced a familiar problem at an unfamiliar scale. They had 232 million users across 79 countries, and their promotional emails were getting opened at roughly the same rate as a generic supermarket flyer. The content was fine. The offers were real. But everything felt addressed to 'Dear Customer.' Their solution, now famous in marketing circles, was Spotify Wrapped. Instead of sending everyone the same message, they used each listener's own data to tell them a story about themselves. Open rates didn't just improve. The campaign became something people screenshot and share. The lesson wasn't about data volume. It was about making the individual feel seen.

Spotify had engineers and data scientists to build that system. You don't need any of that. What you need is the principle underneath it: personalization works when it reflects something true about the specific person receiving it. AI tools like ChatGPT and Claude can help you apply that principle to your emails, proposals, and campaigns, even if you're a team of one. The gap between Spotify's operation and yours is smaller than it looks, because the hard part was never the technology. It was knowing what to say to whom, and why it would matter to them.

This is where most marketing personalization breaks down in practice. Teams spend weeks building audience segments, then write one piece of content and paste the customer's first name at the top. That's not personalization, it's mail merge with extra steps. Real personalization means the message itself changes based on what that segment actually cares about. Their fears, their language, their specific situation. AI makes producing those variations fast enough that it's no longer a resource problem. It's a thinking problem. And thinking is exactly what you need to do first.

Click to enlarge

What a Regional Bank Got Right

A regional bank in the American Midwest, not a fintech startup, not a Silicon Valley darling, wanted to promote a new small business checking account. Their marketing manager had three segments: freelancers, restaurant owners, and retail shop owners. Same product. Completely different concerns. Freelancers worried about separating personal and business finances. Restaurant owners cared about cash flow during slow seasons. Retailers wanted to understand how the account worked with their point-of-sale systems. One email could not honestly speak to all three.

The marketing manager wrote a single detailed briefing for each segment, two paragraphs describing the customer's world, their specific anxiety, and what outcome they actually wanted. She gave each briefing to Claude and asked for a 200-word email that spoke directly to that customer's situation. Forty minutes later, she had three genuinely different emails that shared a product but told three different stories. Click-through rates on the segmented campaign ran 34% higher than their previous single-version send. The AI didn't do the thinking. It executed the thinking she'd already done.

The HR Director Who Stopped Sending the Same Job Post Everywhere

Personalization isn't only a marketing problem. A talent acquisition director at a mid-sized logistics company noticed that the same job posting on LinkedIn attracted very different candidates than the same posting on a trade association job board. LinkedIn candidates responded to career growth language. Trade board candidates responded to stability, benefits, and specific equipment they'd be working with. She'd been posting identical copy everywhere and wondering why quality varied so much by source.

She used ChatGPT to rewrite each job posting for its specific platform audience, keeping the role requirements identical but changing the framing, the opening hook, and the benefits emphasis. For LinkedIn: 'Grow into operations leadership within 18 months.' For the trade board: 'Operate a fleet of late-model Kenworth T680s with full benefits from day one.' Applications from the trade board increased by 41% in the following quarter. Personalization, she said afterward, was just finally respecting that different people care about different things, even for the same job.

The Consultant Who Personalizes Proposals in Real Time

A management consultant in Chicago runs a two-person firm. Before every proposal, she spends 20 minutes pulling three things from a prospective client's LinkedIn company page, their most recent press release, and any notes from her discovery call: their current stated priority, a specific challenge they mentioned, and the language they actually use to describe their business. She pastes those three things into Claude with a template prompt and gets a customized executive summary in about four minutes.

Her close rate on proposals improved from roughly 1 in 5 to closer to 1 in 3 over six months. She attributes most of that to one specific change: prospects now frequently tell her in the first follow-up call that her proposal 'felt like you actually listened.' She did listen. The AI just made sure that listening showed up on the page, in the right language, without requiring her to rewrite from scratch for every client.

Putting It Into Practice

Start with one piece of content you already send to multiple audiences, a newsletter, a proposal template, a follow-up email, a job posting. Identify two distinct groups who receive it and write two short audience briefings: who they are, what they care about, and what would make them act. That's the entire setup. Everything else is just asking AI to execute against those briefings. The discipline is in the briefing, not in the prompting.

Use free tools to start. ChatGPT's free tier handles this well. Claude's free version is strong for longer-form content like proposals and reports. You don't need integrations, automation, or a CRM that talks to your email platform. Write the briefing in a document, paste it into the chat window, get the output, review it, and send it. That workflow takes under 30 minutes for two segment variations once you've done it once.

Track one metric per send, open rate, click-through, reply rate, or close rate, and compare your segmented sends against your previous single-version baseline. You need at least four to six sends before the pattern is meaningful, but most people see a directional difference on the first attempt. When it works, the feedback loop is fast: people respond more, ask better questions, and tell you the message felt relevant. That's the signal. Build from there.

Key Takeaways

1. Personalization means the message itself changes, not just the name at the top. Different segments need different hooks, different emphasis, and different language.
2. The briefing is the real work. Telling AI who the audience is, what they fear, what they want, and how they speak produces dramatically better output than any prompt formula.
3. You don't need automation or integrations to start. A free AI tool, a clear audience description, and one content piece is enough to run your first segmented campaign.
4. Every professional role has segmentation opportunities, marketing emails, job postings, client proposals, internal communications, and sales follow-ups all benefit from audience-specific framing.
5. Track one metric per send and compare against your baseline. Four to six sends gives you a meaningful pattern. Most people see directional improvement on the first attempt.
6. AI executes the thinking you do in advance. The quality of your segmentation thinking determines the quality of the personalized content, not the sophistication of the tool.

Click to enlarge