Across several conversations, one theme was clear: How do you move forward with AI without putting your organization at risk?

Some teams are on pause. Others are deep in circular debates. Most are stuck asking the same questions: What’s allowed? What’s risky? Who decides? Efforts stall out—not because people don’t care, but because no one is quite sure where the lines are, or who gets to draw them.

• If you’re in leadership, your caution is valid. Bad AI use can leak data, erode trust, or create real liabilities.

• If you’re doing the work, so is your frustration. You see opportunities every day, but without guidance, you’re either stuck or improvising without air cover.

The core problem is that most AI policies aren’t built on real concerns or real use cases. A good policy isn’t just a list of rules. It’s a map that makes clear what your organization wants to avoid, and what it’s ready to enable.

This piece is a framework for creating that map.

The operational case for an AI policy

An AI policy isn’t just a safeguard. It’s an operational tool that dictates whether innovation moves forward or stalls out.

• For leadership, a policy sets boundaries before something goes wrong. It protects the company’s data, IP, and reputation.

• For mid-level managers and operators, the policy is about permission. It answers the question: What are we allowed to try? Without clear guidance, people hesitate or, worse, build in the dark.

Done well, a policy enables innovation. It removes the invisible tripwires that stop people from exploring what’s possible.

The shared truth across the org chart is that everyone wants to do the right thing. A clear policy gives people a place to stand and the confidence to move.

Specificity is key to a workable policy

Most AI policy efforts get stuck in a weeks-long game of ping-pong between legal, IT, procurement, and anyone brave enough to propose a use case.

To avoid it, start with outcomes.

• If you’re in leadership, your job is to define the non-negotiables. What are the scenarios you absolutely want to prevent? Data leakage? Public-facing hallucinations? Automation with no human oversight? These are the anchors for any effective policy.

• If you’re managing a team, make the value legible in terms of outcomes. Don’t just say, “We want to use AI for content.” Say: “We’re trying to draft 80% of internal help articles automatically so support leads can focus on edge cases, and we’ll exclude any ticket involving account credentials or billing.” That’s clear, bounded, and defensible.

Focusing on outcomes does two things:

1. It keeps the policy grounded in reality.

2. It gives everyone, from lawyers to line managers, something real to evaluate.

Policies should be built around what actually matters: the consequences you want to avoid, and the value you want to unlock. The “how” can follow.

Match the decision to the level of context

Not every part of an AI policy should be decided at the top. In fact, when too many decisions get centralized, the result is overcaution, underuse, and a document that’s impossible to apply in practice.

• If you’re in leadership, your job is to draw the bright lines: no unreviewed model output in customer-facing channels, no automation in areas with regulatory or contractual risk, and no tools where data flows can’t be clearly audited. These aren’t feature-level decisions. They’re existential guardrails that protect the business.

• If you’re managing a team, your job is to apply those guardrails to the real context of your work. What kind of exposure would actually matter in your domain? Is it customer contact details? Strategic product planning? Internal conversations about deals in motion? What needs human review not just in theory, but in practice?

Policies only work when they reflect how work actually happens. That means pushing implementation details down to the people who live in the specifics. They know the tradeoffs and the opportunities.

The further a policy is from the work, the more likely it is to say “No” to everything. Put decisions closer to the action, and “Yes, as long as…” becomes possible.

Pilots are the path to progress

Almost every organization has someone waiting on someone else. Legal is waiting on product. Product is waiting on leadership. Teams are waiting on “the policy.” Meanwhile, nothing moves.

Pilots are how you break that loop. You don’t need a finished policy to get started. You need a way to learn safely and visibly.

• If you’re in leadership, don’t wait for a perfect policy. Create a lightweight path for teams to propose and run AI pilots. Set the high-trust boundaries: what must be protected, what lines shouldn’t be crossed. Then invite teams to fill in the specifics. You can revise as you go. The important part is sending a clear signal: we’re willing to move.

• If you’re managing a team, don’t wait for permission. Propose something small, specific, and safe. For example: “We want to use LLMs to generate first drafts of product copy. Anything referencing pricing, performance claims, or legal disclaimers will be written or reviewed by a human before publication.” That kind of framing builds trust. It shows you’re being deliberate, and it gives leadership something concrete to react to.

Pilots turn intent into insight. They transform theoretical risk debates into real-world feedback and give your organization a safe, visible way to learn its way into better policy.

Every organization navigating AI is balancing two instincts: protect and progress. A good policy does both.

• If you’re in leadership, your teams need guardrails, but they also need green lights. Set the boundaries that matter most, and then invite people to move within them. Trust doesn’t mean hands-off. It means being clear about what matters and letting others help figure out the “how.”

• If you’re managing a team, don’t wait for top-down clarity. Start small. Define a use case, flag the risks, propose the boundaries. Show that it’s possible to move responsibly. That’s what policy needs more than anything right now: examples.

AI policy shouldn’t be a tug-of-war between caution and ambition. It should be a shared conversation about what your organization actually values, and how it plans to act accordingly.

Policy follows progress. Start making both.

It’s remarkable how it shook up the AI world. Not because it was the smartest model. Not because it was breaking records on academic leaderboards.

Because it was cheaper.

The DeepSeek moment signals a shift in AI. For years, the industry operated under a maximalist philosophy. Starting with GPT-1, each successive generation was orders of magnitude bigger (and more expensive) than the last. The logic became self-reinforcing: a model was considered better because it was bigger, and it had to be bigger because that’s what made it better.

To prove this was true, we relied on tests—esoteric benchmarks understood only by AI researchers, along with more familiar but equally dubious measures, like the SAT and the bar exam, which GPT-4 passed (or did it?).

But did any of this make AI meaningfully better for users? Arguably not. That’s not to say AI hasn’t improved since OpenAI launched ChatGPT in 2022—it has. But the real progress didn’t come from pushing intelligence to its theoretical limits through ever-bigger models. It came from making AI faster, cheaper, and more accessible.

If I’d asked people what they wanted…

…they’d have said a faster horse, said Henry Ford (or did he?).

Depending on who you ask, that maxim is either an indictment of market research—a call for visionary inventors to ignore what consumers say to build what they need—or the opposite: a lesson in deep empathy, showing that true innovation comes from understanding the intent behind what people ask for, not just taking their words at face value.

Both interpretations, though, agree on one thing: faster horses are bad.

But what if they aren’t? What if the real breakthroughs—the ones that reshape industries and change how people live—come not from chasing the grandest vision, but from making things cheaper, easier, and well, faster?

Which AI development do you think made the bigger impact in users’ lives—passing the bar exam or any of the following?

• Longer context windows that improve memory in conversations.

• Lower inference costs that make AI commercially viable.

• Lower latency that makes responses feel instant.

• Higher reliability that makes models more dependable in production.

• Seamless integration into existing tools.

This isn’t a dismissal of fundamental AI research—pushing the boundaries of what’s possible will always matter. And benchmarks aren’t without use—after all, the remarkable thing about DeepSeek is that it was cheap and performed just as well as proprietary models in industry benchmarks.

But when it comes to real-world impact, progress isn’t measured by leaderboard rankings or viral demos. It’s measured by whether AI makes people’s lives easier in tangible ways. A model that can pass the bar exam doesn’t change much for us. A cheaper model does.

At Collected Company, we call this Market-Aligned Product Development (MAP-D)—the idea that AI should be built not just for intelligence or hype, but for impact. That means two things:

• Deep Technical Understanding – Knowing how AI works at a fundamental level to understand what it can do and what it can’t.

• Deep Business Understanding – Identifying where AI can actually move the needle for companies.

This isn’t just a theory—it’s a pattern we’ve seen play out before. And history offers a clear example of what happens when companies mistake technical superiority for real-world impact.

“Better” is an empty ideal. “Faster” is what the people want.

A great example of a leader in market-aligned product development was Ford’s competitor, Alfred Sloan, who headed up GM in the 1920s.

As this HBR blog post points out, GM ate half of Ford’s market share over the course of just 5 years. How? By selling cars “for every purse and purpose.” In other words: products that fit people’s actual needs.

Ford had the better car. The Model T was a marvel of engineering—efficient to produce, technically advanced, and built to last. But better is slippery. GM, on the other hand, made cars that were more practical, more customizable, and more accessible to different kinds of buyers. That’s what actually mattered.

The AI industry has the same problem. It chases better—models that score higher on benchmarks, pack in more parameters, and showcase flashier capabilities. But better is an abstraction, defined and redefined by whatever metric happens to be in vogue.

Faster, though? Faster is real. Faster is felt. It’s the difference between an AI tool that seamlessly integrates into a workflow and one that slows it down. Between something people rely on and something they abandon.

The biggest impact in AI won’t come from pushing intelligence to its theoretical limits. It’ll come from making AI work the way people actually need it to. And most of the time, that means not just making it better, but making it faster, cheaper, and easier to use.

Over the last two months, predictions about AI have flooded the internet. Most posts echo three themes:

• Agentic systems: AI that doesn’t just advise or recommend—it takes action on the user’s behalf.

• Multi-agent systems: Picture a “creative director” agent and a “copywriter” agent collaborating to refine marketing copy.

• On-device models: Systems that can accomplish tasks based on your data without compromising privacy.

These trends will likely prove correct if only because they have momentum: they’re tied to large investments from VCs, tech giants, and advisory firms. It’s no coincidence you see them in 2025 predictions from Google, Snow Flake, Deloitte, PwC, and Salesforce.

Our chief prediction, though, is something you won’t find on those lists:

The AI bubble is about to burst. And that’s a good thing.

It might seem odd for two people who started an AI firm to call their sector a bubble—even more so to welcome a burst.

But the reality is that most AI applications so far have been underwhelming. The hype has forced AI into awkward places—tacked onto existing products or bought in enterprise packages that nobody knows how to use. These are decisions made out of fear of missing out, not real need.

Once the froth subsides, people can focus on the actual problems these tools solve. And make no mistake, those problems exist. Zero interest rates are gone, and everywhere, teams are being asked to do more with less.

AI can help, but only if it’s used thoughtfully.

We believe that using AI effectively starts with identifying workflows that can be improved with models. That’s how you free up teams to do the creative work that drives growth.

Here are four such workflows—ones we at Collected Company have seen teams tackle with real success.

Workflow #1: Wrangling Data

Examples

• Pulling specific numbers from PDFs and converting them into a CSV

• Categorizing files and cleaning up naming conventions

• Extracting contacts or email addresses from messy text logs

Why AI Helps

These tasks have clear, rule-based outputs—“find X and place it in Y”—which is where AI shines. But beware: AI can’t magically impose order. A human must design a taxonomy that fits your organization’s workflows, which is the real interesting part of the problem. By allowing AI to handle the busy work, like manually cleaning and formatting data, you free up your team to focus on strategic work.

Workflow #2: Synthesizing Insights from Multiple Sources

Examples

• Gleaning common themes from dozens of user interview transcripts

• Pulling out recurring pain points across multiple support tickets

• Summarizing findings from a stack of market research reports

Why AI Helps

LLMs excel at spotting patterns and highlighting key topics across large volumes of unstructured text. These tasks typically require sifting through hundreds of pages, which is both tedious and error-prone for humans. By removing this tedium, teams get to spend more time digging into the why and the how—the insights that will actually help your business grow.

Workflow #3: Document Discovery and Tagging

Examples

• Automatically tagging documents with key metadata, like adding client names to contracts, or categorizing invoices by vendor

• Building a “knowledge discovery” portal that lets users quickly locate relevant files

• Continuously updating a knowledge base so important assets don’t get buried

Why AI Helps

AI improves file discovery by analyzing content rather than relying solely on keywords. This means teams spend less time searching for the right document and more time driving projects forward. Whether you’re a film editor tracking down that perfect B-roll shot or a marketer hunting for a specific customer quote, AI minimizes bottlenecks and empowers you to focus on producing high-impact work, not finding the missing puzzle piece.

Workflow #4: Refactoring and Repurposing Existing Information

Examples

• Turning a recording of business development meeting into a written draft of a proposal

• Converting multiple meeting transcripts into a concise weekly executive update

• Transforming long internal memos into bite-sized bullet points for a company-wide newsletter

Why AI Helps

In professional settings, countless hours are wasted reformatting existing information to suit different audiences. AI streamlines this process, converting raw material into polished, audience-ready outputs. By automating this formatting work, teams can redirect their efforts toward creating original, high-value content that drives growth and innovation.

Used thoughtfully, these four workflows unlock genuine efficiency gains—without diluting the human expertise that truly matters.

These studies are silly. Asking someone whether they’re using a tool is not nearly as useful as asking whether that tool is any good—and, most importantly, why.

That’s where this essay comes in. We’re going to walk through the following marketing use cases and qualitatively evaluate how good AI is at each:

• Generating imagery

• Naming products and features

• Writing marketing copy

• Validating marketing hypotheses

• Analyzing data

• Streamlining marketing operations

Importantly, we’ll also dig into why AI fares well or poorly in these areas, focusing on the technical underpinnings that make it so.

We won’t be teaching you prompt engineering or telling you which specific tools are best for each task; rather, we’ll give you a framework for understanding why the technology shines in some contexts and flounders in others.

Methodology

Before we start, three notes on methodology:

• A qualitative lens: There isn’t a simple metric for “good marketing,” so we rely on our experience—both authors started their tech careers as product marketing managers. To evaluate AI’s performance in each use case, we ask, “If an intern produced this work, would we think it’s any good?”

• Focus on models, not prompts: A skilled prompt engineer will, without question, get better results than what we provide below as examples. However, the underlying model constraints remain. Our examples highlight those baked-in limits, regardless of how expertly you phrase the request.

• Tools keep evolving: What’s true in December 2024 might be outdated six months from now. But by focusing on how these models fundamentally function, we hope to give you a clear lens for evaluating any new tools that come along.

Generating Imagery

How it does

Hit-or-miss. Performance really depends on what you’re trying to generate. AI tools are good at softer, organic-looking visuals, like people and animals.

But they struggle with crisp, clean images with sharp contrasts and straight lines. This makes them bad at things like UI and logos.

In the logo below, the plane is asymmetrical, several lines are uneven, and there’s a smudge near the bottom edge—all of which detract from the precision needed for a polished logo.

And the UI is, well, bad.

It also makes them bad at product imagery. While the plane in the example below is convincing, if I were a marketer at Boeing, I’d worry about whether details like the size of the wings or the position of the engines actually match a Boeing plane.

The Technical “Why”

1. Data matters: Most image generation models train on massive datasets of organic, everyday images. That’s great for generating “soft” visual content—natural lighting, fuzzy edges, and broad shapes—but less helpful for the vector-like precision or brand consistency needed for product imagery or logos.

2. Model architecture: Many popular tools use diffusion models. These algorithms start with noise and iteratively refine the image. This means they excel at approximating complex, “messy” images but struggle with structured or rigid designs like infographics, product mockups, or logos—anything needing tight, consistent lines and proportions.

Tools to Try

• DALL-E in ChatGPT: The easiest way to generate images.

• Adobe Photoshop (Generative Fill): Lets you easily add or remove elements in an image.

• Imagen 2: Has been tuned for background swaps. (Note: As of December 2024, Imagen 2 is only available via API, so implementation will require a technical team. We can help.)

Naming

How it does

Very badly. When it comes to naming, AI models usually return bland suggestions that are not ownable (“Stride” for a shoe brand) or awkwardly mashed-up hybrids (“UrbanAthlete” or “FusionStep”).

While this might be okay for initial brainstorming, it rarely yields the kind of fresh, memorable brand names that truly stand out.

The technical “why”

• Tokenization limits creativity. Many names are wholly invented words (“Xerox,” “Expedia”) or altered versions of existing words (“Verizon”). Language models are technically incompatible with this type of thinking, since they work by predicting sequences of “tokens,” which are typically chunks of words (e.g., “strid” + “e”) rather than single letters or syllables. Because of this, the model cannot “invent” words at a granular, character-by-character level—an “X” followed by an “E” followed by an “R” followed by an “O,” etc. Instead, it rearranges existing pieces of language. That’s where we get the recycled feel of AI-generated names.

Tools to try

Don’t expect magic—but if you want to see for yourself, you can test out ChatGPT, Gemini, or Claude.

Writing marketing copy

How it does

It depends. When tasked with writing marketing copy, AI models typically produce text that’s clear and coherent—but also bland and predictable.

This can be helpful for expressing something plainly, or adapting an existing message—shortening a paragraph, or modularly changing the highlighted features based on the audience.

What AI can’t deliver, however, is the kind of unique insight or compelling “hook” that makes your product stand out. Marketing copy isn’t just about clear writing—it’s about telling a story that highlights a product’s unique selling proposition. AI is optimized for what’s statistically common; “unique” is not its strong suit.

The copy below, for instance, could apply to any maker of athletic shoes.

The technical “why”

• Generative AI models are fundamentally designed to generate output that’s “likely,” not “original.” Language models predict text by choosing the most probable next word at each step. This probability-driven method means the final output avoids the unconventional phrases that make for unforgettable marketing. Consider taglines like “Think different” (Apple) or “Impossible is nothing” (Adidas). They break standard grammar or common phrasing to grab attention. Because AI picks what’s statistically likely, it usually steers clear of these bold, memorable choices.

Tools to try

Language models like ChatGPT, Gemini, or Claude.

Validating marketing hypotheses

How it does

Surprisingly good—for now. Recent research from HBS and Haas/Columbia/University of Alberta suggests that LLMs can mimic human survey responses with striking accuracy—up to 75% similarity. In theory, this means marketers might skip sending consumer surveys and simply ask an AI to review and validate concepts.

The technical “why”
At their core, language models choose the most likely next words based on massive datasets and continual human feedback. In other words, they’ve learned to be “mid”—to converge around the average human response. So, if you ask these models what people think of a new product idea or marketing angle, they can provide a reasonable stand-in for an actual focus group.

However, there are a few important caveats:

1. Layered representation problem. Surveys are already a representation of real-world opinion. Using an AI (another layer of representation) on top of that can feel like you’re stacking abstractions—fine in theory, but it may not always capture genuine consumer sentiment.

2. Panel selection still matters. While these studies show LLMs can “impersonate” certain demographics, niche groups are harder to emulate. If the model hasn’t seen enough data from a specific subset of people, its responses are less reliable.

3. Changing data mix. As training data evolves—especially with more synthetic inputs—and as models shift toward machine-based fine-tuning methods, LLMs may drift further from the human “median.” Today’s near-accurate results could become less trustworthy tomorrow.

Tools to try

If you’re curious, you can experiment with ChatGPT, Gemini, or Claude for a quick “pseudo focus group.” Just remember you’re dealing with a statistical approximation, not a direct pipeline to collective human consciousness.

Data analysis

How it does

Pretty darn good. AI has powered data analysis for quite some time, though usually through quantitative methods like regression or clustering. Even so-called “qualitative” analysis still relied on quantitative methods, ignoring the actual semantic content of things like product reviews or interview transcripts.

But with LLMs, you can feed large sets of qualitative data into a model and get back meaningful insights—like recurring themes in user feedback or notable quotes from focus groups. Instead of simply grouping or ranking data objects, LLMs can actually interpret them, generating high-level summaries that capture the essence of what people are saying.

The technical “why”

• Well represented in the data. A lot of content on the internet follows the same structure: data, then conclusion—think scientific papers, well-reasoned opinion pieces, or even Reddit threads (on a good day). LLMs have essentially “learned” this structure from massive training sets.

Tools to try

NotebookLM and Perplexity. These platforms let you upload, query, and summarize your data, helping you move beyond raw numbers into more nuanced, qualitative insights.

Marketing operations

How it does

Good—though we’re using “operations” as a broad umbrella to mean all the small tasks that keep workflows humming. For instance:

• Getting help writing a clearer, more concise email

• Using meeting transcripts to generate executive updates

• Using AI to annotate assets so they’re easier to find later

None of these tasks are especially glamorous, but they can have a noticeable impact on both productivity and output quality when aggregated over time.

The technical “why”

Large AI models excel at tasks that mirror how they’re trained: in discrete, narrowly defined steps with clear feedback. Breaking a real-world job into smaller pieces—like “extract themes from this data” or “polish this paragraph”—mimics that training process, making it easier for the model to perform well.

As a rule of thumb, if you can split a process into smaller, data-heavy components, an LLM can probably handle one or two of those steps effectively—whether that’s summarizing transcripts, labeling images, or writing short blurbs.

Final Thoughts

Marketers should understand why AI works. These models aren’t magic; they excel at tasks that align with their technical underpinnings and struggle when we ask them to invent something genuinely new or nail the tiniest details.

Keep that in mind, and you’ll know exactly where AI can supercharge your workflows.

This is the logic many businesses fall into when building AI solutions: start from scratch, build a custom model, and sprinkle in proprietary magic. But for 90% of companies, the better answer isn’t hiding in some grand invention—it’s already out there, waiting to be used.

The Siren Song of Custom AI

I get why companies are tempted to build their own AI models. Every business is different, with unique challenges and proprietary data that could potentially create a game-changing AI solution. However, this approach comes with significant drawbacks:

1. Astronomical Costs: Training a large language model from scratch is an incredibly expensive endeavor. For context, it's estimated that training BloombergGPT cost between $3-4 million and took 3-4 months.

2. Technical Complexity: Building an AI model requires extensive cross-domain knowledge spanning NLP/ML, subject matter expertise, and software/hardware proficiency.

3. Data Demands: Training an effective LLM requires massive amounts of high-quality, diverse data.

The Middle Ground: Tuning

Many businesses hear about “tuning” an existing model and think it’s the perfect middle ground. Whether it’s full-scale fine-tuning with large datasets, self-implemented parameter-efficient techniques like LoRA, or a simple “fine-tuning” endpoint offered by a paid provider, the goal is the same: take something off-the-shelf and try to make it work better with your data.

Tuning has its advantages, such as improved task-specific performance and greater control over outputs. But it still presents challenges:

1. Data Quality Dependence: The success of fine-tuning heavily relies on the quality and relevance of your training data, and as such, it risks amplifying biases present in your training data.

2. Resource Intensive: Fine-tuning may sound simple, but it still demands significant expertise. Just because an AI provider offers a “tuning” endpoint doesn’t mean you’ll get high-quality results right away. Success depends entirely on your ability to align the right data with clear business objectives—and that’s far harder than it sounds.

3. Ongoing Maintenance: Tuning locks you into a particular version of a model. When new versions are released—and they will be—you’ll need to re-tune to keep up, increasing overhead and slowing down updates.

The Pragmatic Approach: Off-the-Shelf Solutions

For 98% of businesses, an off-the-shelf AI solution is the most practical and effective choice. Here's why:

1. Cost-Effective: Off-the-shelf solutions are significantly cheaper to implement than custom-built models.

2. Rapid Deployment: These solutions can be deployed quickly, allowing you to start deriving value almost immediately.

3. Reduced Risk: Off-the-shelf products are already proven in the marketplace, minimizing the risk of project failure.

4. Ongoing Support: You benefit from the provider's continuous improvements and updates.

Start Smart, Scale Later

While the idea of a bespoke AI model is enticing, the reality is that most businesses don't need—and can't afford—to build one from scratch. By starting with off-the-shelf solutions, you can quickly prototype, learn, and iterate.

Even cutting-edge AI companies like Perplexity began their AI journeys with off-the-shelf models before developing more custom solutions. The key is to start small, focus on solving specific problems, and let your AI strategy evolve based on real-world results and business needs.

Prototyping: The Best First Step

Before even considering building a custom model, it's crucial to prototype with off-the-shelf solutions. This approach offers several advantages:

1. Quick Validation: You can rapidly test whether AI can solve your specific problem without significant investment.

2. Cost-Effective Experimentation: Off-the-shelf models allow you to experiment with different approaches at a fraction of the cost of custom development.

3. Learning Opportunity: Working with existing models helps you understand the nuances of AI applications in your domain, informing future decisions about customization.

At Collected Company, we specialize in helping businesses take this first step. We guide companies through prototyping, helping them explore the best tools and approaches to fit their needs. If you’re facing this challenge yourself, reach out—we’d love to help you find the smartest path forward.

Take it from someone who makes a living with AI tools: That’s snake oil.

Useful technology is not about the tool. It's about the problem you're solving.

At our firm, Collected Company, we’ve been integrating AI tools by first looking at our actual work—where we’re losing time, where processes feel clunky, where human energy is being wasted on repetitive tasks—and only then thinking up ways AI can help.

Here are three examples.

Proposal Generation: Turning Conversation into Content

To write a proposal, Nicole and I would spend an hour on a video call discussing the project. Then, one of us would spend at least another hour capturing all of those points in a document. Essentially, we were drafting the proposal twice—first verbally, then in written form.

We figured we could use AI to jumpstart the document-writing, allowing us to keep our energy and creativity for the live conversations and ideation. Now, our process involves recording and auto-transcribing our conversation, and then using language models to generate a first draft.

We happen to use Grain.co for recording and transcribing our calls and ChatGPT or Gemini to generate the first draft, but the magic isn't in those specific tools—it's in recognizing the redundancy and finding a smarter path.

Interview Insights: From Raw Conversation to Actionable Intelligence

Stakeholder interviews are often a key part of our projects. Whether it’s talking to client teams to understand their processes or diving deep into customers’ experiences, there’s no substitute for hearing from people fist-hand.

AI can never replace the role of a human interviewer, and it cannot generate the level of analysis and insight we want to deliver. But it can dramatically streamline the insight validation process—something that used to require hours of digging through interview transcripts in search of quotes.

Now, we can quickly cross-reference and validate our hypotheses with tools like Google's NotebookLM, which allows us to query up to 50 sources simultaneously. This means we can identify corroborating evidence or contradictions across multiple sources, transforming raw interview data into more robust, well-substantiated findings.

Adaptive Messaging: Making AI Actually Useful in Marketing

Most of the time, creating marketing content with AI doesn't reduce inefficiency; it multiplies it. Wrestling a language model into producing exactly what you want is time-consuming and bypasses the crucial strategic exercise of understanding your offering and positioning.

The real opportunity with language models isn't wholesale content generation, but adaptation. Marketing requires nuanced messaging: a pitch to an investor shouldn't sound like a pitch to a potential client. This is where AI truly shines.

At Collected Company, we use AI as a translation tool, not a content creator. We start with a carefully crafted, human-written narrative and use AI to subtly adjust tone, emphasis, and language for specific contexts. By harnessing AI's ability to refine and adapt, we communicate more precisely and efficiently.

The Real AI Revolution: Problem-Solving, Not Tool-Collecting

AI, at its best, is an extension of human creativity and problem-solving. We shouldn’t start by asking, "What can this tool do?" We should start by asking, "What can we do better?"

Each AI integration is a custom solution, tailored to our specific challenges. It's less about adopting the latest shiny object and more about understanding the unique ecosystem of our work.

Your Turn: A Challenge

Map out your workflows. Where do things get stuck? Where are humans doing work that feels mechanical? Where is time mysteriously disappearing?

Those are your AI integration opportunities.

The future of work isn't about having the shiniest tools. It's about solving real problems with thoughtful, purposeful technology.

Whether you’re leading a team, running a business, or looking to work smarter with AI, we cut through the noise to focus on what actually works. From actionable tips to real-world examples, Collected Thoughts is your guide to making AI work where it matters most.

About Collected Company

Collected Company partners with organizations to bring AI into their product strategy. We turn ideas into reality to solve actual problems, fast.

Our founders, Teo and Nicole, have a cumulative 20 years of experience building products and features in the tech industry. Most recently, Teo was a Senior Product Manager on the API for Gemini, Google's generative AI model; and Nicole was a Senior Product Manager on Google Search.

Now, they enable organizations to envision, build, and test efficiently and effectively. Whether running brainstorms conducting user research, or designing and developing prototypes, Teo and Nicole can help get AI on your roadmap.

To learn more about Collected Company, visit collected.fyi.

Subscribe now