A SaaS founder’s reality check vs linkedin AI hype | Vlad Zhovtenko @ RedTrack

AI is changing how companies work, but that does not mean every business needs to fire its team, build an army of agents, or raise venture capital to survive.

Vlad Zhovtenko, founder of RedTrack, has worked in digital marketing since the early days of the internet. Having seen the dot-com era, the rise of SaaS, the decline of third-party cookies, and the evolution of paid acquisition, he offers a practical view of what AI changes—and what remains the same.

For SaaS founders, marketers, and operators, the central message is straightforward: AI is a powerful tool for processing work, but it is not a replacement for judgment. The companies that benefit most will be those that build strong first-party data foundations, redesign the right workflows, and keep humans responsible for meaningful decisions.

AI Is the New Internet, Only Faster

The current AI moment resembles the arrival of the internet roughly 30 years ago. At first, many people did not see the internet as a serious business tool. It was simply a new way to access information, communicate, and work more efficiently.

Over time, the internet became so embedded in everyday life that the label “internet company” stopped meaning much. Businesses were no longer defined by whether they used the internet. They simply used it.

AI is likely to follow the same path. The major difference is speed. Changes that took five years during the internet era may take one or two years with AI.

Unlike short-lived hype cycles around technologies such as NFTs, AI is already changing how people:

• Find and consume information.
• Produce knowledge and content.
• Complete repetitive work.
• Interact with software and services.
• Make decisions inside companies.

The final form of this shift is still unclear. But the important question for each individual and company is not whether AI will exist. It is whether using it offers a meaningful advantage in the work they do.

Not every role requires the same level of AI adoption. Just as some jobs did not depend heavily on the internet 20 years ago, some roles may not need advanced AI capabilities today. Still, AI will change the nature of work, eliminate some tasks, create new roles, and remain part of the business landscape until the next major technological cycle begins.

AI Has a Cost, Not Just a Benefit

AI adoption is often discussed as though it has no trade-offs. But newer generations are already looking at the technology through a broader lens.

AI can be an excellent learning tool. Used well, it can function like a tutor: answering questions, providing explanations, and helping someone explore concepts they do not yet understand. The most valuable use is not blindly accepting an answer, but asking follow-up questions:

• Why is this the right answer?
• Is this a complete answer?
• What do the unfamiliar concepts in this explanation mean?

At the same time, large-scale AI systems consume resources. The infrastructure behind them requires energy, water, and computing capacity. This does not make AI unusable, but it does mean thoughtful use matters. A useful tool still has a price.

How RedTrack Began: Solving the Paid Acquisition Data Problem

Vlad’s path into digital marketing began around 2000, after discovering early publications covering media buying, referral programs, website development, and online marketing. He left a role at Ernst & Young to join a digital media startup—a move that seemed risky to many people at the time.

That decision led to a career in digital marketing and eventually to leading paid customer acquisition for SaaS businesses.

At the time, paid acquisition for SaaS was still far from standard practice. The challenge was understanding what was happening across multiple marketing channels and connecting ad spend to business outcomes.

The available technology could not handle the level of data analysis required. The workaround was a complex system of interconnected Excel files, reports, and manually updated analytics.

That experience shaped the idea behind RedTrack: media buyers needed better systems for collecting and using their own performance data, without spending years building internal reporting infrastructure.

The company started with ad tracking and conversion analytics, expanded into attribution, and evolved toward automating the repetitive operational work involved in media buying.

The long-term goal is not simply to make media buying faster. It is to remove routine manual work so that media buyers can focus on creative and strategic decisions.

Why First-Party Data Is the Foundation of Modern Media Buying

The decline of third-party cookies and the changes introduced by iOS 14.5 transformed the way paid advertising operates.

Attribution still matters, but it helps to define what attribution is actually for. Marketers do not need perfect, universal knowledge of every conversion and every touchpoint. They need a consistent measurement framework that allows them to compare decisions over time.

If a team makes a change and its consistent data shows that return on investment improves or cost per acquisition falls, that change can be evaluated as a positive one.

Historically, media buyers often relied heavily on reporting from ad platforms such as Meta, Google, TikTok, and others. When third-party cookies were more widely available, platforms had greater access to data and could offer more complete reporting. Even if platforms claimed credit for more conversions than they truly drove, marketers could still use the same data set consistently to make decisions.

That changed as privacy restrictions increased.

Media buying moved away from a model where humans set detailed rules inside platforms and toward a more algorithmic model. Today, media buyers increasingly define the boundaries:

• The target audience.
• The campaign objective.
• The acceptable cost or performance threshold.
• The conversion events that matter.

The platform’s algorithms then make many of the delivery decisions.

For those algorithms to work effectively, ad platforms need conversion data sent back to them. Pixels alone are no longer enough. Companies need to capture, process, enrich, normalize, and send back first-party data about clicks, conversions, and customer outcomes.

This is why first-party data is no longer an optional analytics project. It is part of the operating foundation for efficient paid acquisition.

How AI Could Change Attribution and Media Buying

AI has not yet fully broken attribution, but it is likely to reshape it.

One early shift is happening in search. People are increasingly finding businesses through large language models and AI-generated answers rather than only through traditional search engine result pages. This traffic can still be attributed in some cases, but businesses do not directly control these channels in the way they control paid campaigns.

A more substantial change could happen when AI agents begin taking conversion actions on behalf of people. Consider a future where someone sees an ad, then asks an AI assistant to purchase the product, book a service, or buy a ticket.

The path from ad impression to purchase becomes much harder to connect. The familiar sequence of impression, click, website visit, and conversion may become less common.

Over time, human clicks may decline while actions performed by AI agents increase. Media buying could return to a version of the old “black box” problem: ads go in on one side, business outcomes appear on the other, and teams need to determine what caused the change.

However, the same AI that complicates attribution may also help solve that problem. AI systems can process far more information than a human analyst can reasonably handle. They may become essential for matching actions to outcomes and identifying patterns across increasingly complex customer journeys.

What Every Paid Acquisition Team Should Do Now

Companies engaged in paid advertising should begin collecting and organizing their own first-party data now.

This data is useful for conversion APIs, but its value goes beyond ad-platform optimization. It will also become the foundation for AI agents that support decision-making inside the company.

The advantage will not simply belong to the company with the most advanced agent. Many organizations will have access to similar AI models, cloud infrastructure, and workflow tools. The real advantage will come from the quality of the data available to those systems.

AI needs access to raw business records, not only derivatives created by advertising platforms. That includes data about actual business performance, campaign activity, and the changes made over time.

Companies should capture:

• Clicks and conversions.
• Campaign performance data.
• Revenue and customer outcomes where available.
• Changes made to campaigns, targeting, budgets, and creative.
• A historical record of the actions taken and their results.

With this foundation, a team can eventually ask much better questions: What changed? Which actions preceded an improvement or decline? What patterns have been invisible in weekly and monthly reports?

At RedTrack, connecting AI tools to years of historical data produced insights that had not emerged through routine analysis. The team was already reviewing performance daily, weekly, and monthly. Yet AI condensed an enormous amount of processing work into minutes and surfaced new lines of inquiry.

The key point is that AI did not magically replace expertise. It accelerated the path to insight. Human judgment was still needed to investigate the findings, cross-check them, test the conclusions, and act on them.

AI Should Support Decisions, Not Own Them

There is an important distinction between using AI to prepare a decision and allowing AI to make the decision.

AI can rapidly process records, summarize information, compare outcomes, and suggest patterns. A human could potentially reach many of the same conclusions, but doing so might require days or weeks of digging through data.

That is where AI creates value: it condenses processing work.

But AI systems are not all-knowing. They make highly educated guesses based on patterns in the information available to them. They can be wrong, including when the answer should appear obvious or widely documented.

The internet contains inaccurate information. If an AI system follows an incorrect pattern or starts from faulty assumptions, it can return an incorrect answer with confidence.

For that reason, the person responsible for the outcome must remain responsible for the decision.

A useful principle is simple: AI helps prepare the decision, but the human makes it.

This also makes custom internal knowledge systems increasingly important. Rather than forcing an AI model to infer everything from broad public information each time, companies can give it access to their own validated knowledge, past insights, and operational records.

Processing Work vs. Cognitive Work: Where AI Belongs

Not all work should be treated the same way. For a SaaS company, much of the work falls into two broad categories: processing work and cognitive work.

Processing work involves moving, organizing, transforming, or acting on existing information within known rules. It does not require the creation of genuinely new information or nuanced human judgment at every step.

Examples include:

• Summarizing support tickets.
• Creating call summaries.
• Researching prospects using established criteria.
• Moving information between systems.
• Preparing standardized reports.
• Following defined operational procedures.

This is a strong starting point for AI adoption because the work is often repetitive, time-consuming, and structured.

Cognitive work requires deeper reasoning, judgment, accountability, and the ability to make decisions in uncertain situations. This is where human involvement remains essential.

The goal should not be to hand five employees an AI subscription and expect a transformed organization. Effective adoption requires rebuilding the workflow itself.

Instead of adding AI as an extra step, companies can redesign a process so that AI initiates the processing work, prepares the output, and passes it to a human for validation or approval.

That is fundamentally different from occasional prompting. It is a new operating model.

The Four Levels of AI Adoption

Many companies talk about becoming AI-powered, but most are still in the early stages of adoption. A practical framework can help clarify where a team actually stands.

1. Basic prompting: Using AI like a search engine by asking simple questions and retrieving answers.
2. Intentional prompting: Writing detailed, specific prompts designed to produce a more useful result.
3. Connected data: Giving AI access to relevant company data sources, knowledge repositories, and business context.
4. AI-powered workflows: Redesigning operational processes so AI performs the processing work first and humans validate, approve, or handle exceptions.

The first three levels can largely support individual productivity. The fourth level is where organizational transformation begins.

At this stage, the central question is no longer, “Who in the company uses AI?” Everyone will eventually use it in some form, just as everyone now uses the internet.

The more useful question is: “Which workflow should be rebuilt first?”

How to Build an AI-Powered Company Without Chasing Every Tool

A structured approach is more valuable than a collection of disconnected AI subscriptions.

At RedTrack, the path developed gradually. First came experimentation: encouraging people to use AI tools, gain experience, and build the habit of working with them. The next step was a more formal AI approach for the company.

This included selecting a primary large language model for shared use. The choice was not necessarily about which model was universally “best.” It was about usability and integrations with the company’s existing data sources, such as Slack, Notion, and Gmail.

Individual teams can still use specialized tools where they make sense. For example, one tool may be better for presentations and another better for research or operational workflows. The goal is not rigid standardization for its own sake. It is avoiding unnecessary fragmentation while giving teams practical tools.

To prioritize workflow transformation, companies can look for areas with a large share of processing work. Then they can rebuild one process at a time.

A practical sequence looks like this:

1. Identify a workflow with repetitive, time-intensive processing work.
2. Map how the work is currently completed.
3. Decide what AI can initiate, process, summarize, or prepare.
4. Define where human validation and judgment are required.
5. Test the redesigned workflow.
6. Expand only after the process is working.

Rather than declaring that an entire department must “use AI,” the company can improve one process, then the next, then the next.

Why AI Champions Matter

Workflow redesign works best when it includes people who understand the job from the inside.

In a support team, for example, the people performing the work every day are the best source of information about which processes involve the most repetitive processing. They can identify where AI can genuinely help and where automation would create more problems than it solves.

These people can become AI champions within their teams. Their role is not necessarily to build complex agents themselves. It is to help select the right workflow, support implementation, and help their peers adopt the new process.

This matters because a workflow change only works when the people involved actually use it. If one person refuses to use the shared process, that person can become a bottleneck for the entire team.

AI adoption is therefore not only a technical challenge. It is also an operational and cultural one.

Not Everyone Needs to Manage AI Agents

As AI agents become more common, a new kind of hype has emerged: the idea that every employee must become an “agent manager.”

That is not necessary.

Using an AI-supported process and managing an AI agent are different responsibilities. Building the agent is another level of complexity again.

Most people simply need to do their jobs effectively using the best available tools and workflows. They do not need to manage every piece of technology involved.

A useful comparison comes from the shift from typewriters to word processors and printers. When a better way to print addresses became available, the team did not need to become managers of printers. They simply used a better process.

The same principle applies to AI. A person does not need to be an agent manager any more than a driver needs to be called the manager of a car.

Some roles will involve creating, configuring, and managing agents. But for most employees, the expectation should be simpler: learn to use the AI-supported workflow efficiently.

LinkedIn AI Hype Is Not the Same as Business Reality

Much of the AI conversation on professional social platforms rewards extreme claims. Posts about replacing entire teams, managing dozens of agents, or instantly transforming a business often perform well because they trigger anxiety, excitement, or curiosity.

But popularity does not make those claims representative of what is actually happening inside sustainable businesses.

Social platforms increasingly reward content designed for algorithms rather than content designed to offer careful, useful insight. As a result, many people feel pressure to make AI adoption look more dramatic than it is.

The reality is usually less theatrical and more practical:

• Teams are experimenting with tools.
• Companies are identifying repetitive work.
• Workflows are being rebuilt gradually.
• Humans are still accountable for important outcomes.
• AI creates gains when it is attached to real data and real processes.

The companies that make progress will likely be those that focus less on announcing AI transformation and more on quietly improving how work gets done.

Should SaaS Founders Raise Funding Because of AI?

AI has created a new pressure around fundraising. Even bootstrapped founders may feel that venture capital is necessary to afford models, infrastructure, visibility, and product development.

But funding should not be the default response to a technological shift.

The first question for any founder is: What are we building, and why?

This answer needs to be honest and revisited regularly. A company’s broad direction may remain stable while tactics change significantly over time. Founding teams should revisit their vision at regular intervals and ask whether the original assumptions still hold.

Once the vision is clear, the next question becomes: How would capital help the company reach that vision?

Only then does it make sense to evaluate the appropriate source of capital. Different investors have different expectations:

• Some seek hypergrowth.
• Some prioritize profitability.
• Some focus on specific sectors or niches.
• Some are industry-agnostic.
• Some are better suited to growth-stage businesses than early-stage startups.

Without clarity on what the business is building and how the money will be used, fundraising can become an end in itself rather than a strategic tool.

AI is an important technology, but it is not the first major shift in business and it will not be the last. The internet disrupted many companies, but it did not eliminate every business that was not born online. Businesses continue to succeed when they understand the value they create, the customers they serve, and the resources they need to grow.

The question is not whether AI makes funding fashionable. The question is whether funding is genuinely required to achieve the company’s goals.