AI is expanding what people can produce across organizational boundaries, but the real challenge is no longer simply doing the work. It is knowing whether the work is actually good enough to trust.

You can now ask an AI assistant to write code, analyze data, create a product specification, draft a contract, design a presentation, or reshape a business process in minutes. That sounds like a straightforward productivity breakthrough.

It is also creating a more subtle problem.

Your ability to produce something can now grow faster than your ability to evaluate it.

That gap matters because expertise has traditionally provided a natural safety mechanism. A person who cannot perform every task in an organization can still recognize strong work, challenge weak assumptions, and know when an expert needs to be involved.

AI is weakening that boundary.

For its first CTV campaign, Jennifer Aniston’s DTC haircare brand LolaVie had a few non-negotiables. The campaign had to be simple. It had to demonstrate measurable impact. And it had to be full-funnel.

LolaVie used Roku Ads Manager to test and optimize creatives — reaching millions of potential customers at all stages of their purchase journeys. Roku Ads Manager helped the brand convey LolaVie’s playful voice while helping drive omnichannel sales across both ecommerce and retail touchpoints.

The campaign included an Action Ad overlay that let viewers shop directly from their TVs by clicking OK on their Roku remote. This guided them to the website to buy LolaVie products.

Discover how Roku Ads Manager helped LolaVie drive big sales and customer growth with self-serve TV ads.

The DTC beauty category is crowded. To break through, Jennifer Aniston’s brand LolaVie, worked with Roku Ads Manager to easily set up, test, and optimize CTV ad creatives. The campaign helped drive a big lift in sales and customer growth, helping LolaVie break through in the crowded beauty category.

Learn more.

The result is a workplace where more people can participate in more types of work, but where the question of who can actually judge the quality of that work becomes increasingly important.

For most of your career, professional roles have been defined partly by what you could personally do.

A product manager did not need to be a software engineer. A designer did not need to understand every technical implementation detail. An engineer was not expected to be an expert in market segmentation or product positioning.

That did not mean these people operated in isolation.

A strong product manager could evaluate a proposed technical solution well enough to challenge it when it created a poor customer experience. An engineer could question a product requirement when it introduced unnecessary complexity. A designer could identify when a technical decision compromised usability.

In other words, people traditionally had a larger ability to judge than to perform.

AI changes the equation.

A product manager with an AI coding assistant can now produce working software without years of programming experience. A designer can use AI to analyze customer feedback or generate product concepts. An engineer can ask an AI system to develop positioning ideas, summarize research, or create a first version of a product specification.

That is useful because not every task requires expert-level execution.

The problem begins when producing something becomes easier than understanding whether it is correct.

Tip: Use AI to expand the range of work you can attempt, but treat unfamiliar output as something that needs additional validation rather than automatic trust.

Imagine someone who has never worked deeply with security suddenly generating production code with AI.

The code may compile.

It may even work during initial testing.

But that person may not recognize a subtle authentication weakness, an unsafe dependency, an insecure data flow, or an architectural decision that creates problems later.

The AI has increased the person's doing capability without proportionally increasing their judgment capability.

That is the dangerous zone.

There is a safer zone, too. If you already understand the underlying discipline, AI can dramatically accelerate your work because you can inspect the output and identify what needs improvement.

A developer using AI to generate a familiar piece of code can usually spot obvious mistakes. A marketer using AI to rewrite copy can recognize when the language sounds unnatural or fails to match the audience. A product leader can ask AI to improve a roadmap while still applying knowledge about customers, strategy, and organizational constraints.

The problem is not AI-generated work itself.

The problem is AI-generated work that looks credible to someone who lacks the expertise to challenge it.

That distinction becomes increasingly important as AI systems become better at producing polished answers. A convincing output can make a weak decision harder to recognize because presentation quality and actual quality are not the same thing.

Tip: Before asking AI to perform an unfamiliar task, ask what expertise would normally be required to evaluate the result. That tells you where human review becomes especially important.

There is a fascinating consequence to this shift.

AI does not merely make specialists faster. It allows people to temporarily cross professional boundaries.

That can be extremely valuable.

A product manager can prototype an internal tool instead of waiting weeks for engineering capacity. An engineer can investigate customer segments before bringing a product proposal forward. A designer can explore technical possibilities without requiring an engineer for every experiment.

The organization becomes more flexible because fewer tasks are trapped inside a single department.

But there is a cost.

When everyone can produce work across disciplines, everyone can also begin judging work outside their expertise.

That creates the possibility of a workplace where every function starts reviewing, modifying, and challenging every other function with AI assistance.

A marketing leader can ask an AI system to critique a product roadmap. A development team can ask an AI model to challenge product priorities. A manager can have an AI assistant rewrite another team's objectives. Each person may arrive with an apparently rational argument supported by an AI-generated analysis.

The question becomes much harder:

Who actually owns the decision?

AI can provide another opinion, but an opinion is not the same thing as accountability.

If an AI-assisted recommendation turns out to be wrong, responsibility still belongs to the people and organization that accepted it.

Tip: Keep decision ownership attached to humans even when AI participates heavily in research, analysis, drafting, and critique.

There is an important upside here.

AI does not have to be the thing that makes the judgment gap worse. It can also help close it.

Suppose you are reviewing work in your own area of expertise. AI can act as another reviewer and challenge assumptions you might have missed.

That is particularly useful because human review has limits.

When people repeatedly review AI-generated material, the process can become monotonous. The more suggestions that look reasonable, the easier it becomes to approve them without carefully examining each one.

This creates what can broadly be described as approval fatigue.

Automated review can help by examining large volumes of generated work for patterns that deserve human attention. It can act as a filter rather than replacing human accountability.

But this works best when the human already understands the domain.

An experienced engineer can use AI to challenge an implementation because the engineer knows what questions to ask. A legal professional can use AI to identify clauses worth investigating because they understand the consequences. A product leader can ask AI to attack a strategy because they understand the business context behind the decision.

The AI becomes a second set of eyes.

It does not become the final authority.

Tip: Use AI reviewers to challenge your thinking, not merely confirm it. Ask for failure modes, contradictory evidence, hidden assumptions, and reasons the proposed solution could be wrong.

This is where the situation becomes uncomfortable.

AI can help someone discover what they do not know.

But it can also make someone feel like they know more than they actually do.

A confident explanation can create the appearance of expertise without providing the underlying understanding required to evaluate it.

That is especially dangerous in areas where mistakes are difficult to detect.

Someone unfamiliar with software security might accept an AI explanation about authentication because it sounds technically sophisticated. Someone without accounting expertise might accept a generated financial model because the formulas appear reasonable. Someone unfamiliar with product discovery might accept an AI-generated prioritization framework because it produces a clean table and convincing language.

The output can be polished while the reasoning remains flawed.

This is why cognitive biases become relevant. People are not perfectly calibrated judges of their own competence, particularly in areas where they lack enough knowledge to recognize what they are missing.

AI can therefore have two opposite effects.

It can reduce the competence gap by giving people useful access to expertise.

Or it can increase the judgment gap by giving people convincing answers they are not qualified to evaluate.

The same technology can do both.

Tip: The less familiar you are with a domain, the more you should treat AI output as a starting hypothesis rather than an answer.

With Wati, every message across Instagram DM, Messenger, WhatsApp, SMS, RCS, and web chat gets an instant AI-powered response. Automations qualify leads, route conversations, and log everything automatically, so your team only steps in when it’s time to close.

See Wati In Action

There is another organizational problem hiding underneath all of this.

If AI allows hundreds of employees to generate work outside their specialties, somebody still has to review the important outputs.

That often means the organization's experts.

Developers may suddenly receive large amounts of AI-generated code from colleagues who previously would never have touched the codebase. Designers may be asked to review AI-generated interfaces. Analysts may need to validate machine-generated models. Legal teams may become the final checkpoint for contracts produced elsewhere.

The technology makes production cheaper, but it does not necessarily make expert judgment cheaper.

That can create a new bottleneck.

An organization might celebrate how quickly AI allows employees to generate material while quietly overwhelming the small group of people capable of validating it.

This is why simply telling everyone to "use AI" is not a complete operating model.

The organization also needs to decide which work requires expert review, what level of review is appropriate, and how much review capacity actually exists.

Tip: Identify high-risk categories of AI-generated work and establish explicit review paths before the volume of generated output becomes a problem.

One tempting response is to say that everyone now needs to become a generalist.

A product manager should learn programming. Engineers should learn product management. Designers should understand analytics. Everyone should learn security, finance, legal basics, and data science.

Some cross-functional knowledge is absolutely valuable.

But expecting every person to become competent in every discipline is unrealistic.

Deep expertise takes time.

A better goal is often judgment literacy.

You do not necessarily need to become a senior engineer to understand that production code should go through security review. You do not need to become an accountant to recognize that a financial model requires certain assumptions to be validated. You do not need to become a lawyer to know that an AI-generated contract should not be treated as legally reliable simply because it sounds professional.

The objective is to understand enough about a domain to recognize its boundaries, ask intelligent questions, and know when specialist involvement is necessary.

That is a much more achievable skill.

Tip: Build enough cross-functional knowledge to recognize risk and ask better questions, rather than trying to master every discipline AI can now expose you to.

The most productive relationship with AI may not be "AI does the work and humans approve it."

It may be closer to AI produces, AI critiques, humans decide.

That distinction preserves an important organizational principle: accountability should remain with people who understand the consequences of the decision.

For example, AI can generate several roadmap alternatives, identify trade-offs, estimate implementation considerations, and challenge assumptions.

But deciding which roadmap aligns with the company's customers and strategy still requires human judgment.

AI can generate code, identify possible vulnerabilities, compare implementation approaches, and explain technical trade-offs.

But determining whether a particular change is safe for a production system remains a human responsibility.

The goal is not to keep AI away from important work.

It is to make sure that increasing AI participation does not accidentally eliminate meaningful human judgment.

AI is making competence more accessible.

That is one of its most powerful effects.

But accessibility is not the same as expertise.

You can now produce something that previously required years of experience. That does not automatically mean you have acquired the experience required to recognize every subtle problem in what you produced.

That creates a new professional skill that may become increasingly valuable: knowing the boundary between what you can produce and what you can responsibly judge.

The people who understand that boundary will be better positioned to use AI aggressively without becoming careless.

They will know when to experiment independently, when to ask AI for another perspective, when to involve a specialist, and when a decision is too consequential to outsource to a chatbot.

And that may become one of the defining differences between simply having access to powerful AI and actually knowing how to work with it.

Tip: Before trusting an AI-assisted result, ask one simple question: What would I need to know to confidently prove this is correct? If the answer is outside your expertise, bring in someone who has that judgment.

What’s your next spark? A new platform engineering skill? A bold pitch? A team ready to rise? Share your ideas or challenges at Tiny Big Spark. Let’s build your pyramid—together.

That’s it!

Keep innovating and stay inspired!

If you think your colleagues and friends would find this content valuable, we’d love it if you shared our newsletter with them!

As the world becomes increasingly digital, this question will be on the minds of millions of people seeking new income streams in 2026.

The answer is—Absolutely!

Thank you for taking the time to read today’s email! Your support allows me to send out this newsletter for free every day. 

 What do you think for today’s episode? Please provide your feedback in the poll below.

Share the newsletter with your friends and colleagues if you find it valuable.

Disclaimer: The "Tiny Big Spark" newsletter is for informational and educational purposes only, not a substitute for professional advice, including financial, legal, medical, or technical. We strive for accuracy but make no guarantees about the completeness or reliability of the information provided. Any reliance on this information is at your own risk. The views expressed are those of the authors and do not reflect any organization's official position. This newsletter may link to external sites we don't control; we do not endorse their content. We are not liable for any losses or damages from using this information.