How game mechanics are changing B2B onboarding in 2026 | Karel Papik @ Product Fruits

Karel did not come to SaaS through the usual product or marketing route. He spent roughly two decades in gaming, building and producing titles across PC and console, from hardcore games to casual games aimed at a very specific audience. That experience shaped the way he thinks about user behavior, first-hour engagement, and habit formation.

At Product Fruits, that background became unexpectedly useful. His core argument is simple: games have had to master adoption far earlier and far better than most B2B software. If a game cannot pull someone in quickly, it loses them. B2B products, by contrast, have often tolerated weak onboarding because users “need” the software anyway. That gap is now closing fast.

From games to digital adoption

Karel’s path into Product Fruits started after several years in Vietnam, where he worked with startups in an advisory capacity. During that time, he encountered Product Fruits inside another company’s product and was struck by its potential. After meeting Slava Shalov, the original force behind the company, he decided to join. The fit was immediate.

Today, Product Fruits helps software companies add an adoption layer on top of their applications. In practice, that means step-by-step tours, hints, announcements, guidance flows, invitations, and contextual help that sit between the product and the end user.

But Karel increasingly frames the category as adoption rather than onboarding. Onboarding is only the start. Real product success depends on helping users discover value continuously, navigate change, learn features in context, and get answers exactly when they need them.

Why games are still ahead of B2B software

Gaming taught Karel a discipline that many B2B teams still lack: the first session matters enormously. In his gaming businesses, the business model often depended on whether a player became engaged within the first hour. Everything had to be designed around that moment:

• How quickly users understand the core value
• How fast they experience momentum
• How carefully friction is introduced
• How rewards are timed
• How the product earns continued attention

That mindset translates cleanly into SaaS. Most teams are not trying to make products addictive in a shallow sense. They are trying to help users adopt the product, reach outcomes faster, and build confidence. The useful lesson from games is not gimmicky gamification. It is behavioral design.

A gaming mechanic that works surprisingly well in SaaS

One of Karel’s favorite examples comes from free-to-play design. Imagine a game where a player discovers a powerful item and gets to enjoy it immediately. After ten minutes of fun, the game presents a choice: keep using it by paying, or give it up.

That approach is much stronger than asking for payment upfront before the person has experienced the value.

In B2B SaaS, the same logic can apply. Instead of asking someone whether they want to buy a capability they do not yet understand, let them use it in context first. If the feature proves useful at the right moment, conversion becomes more natural.

This is a subtle but important shift:

• Do not sell abstract functionality.
• Let people experience a meaningful outcome first.
• Present the upgrade decision when the value is already obvious.

Karel also points out that some gaming mechanics became so effective that regulators eventually restricted them. His point is not that SaaS should copy exploitative systems. It is that games have spent years refining adoption loops that software companies are only beginning to study seriously.

Why Product Fruits went all-in on AI

About a year before this conversation, Product Fruits reached an uncomfortable conclusion: the market was changing fast, and an “analog” digital adoption platform would not be enough for long.

The team initially reacted with fear. AI seemed like the kind of shift that could wipe out existing products in the category. Instead of treating that as a reason to retreat, they took it as a reason to rebuild.

Karel presented investors with a bold plan: slow down development of the traditional product and focus on building a new AI-native version to bring to market within months.

That decision could easily have looked risky from the outside. Existing customers still wanted improvements to the current product. Monthly recurring revenue might have stalled. The roadmap was being disrupted on purpose.

Instead, investors responded almost immediately with support and additional capital. Their logic was straightforward: they wanted to back a company behaving like a winner, not one merely trying to survive.

Importantly, the business did not collapse during the transition. Revenue kept growing while the AI product was being built. By late 2025, Product Fruits had started releasing its new AI capabilities, and by early 2026 Karel believed they had become the most advanced AI-first digital adoption platform in their category.

From segments to true personalization

Traditional onboarding personalization usually relies on segmentation. Teams define user groups such as admins, end users, small accounts, enterprise accounts, or trial users, then try to build separate flows for each.

That works to a point. But it becomes unwieldy quickly.

As products grow, teams face too many combinations of roles, goals, pages, use cases, pricing tiers, and lifecycle stages. The result is complexity that is hard to manage and even harder to maintain.

Karel sees AI changing this model entirely. Instead of building countless predefined paths, the product can adapt to the person in real time.

At Product Fruits, this starts with a discovery conversation. When someone signs up for a product using the platform, the AI can ask practical questions such as:

• What are you trying to achieve?
• What is your use case?
• What tools have you used before?
• How large is your team?

This resembles a smart qualification conversation, but it happens inside the product experience. Based on those answers, the system can personalize guidance immediately.

That is a meaningful leap beyond segmentation. It does not just place a person into a static bucket. It adapts to context, intent, and prior behavior.

How Elvin works inside the product

One of Product Fruits’ AI components is Elvin, which acts as an in-product assistant. The idea is that the adoption layer should feel inseparable from the application itself. Users should not have to think about where the app ends and the support layer begins.

Elvin can do several things at once:

• Understand the current page and screen context
• Remember previous conversations
• Answer questions using product-specific knowledge
• Guide someone visually by pointing to the relevant part of the interface
• Handle voice interactions as well as typed prompts

If someone asks for integrations, for example, the assistant can respond with awareness of what has already been discussed, infer likely intent, identify where the integration settings live, explain whether a feature requires an upgrade, and tie that answer back to the user’s goals.

That makes the support experience far more dynamic than a static help center article or linear product tour.

AI for users is only half the story

One of the more interesting ideas in Product Fruits is that AI should not only help users. It should also help the people managing adoption.

Karel described using Product Fruits on top of Product Fruits itself. In other words, the company relies on its own platform to understand how customers experience the product.

That creates a second layer of value for admins and product teams. Instead of manually reviewing support patterns, they can ask the AI questions such as:

• Where are users struggling the most?
• What has caused confusion over the past few weeks?
• Which product changes are creating friction?

The AI can read through conversations, identify patterns, and surface likely root causes. In one example, it flagged confusion around new pricing and suggested that the team investigate that area more closely.

This turns AI from a front-end helper into an operational feedback system.

The annotation method: teaching AI your product

There is a catch, of course. AI is only as useful as the information it has access to.

Karel is blunt about this. If a company’s documentation is weak, the AI cannot magically invent clear, reliable answers. Human support teams may still know the product well through tribal knowledge, but if that knowledge is not captured, the system has little to work with.

That challenge led Product Fruits to build what it calls annotation.

Annotation is designed around a practical assumption: most product managers can describe their application well, even if they are not experts in onboarding design.

So instead of asking teams to craft complex guidance flows from scratch, Product Fruits lets them describe the product in plain terms:

• What the application does
• What each area is for
• What specific buttons and actions do
• What the key use cases are
• Who the product is meant for

That descriptive layer is then combined with other sources such as CRM data and the knowledge base. From there, the AI can generate relevant tours and contextual guidance on the fly.

This is an important distinction in Karel’s thinking. Product teams do not necessarily know how to design great onboarding systems. But they do know how to explain their own products. The platform should bridge that gap for them.

Why documentation quality still matters

Despite the excitement around AI, Product Fruits does not pretend the technology erases operational discipline. If the knowledge base is messy, inconsistent, or incomplete, answers will be weaker. If product information is scattered informally across conversations, support threads, and internal lore, the AI will struggle.

That means companies still need to invest in foundational clarity. AI can amplify strong inputs. It cannot reliably rescue poor ones.

Data privacy and separation

As more teams feed product information into AI systems, trust becomes a central concern. Karel’s position is that each customer’s Product Fruits environment remains its own. Data is not shared across products or mixed together for other customers’ benefit.

He also notes that, in practice, cross-product sharing would not even make much sense. The details of adoption are highly specific. What works in one application may not transfer cleanly to another because the workflows, language, and user intent are different.

So the privacy model is not only about safety. It is also aligned with how product adoption actually works: at a very granular, product-specific level.

The real challenge with AI is uncertainty

Interestingly, Karel does not think the biggest barrier is data privacy. He thinks it is uncertainty.

Most software teams are used to deterministic systems. They expect fixed outputs, predictable logic, and exact repeatability. AI behaves differently. Responses can vary. The experience is more flexible, more contextual, and in that sense more human.

That creates discomfort. Companies often want the AI to answer every question in exactly one approved way. Product Fruits can steer and improve responses, but Karel does not believe the right goal is to force rigid script-like behavior.

Instead, the product lets teams review conversations and guide the assistant over time. Admins can mark answers as good, flag weaker ones, and suggest how future responses should improve. The system can be tuned, but not reduced to copy-paste automation.

This is another major mindset change for software teams: moving from full control toward guided adaptation.

Why pricing had to change for AI

Uncertainty also showed up in pricing.

AI conversations create variable costs in the background, but many SaaS buyers still want fixed annual numbers that can pass procurement cleanly. “Maybe” is not a budget line item.

Product Fruits initially struggled with this mismatch. If usage depends on how often end users interact with AI, the company cannot know exact future costs in advance.

But customers still needed predictable pricing. So Product Fruits changed its approach and began taking more of that risk on itself, allowing prepayment models that give buyers a clearer number while the vendor absorbs more of the variability.

That is a useful reminder that AI product design is not only about the interface. It affects packaging, billing, and internal economics too.

Growth strategy: PPC over content and no outbound

On go-to-market, Product Fruits made another contrarian choice. From the beginning, the team tried to avoid short-lived founder-led tactics that would later have to be replaced. Instead, they wanted a growth model they could sustain as they scaled.

The channel they chose was PPC.

Karel preferred paid acquisition because it produced faster feedback loops and clearer control over spending. Content, by contrast, often came with long delays, vague promises, and no certainty that results would arrive.

That decision also reflected timing. In a category already crowded with well-funded players producing large amounts of educational content, Product Fruits did not believe it could outproduce teams with far bigger editorial operations.

So rather than compete head-on in that format, they focused on inbound demand generated through paid channels. The sales team remains relatively small and handles inbound only. They do not do outbound.

Karel’s reasoning on content is especially pragmatic. Even when educational content succeeds in creating category awareness, a buyer will often still compare multiple vendors before choosing. In that scenario, the company doing the educational heavy lifting does not always capture the sale. Product Fruits is comfortable letting larger competitors warm the market and then competing at the point of decision.

Does AI search change that equation?

The team has also explored visibility inside AI-driven search experiences such as ChatGPT and Google’s AI surfaces. Karel’s conclusion so far is that the fundamentals are not dramatically different from SEO.

His simplified view is that if a company is discoverable in traditional search, it is likely to show up in AI search as well. The mechanics may evolve, but the underlying signals still look similar enough that he does not treat AI discovery as an entirely separate discipline yet.

That perspective is notably less hype-driven than much of the current discussion around ranking in AI tools.

Product Fruits’ biggest win

When asked about the company’s biggest win, Karel points to the decision to rebuild around AI at the core rather than layering superficial features on top.

In his view, small additions like summaries, copy helpers, or image generation are not enough. The more meaningful move was to rethink the platform from the ground up for a different era of adoption.

That required conviction, timing, and willingness to disrupt an already functioning business. But he sees that boldness as the turning point.

And its biggest failure

His clearest failure was following the crowd into content more aggressively than his instincts supported.

Although Product Fruits still has a blog and still produces some content, the attempt to invest more heavily in content-led growth did not pay off. Karel sees that as partly a strategic mismatch and partly a realism problem: the company could not reasonably compete with larger teams dedicated to publishing at scale in the same space.

The lesson was not that content never works. It was that copying a common playbook without a real edge is usually a mistake.

The founder lesson behind all of this: critical thinking

The final theme running through Karel’s thinking is critical thinking. He treats it almost like both a superpower and a burden.

He is skeptical of popular startup narratives, skeptical of crowded advice loops, and skeptical of the endless stream of polished certainty that fills founder circles and social feeds. His view is that many founders chase silver bullets because they want safety, predictability, and social proof.

But startups rarely reward that behavior.

By definition, a startup is making a risky bet with a low probability of success and potentially large upside. Following consensus too closely may feel safer, but it often leads to average decisions in a market that rewards uncommon insight.

That also shapes how he thinks about customer feedback.

Karel is not arguing that founders should ignore customers entirely. Product Fruits talks to customers constantly and uses early access releases heavily. But he draws a line between learning from present-day friction and outsourcing the future of the product.

Customers understand their own needs within their own context. They do not necessarily understand what the category will look like in two years. If product adoption is your domain, then it is your job to build that point of view rather than expecting customers to do it for you.

In short:

• Listen closely to today’s problems.
• Do not expect customers to define tomorrow’s category.
• Develop expertise deep enough to hold an independent view.
• Be willing to act against consensus when the evidence points that way.

Karel also acknowledges the cost of that approach. Independent thinking can be lonely. People naturally seek safety in shared opinion, and consensus is often mistaken for truth. But in startups, comfort and correctness are not the same thing.