From Co-pilot to Co-founder. The 2026 Startup Playbook | BelTech 4.0, Warsaw

Summary: From Co-pilot to Co-founder — The 2026 Startup Playbook

The panel’s central argument is that AI has made it much cheaper and faster to build early products, enabling solo founders and very small teams. But building a prototype is only part of building a company: industry knowledge, customer relationships, sales, distribution, judgment, and human accountability still matter. The speakers encouraged founders to use AI from the start, test ideas quickly, and scale only when the evidence supports it.

Startup Playbook and Operating Tactics

• Start with domain expertise, not a large founding team. AI agents can increasingly handle technical, marketing, recruiting, and operational tasks. The panel argued that a founder who deeply understands an industry, its customers, and its problems may be more valuable at the outset than assembling conventional technical and marketing co-founders.
• Use AI as core infrastructure. One speaker compared starting a company without AI to starting one without a computer. AI can let one person work across multiple functions, though human co-founders can still provide accountability, emotional support, and complementary expertise.
• Prototype first, then decide what to build around it. The suggested approach was to spend a weekend building a basic version, put it in front of potential customers, and test demand—potentially with a small Google Ads campaign. Don’t wait for certainty or a perfect competitive analysis before trying.
• Bootstrap early; raise for scaling. Since prototypes and lean teams cost less to create, founders may not need early capital just to get started. Funding may be more useful when the business has evidence of demand and needs to scale distribution or operations.
• Treat make-versus-buy as a strategic choice. AI makes custom software easier to create, but founders may still prefer to buy tools or services that handle infrastructure, privacy, security, or maintenance so they can focus on their core business.
• Measure AI by business outcomes, not agent counts. The number of agents or functions is not a useful target by itself. The panel raised the need to connect AI usage and spending to productivity, revenue, profit, and the company’s P&L, but noted that many teams do not yet have a consistent measurement system.
• Test distribution earlier. As more people can build similar products, competition rises and customer acquisition becomes more important. The panel stressed that distribution, sales, and customer access can be more defensible than the product itself.
• Rethink early-stage investing. One proposal was for investors to provide AI credits, run hackathons, identify people who can build quickly, and then support promising teams through incubators or accelerators—not just write small checks.

Competitive Advantage in an AI-Heavy Market

The panel said generic AI “wrappers” are vulnerable, especially if a model provider can reproduce their features. Potential sources of differentiation include:

• Proprietary or hard-to-access data, particularly data accumulated through real customer use.
• Industry expertise and a detailed understanding of customer problems.
• Sales capability, distribution, and trusted customer relationships.
• Brand, product taste, curation, and customer experience.
• Regulatory or licensing barriers, such as in financial services or healthcare.
• Connections to the physical world, where AI alone cannot easily replicate the work.
• Cybersecurity and infrastructure, areas where growing AI-enabled attack capabilities may increase demand.

Examples and Case Studies

• Mara Radio: Its founder built an initial personal-radio MVP locally on a Mac, despite the prototype’s rough technical setup. After showing it to a technical co-founder, the team recognized that the prototype would not scale without deeper engineering expertise. The example illustrates using a founder-built prototype to clarify the product before adding specialist capability.
• OneSoil: The company’s satellite imagery is publicly available, but it also trains models on farmer data collected through its app. The panel discussed soil sampling as a possible way to build an even more distinctive dataset, since it requires physical samples that cannot be obtained from satellite imagery alone.
• Doctrina: The company used an AI colleague named Fred to take on operational tasks, including routing requests, reminding employees, and tracking progress. The example showed how an AI agent can become part of day-to-day management workflows.
• One-person company example: A founder reportedly built a website, technology, and marketing funnel alone. The panel used this to illustrate how AI can enable a single founder to perform work that previously required a team.
• DIY community: One speaker described launching a construction and home-improvement forum before discovering that another similar forum already existed. The lesson was that acting quickly can reveal opportunities that overthinking might prevent a founder from pursuing.

Business Models, SaaS, and Management Questions

• B2B SaaS faces pressure from custom AI solutions. Panelists said AI integrations may replace some existing software, particularly where companies can build solutions tailored to their own workflows. However, they did not conclude that SaaS is finished.
• Per-seat pricing may become less suitable. The panel questioned whether traditional seat-based pricing will persist as AI agents perform work. Pricing by agent or another model was raised, but no clear replacement model was established.
• Implementation services may be an opportunity. Many small businesses want to use AI but lack the knowledge to deploy it. The panel compared this opportunity to earlier waves of website creation and IT support.
• Security, infrastructure, and trusted vendors may remain valuable. Businesses may continue buying software and services where they need reliable security, privacy, infrastructure, or support rather than building everything themselves.

Metrics and Figures Mentioned

• Pre-seed team size: A panelist cited Crunchbase statistics suggesting that the average number of employees at pre-seed startups fell from roughly four to three over the prior year.
• AI spending: One founder said their company had spent about $6,000 in a month on AI tools and reacted after noticing the cost was too high.
• Enterprise AI usage: A panelist cited a company with about 450 employees spending roughly $6,000–$7,000 per engineer per month on tokens—approximately comparable to an engineer’s monthly salary. This was presented as an example, not as a general benchmark.
• Subscription usage: A startup of roughly seven or eight people was said to be using 91 Claude Max subscriptions through a managed setup.
• Funding figures: The introduction cited Gero as having raised about $34 million in total, after confusion over other figures. It also described Zubr Capital private-equity checks of $5 million–$15 million.
• SaaS valuation anecdote: A speaker said Miro had been valued at around $17.5 billion at its peak and later acquired for about $1.6 billion. This was raised as an example of changing software-market sentiment.
• Venture returns: The panel discussed expectations potentially moving from roughly 10 years toward four or five years, but did not present this as a settled industry rule.

Bottom Line

The practical recommendation was to build early, use AI to keep the initial team lean, and test customer demand and distribution quickly. Founders should not treat the number of agents or lines of code as success metrics; they should focus on customer value, revenue, productivity, and defensible advantages. Human expertise and relationships remain important even as AI takes on more execution.

Presenters and Sources

• Sergey Osipov — moderator; identified in the introduction as a co-founder of PlaySense and Playsee AI, and an early-stage investor who also invests with Zubr Capital.
• Yury Melnichek — serial entrepreneur; associated in the introduction with OneSoil, Doctrina, Rainbow, and Gero.
• Alexander Chekan — venture builder and entrepreneur; the subtitles also render his name as “Alex Chernenko.” The introduction associates him with the personal-radio startup Mara Radio.
• Other sources cited in discussion: Crunchbase statistics; company and founder anecdotes from OneSoil, Doctrina, and other AI-focused startups.